pankajmathur

"Obsessed with Open Source GenAI's potential? So am I ! Let's Contribute together 🚀 https://www.linkedin.com/in/pankajam " 1) https://huggingface.co/TheBloke/orcaminiv313B-GGML 2) https://huggingface.co/TheBloke/orcaminiv313B-GPTQ This model is bound by the license & usage restrictions of the original Llama-2 model. And comes with no warranty or gurantees of any kind. We evaluated orcaminiv313b on a wide range of tasks using Language Model Evaluation Harness from EleutherAI. Here are the results on metrics used by HuggingFaceH4 Open LLM Leaderboard ||||| |:------:|:--------:|:-------:|:--------:| |Task|Metric|Value|Stderr| |arcchallenge|accnorm|0.6314|0.0141| |hellaswag|accnorm|0.8242|0.0038| |mmlu|accnorm|0.5637|0.0351| |truthfulqamc|mc2|0.5127|0.0157| |Total Average|-|0.6329877193|| Below shows a code example on how to use this model While this model aims for accuracy, it can occasionally produce inaccurate or misleading results. Despite diligent efforts in refining the pretraining data, there remains a possibility for the generation of inappropriate, biased, or offensive content. Exercise caution and cross-check information when necessary. Open LLM Leaderboard Evaluation Results Detailed results can be found here | Metric | Value | |-----------------------|---------------------------| | Avg. | 52.23 | | ARC (25-shot) | 63.14 | | HellaSwag (10-shot) | 82.35 | | MMLU (5-shot) | 56.52 | | TruthfulQA (0-shot) | 51.81 | | Winogrande (5-shot) | 76.48 | | GSM8K (5-shot) | 13.12 | | DROP (3-shot) | 22.23 | Open LLM Leaderboard Evaluation Results Detailed results can be found here | Metric |Value| |---------------------------------|----:| |Avg. |57.24| |AI2 Reasoning Challenge (25-Shot)|63.14| |HellaSwag (10-Shot) |82.35| |MMLU (5-Shot) |56.52| |TruthfulQA (0-shot) |51.81| |Winogrande (5-shot) |76.48| |GSM8k (5-shot) |13.12| Open LLM Leaderboard Evaluation Results Detailed results can be found here | Metric |Value| |-------------------|----:| |Avg. |15.00| |IFEval (0-Shot) |28.97| |BBH (3-Shot) |25.55| |MATH Lvl 5 (4-Shot)| 1.89| |GPQA (0-shot) | 2.01| |MuSR (0-shot) |17.11| |MMLU-PRO (5-shot) |14.50|

"Obsessed with GenAI's potential? So am I ! Let's create together 🚀 https://www.linkedin.com/in/pankajam " 1) https://huggingface.co/TheBloke/orcaminiv370B-GGML 2) https://huggingface.co/TheBloke/orcaminiv370B-GPTQ This model is bound by the license & usage restrictions of the original Llama-2 model. And comes with no warranty or gurantees of any kind. We evaluated orcaminiv370b on a wide range of tasks using Language Model Evaluation Harness from EleutherAI. Here are the results on metrics used by HuggingFaceH4 Open LLM Leaderboard ||| |:------:|:--------:| |Task|Value| |ARC|0.7125| |HellaSwag|0.8785| |MMLU|0.7018| |TruthfulQA|0.6127| |Winogrande|0.8272| |GSM8K|0.4086| |DROP|0.4017| |Total Average|0.649| This model required upto 45GB GPU VRAM in 4bit so it can be loaded directly on Single RTX 6000/L40/A40/A100/H100 GPU or Double RTX 4090/L4/A10/RTX 3090/RTX A5000 So, if you have access to Machine with 45GB GPU VRAM and have installed OobaBooga Web UI on it. You can just download this model by using HF repo link directly on OobaBooga Web UI "Model" Tab/Page & Just use load-in-4bit option in it. After that go to Default Tab/Page on OobaBooga Web UI and copy paste above prompt format into Input and Enjoy! Below shows a code example on how to use this model While this model aims for accuracy, it can occasionally produce inaccurate or misleading results. Despite diligent efforts in refining the pretraining data, there remains a possibility for the generation of inappropriate, biased, or offensive content. Exercise caution and cross-check information when necessary. Open LLM Leaderboard Evaluation Results Detailed results can be found here | Metric | Value | |-----------------------|---------------------------| | Avg. | 64.9 | | ARC (25-shot) | 71.25 | | HellaSwag (10-shot) | 87.85 | | MMLU (5-shot) | 70.18 | | TruthfulQA (0-shot) | 61.27 | | Winogrande (5-shot) | 82.72 | | GSM8K (5-shot) | 40.86 | | DROP (3-shot) | 40.17 | Open LLM Leaderboard Evaluation Results Detailed results can be found here | Metric |Value| |---------------------------------|----:| |Avg. |69.02| |AI2 Reasoning Challenge (25-Shot)|71.25| |HellaSwag (10-Shot) |87.85| |MMLU (5-Shot) |70.18| |TruthfulQA (0-shot) |61.27| |Winogrande (5-shot) |82.72| |GSM8k (5-shot) |40.86| Open LLM Leaderboard Evaluation Results Detailed results can be found here | Metric |Value| |-------------------|----:| |Avg. |25.26| |IFEval (0-Shot) |40.15| |BBH (3-Shot) |42.98| |MATH Lvl 5 (4-Shot)| 3.63| |GPQA (0-shot) | 9.06| |MuSR (0-shot) |25.12| |MMLU-PRO (5-shot) |30.64|

OrcaMiniv96Llama-3.2-3B-Instruct-GGUF is static quant version of orcaminiv963B-Instruct "Obsessed with Open Source GenAI's potential? So am I ! Let's Contribute together 🚀 https://www.linkedin.com/in/pankajam " NOTICE By providing proper credit and attribution, you are granted permission to use this model as a foundational base for further Full fine tuning, DPO, PPO or ORPO tuning and any kind of Merges. I actively encourage users to customize and enhance the model according to their specific needs, as this version is designed to be a comprehensive general model. Dive in and innovate! My preferred method is Ollama but you can choose other ways to use this modle locally on your PC by going to right side of this repo card and select "Use this model". Here are the steps for Ollama: 1. Install Ollama https://github.com/ollama/ollama 2. After Installing Ollama run below command on your terminal 4. Start Chatting with this model! 5. Optional: For ChatGPT like experience (or better), Install Open-Webui https://github.com/open-webui/open-webui 6. open-webui will auto connect to Ollama running this model and you can start chatting right away As part of our Responsible release approach, we followed a three-pronged strategy to managing trust & safety risks: 1. Enable developers to deploy helpful, safe and flexible experiences for their target audience and for the use cases supported by Llama 2. Protect developers against adversarial users aiming to exploit Llama capabilities to potentially cause harm 3. Provide protections for the community to help prevent the misuse of our models Approach: Llama is a foundational technology designed to be used in a variety of use cases. Examples on how Meta’s Llama models have been responsibly deployed can be found in our Community Stories webpage. Our approach is to build the most helpful models, enabling the world to benefit from the technology power, by aligning our model safety for generic use cases and addressing a standard set of harms. Developers are then in the driver’s seat to tailor safety for their use cases, defining their own policies and deploying the models with the necessary safeguards in their Llama systems. Llama 3.2 was developed following the best practices outlined in our Responsible Use Guide. Objective: Our main objectives for conducting safety fine-tuning are to provide the research community with a valuable resource for studying the robustness of safety fine-tuning, as well as to offer developers a readily available, safe, and powerful model for various applications to reduce the developer workload to deploy safe AI systems. We implemented the same set of safety mitigations as in Llama 3, and you can learn more about these in the Llama 3 paper. Fine-Tuning Data: We employ a multi-faceted approach to data collection, combining human-generated data from our vendors with synthetic data to mitigate potential safety risks. We’ve developed many large language model (LLM)-based classifiers that enable us to thoughtfully select high-quality prompts and responses, enhancing data quality control. Refusals and Tone: Building on the work we started with Llama 3, we put a great emphasis on model refusals to benign prompts as well as refusal tone. We included both borderline and adversarial prompts in our safety data strategy, and modified our safety data responses to follow tone guidelines. Safety as a System: Large language models, including Llama 3.2, are not designed to be deployed in isolation but instead should be deployed as part of an overall AI system with additional safety guardrails as required. Developers are expected to deploy system safeguards when building agentic systems. Safeguards are key to achieve the right helpfulness-safety alignment as well as mitigating safety and security risks inherent to the system and any integration of the model or system with external tools. As part of our responsible release approach, we provide the community with safeguards that developers should deploy with Llama models or other LLMs, including Llama Guard, Prompt Guard and Code Shield. All our reference implementations demos contain these safeguards by default so developers can benefit from system-level safety out-of-the-box. Technological Advancement: Llama releases usually introduce new capabilities that require specific considerations in addition to the best practices that generally apply across all Generative AI use cases. For prior release capabilities also supported by Llama 3.2, see Llama 3.1 Model Card, as the same considerations apply here as well. Constrained Environments: Llama 3.2 1B and 3B models are expected to be deployed in highly constrained environments, such as mobile devices. LLM Systems using smaller models will have a different alignment profile and safety/helpfulness tradeoff than more complex, larger systems. Developers should ensure the safety of their system meets the requirements of their use case. We recommend using lighter system safeguards for such use cases, like Llama Guard 3-1B or its mobile-optimized version. Scaled Evaluations: We built dedicated, adversarial evaluation datasets and evaluated systems composed of Llama models and Purple Llama safeguards to filter input prompt and output response. It is important to evaluate applications in context, and we recommend building dedicated evaluation dataset for your use case. Red Teaming: We conducted recurring red teaming exercises with the goal of discovering risks via adversarial prompting and we used the learnings to improve our benchmarks and safety tuning datasets. We partnered early with subject-matter experts in critical risk areas to understand the nature of these real-world harms and how such models may lead to unintended harm for society. Based on these conversations, we derived a set of adversarial goals for the red team to attempt to achieve, such as extracting harmful information or reprogramming the model to act in a potentially harmful capacity. The red team consisted of experts in cybersecurity, adversarial machine learning, responsible AI, and integrity in addition to multilingual content specialists with background in integrity issues in specific geographic markets. In addition to our safety work above, we took extra care on measuring and/or mitigating the following critical risk areas: 1\. CBRNE (Chemical, Biological, Radiological, Nuclear, and Explosive Weapons): Llama 3.2 1B and 3B models are smaller and less capable derivatives of Llama 3.1. For Llama 3.1 70B and 405B, to assess risks related to proliferation of chemical and biological weapons, we performed uplift testing designed to assess whether use of Llama 3.1 models could meaningfully increase the capabilities of malicious actors to plan or carry out attacks using these types of weapons and have determined that such testing also applies to the smaller 1B and 3B models. 2\. Child Safety: Child Safety risk assessments were conducted using a team of experts, to assess the model’s capability to produce outputs that could result in Child Safety risks and inform on any necessary and appropriate risk mitigations via fine tuning. We leveraged those expert red teaming sessions to expand the coverage of our evaluation benchmarks through Llama 3 model development. For Llama 3, we conducted new in-depth sessions using objective based methodologies to assess the model risks along multiple attack vectors including the additional languages Llama 3 is trained on. We also partnered with content specialists to perform red teaming exercises assessing potentially violating content while taking account of market specific nuances or experiences. 3\. Cyber Attacks: For Llama 3.1 405B, our cyber attack uplift study investigated whether LLMs can enhance human capabilities in hacking tasks, both in terms of skill level and speed. Our attack automation study focused on evaluating the capabilities of LLMs when used as autonomous agents in cyber offensive operations, specifically in the context of ransomware attacks. This evaluation was distinct from previous studies that considered LLMs as interactive assistants. The primary objective was to assess whether these models could effectively function as independent agents in executing complex cyber-attacks without human intervention. Because Llama 3.2’s 1B and 3B models are smaller and less capable models than Llama 3.1 405B, we broadly believe that the testing conducted for the 405B model also applies to Llama 3.2 models. Industry Partnerships: Generative AI safety requires expertise and tooling, and we believe in the strength of the open community to accelerate its progress. We are active members of open consortiums, including the AI Alliance, Partnership on AI and MLCommons, actively contributing to safety standardization and transparency. We encourage the community to adopt taxonomies like the MLCommons Proof of Concept evaluation to facilitate collaboration and transparency on safety and content evaluations. Our Purple Llama tools are open sourced for the community to use and widely distributed across ecosystem partners including cloud service providers. We encourage community contributions to our Github repository. Grants: We also set up the Llama Impact Grants program to identify and support the most compelling applications of Meta’s Llama model for societal benefit across three categories: education, climate and open innovation. The 20 finalists from the hundreds of applications can be found here. Reporting: Finally, we put in place a set of resources including an output reporting mechanism and bug bounty program to continuously improve the Llama technology with the help of the community. Values: The core values of Llama 3.2 are openness, inclusivity and helpfulness. It is meant to serve everyone, and to work for a wide range of use cases. It is thus designed to be accessible to people across many different backgrounds, experiences and perspectives. Llama 3.2 addresses users and their needs as they are, without insertion unnecessary judgment or normativity, while reflecting the understanding that even content that may appear problematic in some cases can serve valuable purposes in others. It respects the dignity and autonomy of all users, especially in terms of the values of free thought and expression that power innovation and progress. Testing: Llama 3.2 is a new technology, and like any new technology, there are risks associated with its use. Testing conducted to date has not covered, nor could it cover, all scenarios. For these reasons, as with all LLMs, Llama 3.2’s potential outputs cannot be predicted in advance, and the model may in some instances produce inaccurate, biased or other objectionable responses to user prompts. Therefore, before deploying any applications of Llama 3.2 models, developers should perform safety testing and tuning tailored to their specific applications of the model. Please refer to available resources including our Responsible Use Guide, Trust and Safety solutions, and other resources to learn more about responsible development.

"Obsessed with GenAI's potential? So am I ! Let's create together 🚀 https://www.linkedin.com/in/pankajam " Use orca-mini-3b for Free on Google Colab with T4 GPU :) An OpenLLaMa-3B model model trained on explain tuned datasets, created using Instructions and Input from WizardLM, Alpaca & Dolly-V2 datasets and applying Orca Research Paper dataset construction approaches. We build explain tuned WizardLM dataset ~70K, Alpaca dataset ~52K & Dolly-V2 dataset ~15K created using approaches from Orca Research Paper. We leverage all of the 15 system instructions provided in Orca Research Paper. to generate custom datasets, in contrast to vanilla instruction tuning approaches used by original datasets. This helps student model aka this model to learn thought process from teacher model, which is ChatGPT (gpt-3.5-turbo-0301 version). Please see below example usage how the System prompt is added before each instruction. The training configurations are provided in the table below. The training takes on 8x A100(80G) GPUs and lasts for around 4 Hours for cost of $48 using Lambda Labs We used DeepSpeed with fully sharded data parallelism, also know as ZeRO stage 3 by writing our own fine tunning scripts plus leveraging some of the model training code provided by amazing OpenAlpaca repo ||| |:-------------:|:-------------:| |batchsize|64| |trainmicrobatchsizepergpu|4| |gradientaccumulationsteps|2| |Learning rate|2e-5| |Max length|1024| |Epochs|3| |Optimizer|AdamW| Next Goals: 1) Try more data like actually using FLAN-v2, just like Orka Research Paper (I am open for suggestions) 2) Provide more options for Text generation UI. (may be https://github.com/oobabooga/text-generation-webui) 3) Provide 4bit GGML/GPTQ quantized model (may be TheBloke can help here) This model can produce factually incorrect output, and should not be relied on to produce factually accurate information. This model was trained on various public datasets. While great efforts have been taken to clean the pretraining data, it is possible that this model could generate lewd, biased or otherwise offensive outputs. The license on this model does not constitute legal advice. We are not responsible for the actions of third parties who use this model. Please cosult an attorney before using this model for commercial purposes. If you found wizardlmalpacadollyorcaopenllama3b useful in your research or applications, please kindly cite using the following BibTeX: Open LLM Leaderboard Evaluation Results Detailed results can be found here | Metric | Value | |-----------------------|---------------------------| | Avg. | 35.5 | | ARC (25-shot) | 41.55 | | HellaSwag (10-shot) | 61.52 | | MMLU (5-shot) | 26.79 | | TruthfulQA (0-shot) | 42.42 | | Winogrande (5-shot) | 61.8 | | GSM8K (5-shot) | 0.08 | | DROP (3-shot) | 14.33 | Open LLM Leaderboard Evaluation Results Detailed results can be found here | Metric |Value| |---------------------------------|----:| |Avg. |39.03| |AI2 Reasoning Challenge (25-Shot)|41.55| |HellaSwag (10-Shot) |61.52| |MMLU (5-Shot) |26.79| |TruthfulQA (0-shot) |42.42| |Winogrande (5-shot) |61.80| |GSM8k (5-shot) | 0.08|

orcaminiphi-4-GGUF is static quant version of orcaminiphi-4 "Obsessed with Open Source GenAI's potential? So am I ! Let's Contribute together 🚀 https://www.linkedin.com/in/pankajam " NOTICE By providing proper credit and attribution, you are granted permission to use this model as a foundational base for further Full fine tuning, DPO, PPO or ORPO tuning and any kind of Merges. I actively encourage users to customize and enhance the model according to their specific needs, as this version is designed to be a comprehensive general model. Dive in and innovate! My preferred method is Ollama but you can choose other ways to use this modle locally on your PC by going to right side of this repo card and select "Use this model". Here are the steps for Ollama: 1. Install Ollama https://github.com/ollama/ollama 2. After Installing Ollama (choose various available quants size, depending upon your PC RAM) run below command on your terminal, for example for Nov 2023 MacBook Pro with 16GB RAM: 4. Start Chatting with this model! 5. Optional: For ChatGPT like experience (or better), Install Open-Webui https://github.com/open-webui/open-webui 6. open-webui will auto connect to Ollama running this model and you can start chatting right away

OrcaMiniv81Llama-3.3-70B-Instruct is trained with various SFT Datasets on Llama-3.3-70B-Instruct "Obsessed with Open Source GenAI's potential? So am I ! Let's Contribute together 🚀 https://www.linkedin.com/in/pankajam " NOTICE By providing proper credit and attribution, you are granted permission to use this model as a foundational base for further Full fine tuning, DPO, PPO or ORPO tuning and any kind of Merges. I actively encourage users to customize and enhance the model according to their specific needs, as this version is designed to be a comprehensive general model. Dive in and innovate! Below shows a code example on how to use this model in default half precision (bfloat16) format, it requires around ~133GB VRAM Below shows a code example on how to use this model in 4-bit format via bitsandbytes library, it requires around ~39GB VRAM Below shows a code example on how to use this model in 8-bit format via bitsandbytes library, it requires around ~69GB VRAM Below shows a code example on how to do a tool use with this model and tranformer library Since orcaminiv8170b is based upon LLaMA-3.3, it supports multiple tool use formats. You can see a full guide to prompt formatting here. Tool use is also supported through chat templates in Transformers. Here is a quick example showing a single simple tool: You can then generate text from this input as normal. If the model generates a tool call, you should add it to the chat like so: and then call the tool and append the result, with the `tool` role, like so: After that, you can `generate()` again to let the model use the tool result in the chat. Note that this was a very brief introduction to tool calling - for more information, see the LLaMA prompt format docs and the Transformers tool use documentation. As part of our Responsible release approach, we followed a three-pronged strategy to managing trust & safety risks: Enable developers to deploy helpful, safe and flexible experiences for their target audience and for the use cases supported by Llama. Protect developers against adversarial users aiming to exploit Llama capabilities to potentially cause harm. Provide protections for the community to help prevent the misuse of our models. Llama is a foundational technology designed to be used in a variety of use cases, examples on how Meta’s Llama models have been responsibly deployed can be found in our Community Stories webpage. Our approach is to build the most helpful models enabling the world to benefit from the technology power, by aligning our model safety for the generic use cases addressing a standard set of harms. Developers are then in the driver seat to tailor safety for their use case, defining their own policy and deploying the models with the necessary safeguards in their Llama systems. Llama 3.3 was developed following the best practices outlined in our Responsible Use Guide, you can refer to the Responsible Use Guide to learn more. Our main objectives for conducting safety fine-tuning are to provide the research community with a valuable resource for studying the robustness of safety fine-tuning, as well as to offer developers a readily available, safe, and powerful model for various applications to reduce the developer workload to deploy safe AI systems. For more details on the safety mitigations implemented please read the Llama 3 paper. Fine-tuning data We employ a multi-faceted approach to data collection, combining human-generated data from our vendors with synthetic data to mitigate potential safety risks. We’ve developed many large language model (LLM)-based classifiers that enable us to thoughtfully select high-quality prompts and responses, enhancing data quality control. Refusals and Tone Building on the work we started with Llama 3, we put a great emphasis on model refusals to benign prompts as well as refusal tone. We included both borderline and adversarial prompts in our safety data strategy, and modified our safety data responses to follow tone guidelines. Large language models, including Llama 3.3, are not designed to be deployed in isolation but instead should be deployed as part of an overall AI system with additional safety guardrails as required. Developers are expected to deploy system safeguards when building agentic systems. Safeguards are key to achieve the right helpfulness-safety alignment as well as mitigating safety and security risks inherent to the system and any integration of the model or system with external tools. As part of our responsible release approach, we provide the community with safeguards that developers should deploy with Llama models or other LLMs, including Llama Guard 3, Prompt Guard and Code Shield. All our reference implementations demos contain these safeguards by default so developers can benefit from system-level safety out-of-the-box. Note that this release introduces new capabilities, including a longer context window, multilingual inputs and outputs and possible integrations by developers with third party tools. Building with these new capabilities requires specific considerations in addition to the best practices that generally apply across all Generative AI use cases. Tool-use: Just like in standard software development, developers are responsible for the integration of the LLM with the tools and services of their choice. They should define a clear policy for their use case and assess the integrity of the third party services they use to be aware of the safety and security limitations when using this capability. Refer to the Responsible Use Guide for best practices on the safe deployment of the third party safeguards. Multilinguality: Llama 3.3 supports 7 languages in addition to English: French, German, Hindi, Italian, Portuguese, Spanish, and Thai. Llama may be able to output text in other languages than those that meet performance thresholds for safety and helpfulness. We strongly discourage developers from using this model to converse in non-supported languages without implementing finetuning and system controls in alignment with their policies and the best practices shared in the Responsible Use Guide. We evaluated Llama models for common use cases as well as specific capabilities. Common use cases evaluations measure safety risks of systems for most commonly built applications including chat bot, coding assistant, tool calls. We built dedicated, adversarial evaluation datasets and evaluated systems composed of Llama models and Llama Guard 3 to filter input prompt and output response. It is important to evaluate applications in context, and we recommend building dedicated evaluation dataset for your use case. Prompt Guard and Code Shield are also available if relevant to the application. Capability evaluations measure vulnerabilities of Llama models inherent to specific capabilities, for which were crafted dedicated benchmarks including long context, multilingual, tools calls, coding or memorization. Red teaming For both scenarios, we conducted recurring red teaming exercises with the goal of discovering risks via adversarial prompting and we used the learnings to improve our benchmarks and safety tuning datasets. We partnered early with subject-matter experts in critical risk areas to understand the nature of these real-world harms and how such models may lead to unintended harm for society. Based on these conversations, we derived a set of adversarial goals for the red team to attempt to achieve, such as extracting harmful information or reprogramming the model to act in a potentially harmful capacity. The red team consisted of experts in cybersecurity, adversarial machine learning, responsible AI, and integrity in addition to multilingual content specialists with background in integrity issues in specific geographic markets. . We specifically focused our efforts on mitigating the following critical risk areas: 1- CBRNE (Chemical, Biological, Radiological, Nuclear, and Explosive materials) helpfulness To assess risks related to proliferation of chemical and biological weapons, we performed uplift testing designed to assess whether use of the Llama 3.3 model could meaningfully increase the capabilities of malicious actors to plan or carry out attacks using these types of weapons. Child Safety risk assessments were conducted using a team of experts, to assess the model’s capability to produce outputs that could result in Child Safety risks and inform on any necessary and appropriate risk mitigations via fine tuning. We leveraged those expert red teaming sessions to expand the coverage of our evaluation benchmarks through Llama 3 model development. For Llama 3, we conducted new in-depth sessions using objective based methodologies to assess the model risks along multiple attack vectors including the additional languages Llama 3 is trained on. We also partnered with content specialists to perform red teaming exercises assessing potentially violating content while taking account of market specific nuances or experiences. 3\. Cyber attack enablement Our cyber attack uplift study investigated whether LLMs can enhance human capabilities in hacking tasks, both in terms of skill level and speed. Our attack automation study focused on evaluating the capabilities of LLMs when used as autonomous agents in cyber offensive operations, specifically in the context of ransomware attacks. This evaluation was distinct from previous studies that considered LLMs as interactive assistants. The primary objective was to assess whether these models could effectively function as independent agents in executing complex cyber-attacks without human intervention. Generative AI safety requires expertise and tooling, and we believe in the strength of the open community to accelerate its progress. We are active members of open consortiums, including the AI Alliance, Partnership on AI and MLCommons, actively contributing to safety standardization and transparency. We encourage the community to adopt taxonomies like the MLCommons Proof of Concept evaluation to facilitate collaboration and transparency on safety and content evaluations. Our Purple Llama tools are open sourced for the community to use and widely distributed across ecosystem partners including cloud service providers. We encourage community contributions to our Github repository. We also set up the Llama Impact Grants program to identify and support the most compelling applications of Meta’s Llama model for societal benefit across three categories: education, climate and open innovation. The 20 finalists from the hundreds of applications can be found here. Finally, we put in place a set of resources including an output reporting mechanism and bug bounty program to continuously improve the Llama technology with the help of the community. The core values of Llama 3.3 are openness, inclusivity and helpfulness. It is meant to serve everyone, and to work for a wide range of use cases. It is thus designed to be accessible to people across many different backgrounds, experiences and perspectives. Llama 3.3 addresses users and their needs as they are, without insertion unnecessary judgment or normativity, while reflecting the understanding that even content that may appear problematic in some cases can serve valuable purposes in others. It respects the dignity and autonomy of all users, especially in terms of the values of free thought and expression that power innovation and progress. But Llama 3.3 is a new technology, and like any new technology, there are risks associated with its use. Testing conducted to date has not covered, nor could it cover, all scenarios. For these reasons, as with all LLMs, Llama 3.3’s potential outputs cannot be predicted in advance, and the model may in some instances produce inaccurate, biased or other objectionable responses to user prompts. Therefore, before deploying any applications of Llama 3.3 model, developers should perform safety testing and tuning tailored to their specific applications of the model. Please refer to available resources including our Responsible Use Guide, Trust and Safety solutions, and other resources to learn more about responsible development.

A hybrid (explain + instruct) style Llama2-13b model, Pleae check examples below for both style prompts, Here is the list of datasets used: Open-Platypus Alpaca WizardLM Dolly-V2 Dolphin Samples (~200K) Orcaminisv1 Alpacaorca WizardLMorca Dolly-V2orca Passionate about Generative AI? I help companies to privately train and deploy custom LLM/MLLM affordably. For startups, I can even assist with securing GPU grants to get you started. Let's chat! This model is bound by the license & usage restrictions of the original Llama-2 model. And comes with no warranty or gurantees of any kind. We evaluated model00713bv2 on a wide range of tasks using Language Model Evaluation Harness from EleutherAI. Here are the results on metrics used by HuggingFaceH4 Open LLM Leaderboard ||||| |:------:|:--------:|:-------:|:--------:| |Task|Metric|Value|Stderr| |arcchallenge|accnorm|0.6314|0.0141| |hellaswag|accnorm|0.8242|0.0038| |mmlu|accnorm|0.5637|0.0351| |truthfulqamc|mc2|0.5127|0.0157| |Total Average|-|0.6329877193|| Below shows a code example on how to use this model Below shows a code example on how to use this model While this model aims for accuracy, it can occasionally produce inaccurate or misleading results. Despite diligent efforts in refining the pretraining data, there remains a possibility for the generation of inappropriate, biased, or offensive content. Exercise caution and cross-check information when necessary. Open LLM Leaderboard Evaluation Results Detailed results can be found here | Metric | Value | |-----------------------|---------------------------| | Avg. | 53.78 | | ARC (25-shot) | 61.95 | | HellaSwag (10-shot) | 82.48 | | MMLU (5-shot) | 57.32 | | TruthfulQA (0-shot) | 53.5 | | Winogrande (5-shot) | 75.85 | | GSM8K (5-shot) | 1.36 | | DROP (3-shot) | 43.97 | Open LLM Leaderboard Evaluation Results Detailed results can be found here | Metric |Value| |---------------------------------|----:| |Avg. |55.41| |AI2 Reasoning Challenge (25-Shot)|61.95| |HellaSwag (10-Shot) |82.48| |MMLU (5-Shot) |57.32| |TruthfulQA (0-shot) |53.50| |Winogrande (5-shot) |75.85| |GSM8k (5-shot) | 1.36| Open LLM Leaderboard Evaluation Results Detailed results can be found here | Metric |Value| |-------------------|----:| |Avg. |15.86| |IFEval (0-Shot) |30.56| |BBH (3-Shot) |25.45| |MATH Lvl 5 (4-Shot)| 1.21| |GPQA (0-shot) | 4.47| |MuSR (0-shot) |17.20| |MMLU-PRO (5-shot) |16.23|

OrcaMiniv96Llama-3.2-1B-Instruct is trained with various SFT Datasets on Llama-3.2-1B-Instruct "Obsessed with Open Source GenAI's potential? So am I ! Let's Contribute together 🚀 https://www.linkedin.com/in/pankajam " NOTICE By providing proper credit and attribution, you are granted permission to use this model as a foundational base for further Full fine tuning, DPO, PPO or ORPO tuning and any kind of Merges. I actively encourage users to customize and enhance the model according to their specific needs, as this version is designed to be a comprehensive general model. Dive in and innovate! Use this model for Free on Google Colab with T4 GPU :) Download GGUF version here and Follow Ollama instructions: orcaminiv961B-Instruct-GGUF Below shows a code example on how to use this model in default half precision (bfloat16) format Below shows a code example on how to use this model in 4-bit format via bitsandbytes library Below shows a code example on how to use this model in 8-bit format via bitsandbytes library As part of our Responsible release approach, we followed a three-pronged strategy to managing trust & safety risks: 1. Enable developers to deploy helpful, safe and flexible experiences for their target audience and for the use cases supported by Llama 2. Protect developers against adversarial users aiming to exploit Llama capabilities to potentially cause harm 3. Provide protections for the community to help prevent the misuse of our models Approach: Llama is a foundational technology designed to be used in a variety of use cases. Examples on how Meta’s Llama models have been responsibly deployed can be found in our Community Stories webpage. Our approach is to build the most helpful models, enabling the world to benefit from the technology power, by aligning our model safety for generic use cases and addressing a standard set of harms. Developers are then in the driver’s seat to tailor safety for their use cases, defining their own policies and deploying the models with the necessary safeguards in their Llama systems. Llama 3.2 was developed following the best practices outlined in our Responsible Use Guide. Objective: Our main objectives for conducting safety fine-tuning are to provide the research community with a valuable resource for studying the robustness of safety fine-tuning, as well as to offer developers a readily available, safe, and powerful model for various applications to reduce the developer workload to deploy safe AI systems. We implemented the same set of safety mitigations as in Llama 3, and you can learn more about these in the Llama 3 paper. Fine-Tuning Data: We employ a multi-faceted approach to data collection, combining human-generated data from our vendors with synthetic data to mitigate potential safety risks. We’ve developed many large language model (LLM)-based classifiers that enable us to thoughtfully select high-quality prompts and responses, enhancing data quality control. Refusals and Tone: Building on the work we started with Llama 3, we put a great emphasis on model refusals to benign prompts as well as refusal tone. We included both borderline and adversarial prompts in our safety data strategy, and modified our safety data responses to follow tone guidelines. Safety as a System: Large language models, including Llama 3.2, are not designed to be deployed in isolation but instead should be deployed as part of an overall AI system with additional safety guardrails as required. Developers are expected to deploy system safeguards when building agentic systems. Safeguards are key to achieve the right helpfulness-safety alignment as well as mitigating safety and security risks inherent to the system and any integration of the model or system with external tools. As part of our responsible release approach, we provide the community with safeguards that developers should deploy with Llama models or other LLMs, including Llama Guard, Prompt Guard and Code Shield. All our reference implementations demos contain these safeguards by default so developers can benefit from system-level safety out-of-the-box. Technological Advancement: Llama releases usually introduce new capabilities that require specific considerations in addition to the best practices that generally apply across all Generative AI use cases. For prior release capabilities also supported by Llama 3.2, see Llama 3.1 Model Card, as the same considerations apply here as well. Constrained Environments: Llama 3.2 1B and 3B models are expected to be deployed in highly constrained environments, such as mobile devices. LLM Systems using smaller models will have a different alignment profile and safety/helpfulness tradeoff than more complex, larger systems. Developers should ensure the safety of their system meets the requirements of their use case. We recommend using lighter system safeguards for such use cases, like Llama Guard 3-1B or its mobile-optimized version. Scaled Evaluations: We built dedicated, adversarial evaluation datasets and evaluated systems composed of Llama models and Purple Llama safeguards to filter input prompt and output response. It is important to evaluate applications in context, and we recommend building dedicated evaluation dataset for your use case. Red Teaming: We conducted recurring red teaming exercises with the goal of discovering risks via adversarial prompting and we used the learnings to improve our benchmarks and safety tuning datasets. We partnered early with subject-matter experts in critical risk areas to understand the nature of these real-world harms and how such models may lead to unintended harm for society. Based on these conversations, we derived a set of adversarial goals for the red team to attempt to achieve, such as extracting harmful information or reprogramming the model to act in a potentially harmful capacity. The red team consisted of experts in cybersecurity, adversarial machine learning, responsible AI, and integrity in addition to multilingual content specialists with background in integrity issues in specific geographic markets. In addition to our safety work above, we took extra care on measuring and/or mitigating the following critical risk areas: 1\. CBRNE (Chemical, Biological, Radiological, Nuclear, and Explosive Weapons): Llama 3.2 1B and 3B models are smaller and less capable derivatives of Llama 3.1. For Llama 3.1 70B and 405B, to assess risks related to proliferation of chemical and biological weapons, we performed uplift testing designed to assess whether use of Llama 3.1 models could meaningfully increase the capabilities of malicious actors to plan or carry out attacks using these types of weapons and have determined that such testing also applies to the smaller 1B and 3B models. 2\. Child Safety: Child Safety risk assessments were conducted using a team of experts, to assess the model’s capability to produce outputs that could result in Child Safety risks and inform on any necessary and appropriate risk mitigations via fine tuning. We leveraged those expert red teaming sessions to expand the coverage of our evaluation benchmarks through Llama 3 model development. For Llama 3, we conducted new in-depth sessions using objective based methodologies to assess the model risks along multiple attack vectors including the additional languages Llama 3 is trained on. We also partnered with content specialists to perform red teaming exercises assessing potentially violating content while taking account of market specific nuances or experiences. 3\. Cyber Attacks: For Llama 3.1 405B, our cyber attack uplift study investigated whether LLMs can enhance human capabilities in hacking tasks, both in terms of skill level and speed. Our attack automation study focused on evaluating the capabilities of LLMs when used as autonomous agents in cyber offensive operations, specifically in the context of ransomware attacks. This evaluation was distinct from previous studies that considered LLMs as interactive assistants. The primary objective was to assess whether these models could effectively function as independent agents in executing complex cyber-attacks without human intervention. Because Llama 3.2’s 1B and 3B models are smaller and less capable models than Llama 3.1 405B, we broadly believe that the testing conducted for the 405B model also applies to Llama 3.2 models. Industry Partnerships: Generative AI safety requires expertise and tooling, and we believe in the strength of the open community to accelerate its progress. We are active members of open consortiums, including the AI Alliance, Partnership on AI and MLCommons, actively contributing to safety standardization and transparency. We encourage the community to adopt taxonomies like the MLCommons Proof of Concept evaluation to facilitate collaboration and transparency on safety and content evaluations. Our Purple Llama tools are open sourced for the community to use and widely distributed across ecosystem partners including cloud service providers. We encourage community contributions to our Github repository. Grants: We also set up the Llama Impact Grants program to identify and support the most compelling applications of Meta’s Llama model for societal benefit across three categories: education, climate and open innovation. The 20 finalists from the hundreds of applications can be found here. Reporting: Finally, we put in place a set of resources including an output reporting mechanism and bug bounty program to continuously improve the Llama technology with the help of the community. Values: The core values of Llama 3.2 are openness, inclusivity and helpfulness. It is meant to serve everyone, and to work for a wide range of use cases. It is thus designed to be accessible to people across many different backgrounds, experiences and perspectives. Llama 3.2 addresses users and their needs as they are, without insertion unnecessary judgment or normativity, while reflecting the understanding that even content that may appear problematic in some cases can serve valuable purposes in others. It respects the dignity and autonomy of all users, especially in terms of the values of free thought and expression that power innovation and progress. Testing: Llama 3.2 is a new technology, and like any new technology, there are risks associated with its use. Testing conducted to date has not covered, nor could it cover, all scenarios. For these reasons, as with all LLMs, Llama 3.2’s potential outputs cannot be predicted in advance, and the model may in some instances produce inaccurate, biased or other objectionable responses to user prompts. Therefore, before deploying any applications of Llama 3.2 models, developers should perform safety testing and tuning tailored to their specific applications of the model. Please refer to available resources including our Responsible Use Guide, Trust and Safety solutions, and other resources to learn more about responsible development.

OrcaMiniv95Llama-3.2-3B-Instruct is trained with various SFT Datasets on Llama-3.2-3B-Instruct "Obsessed with Open Source GenAI's potential? So am I ! Let's Contribute together 🚀 https://www.linkedin.com/in/pankajam " NOTICE By providing proper credit and attribution, you are granted permission to use this model as a foundational base for further Full fine tuning, DPO, PPO or ORPO tuning and any kind of Merges. I actively encourage users to customize and enhance the model according to their specific needs, as this version is designed to be a comprehensive general model. Dive in and innovate! Use this model for Free on Google Colab with T4 GPU :) Below shows a code example on how to use this model in default half precision (bfloat16) format Below shows a code example on how to use this model in 4-bit format via bitsandbytes library Below shows a code example on how to use this model in 8-bit format via bitsandbytes library As part of our Responsible release approach, we followed a three-pronged strategy to managing trust & safety risks: 1. Enable developers to deploy helpful, safe and flexible experiences for their target audience and for the use cases supported by Llama 2. Protect developers against adversarial users aiming to exploit Llama capabilities to potentially cause harm 3. Provide protections for the community to help prevent the misuse of our models Approach: Llama is a foundational technology designed to be used in a variety of use cases. Examples on how Meta’s Llama models have been responsibly deployed can be found in our Community Stories webpage. Our approach is to build the most helpful models, enabling the world to benefit from the technology power, by aligning our model safety for generic use cases and addressing a standard set of harms. Developers are then in the driver’s seat to tailor safety for their use cases, defining their own policies and deploying the models with the necessary safeguards in their Llama systems. Llama 3.2 was developed following the best practices outlined in our Responsible Use Guide. Objective: Our main objectives for conducting safety fine-tuning are to provide the research community with a valuable resource for studying the robustness of safety fine-tuning, as well as to offer developers a readily available, safe, and powerful model for various applications to reduce the developer workload to deploy safe AI systems. We implemented the same set of safety mitigations as in Llama 3, and you can learn more about these in the Llama 3 paper. Fine-Tuning Data: We employ a multi-faceted approach to data collection, combining human-generated data from our vendors with synthetic data to mitigate potential safety risks. We’ve developed many large language model (LLM)-based classifiers that enable us to thoughtfully select high-quality prompts and responses, enhancing data quality control. Refusals and Tone: Building on the work we started with Llama 3, we put a great emphasis on model refusals to benign prompts as well as refusal tone. We included both borderline and adversarial prompts in our safety data strategy, and modified our safety data responses to follow tone guidelines. Safety as a System: Large language models, including Llama 3.2, are not designed to be deployed in isolation but instead should be deployed as part of an overall AI system with additional safety guardrails as required. Developers are expected to deploy system safeguards when building agentic systems. Safeguards are key to achieve the right helpfulness-safety alignment as well as mitigating safety and security risks inherent to the system and any integration of the model or system with external tools. As part of our responsible release approach, we provide the community with safeguards that developers should deploy with Llama models or other LLMs, including Llama Guard, Prompt Guard and Code Shield. All our reference implementations demos contain these safeguards by default so developers can benefit from system-level safety out-of-the-box. Technological Advancement: Llama releases usually introduce new capabilities that require specific considerations in addition to the best practices that generally apply across all Generative AI use cases. For prior release capabilities also supported by Llama 3.2, see Llama 3.1 Model Card, as the same considerations apply here as well. Constrained Environments: Llama 3.2 1B and 3B models are expected to be deployed in highly constrained environments, such as mobile devices. LLM Systems using smaller models will have a different alignment profile and safety/helpfulness tradeoff than more complex, larger systems. Developers should ensure the safety of their system meets the requirements of their use case. We recommend using lighter system safeguards for such use cases, like Llama Guard 3-1B or its mobile-optimized version. Scaled Evaluations: We built dedicated, adversarial evaluation datasets and evaluated systems composed of Llama models and Purple Llama safeguards to filter input prompt and output response. It is important to evaluate applications in context, and we recommend building dedicated evaluation dataset for your use case. Red Teaming: We conducted recurring red teaming exercises with the goal of discovering risks via adversarial prompting and we used the learnings to improve our benchmarks and safety tuning datasets. We partnered early with subject-matter experts in critical risk areas to understand the nature of these real-world harms and how such models may lead to unintended harm for society. Based on these conversations, we derived a set of adversarial goals for the red team to attempt to achieve, such as extracting harmful information or reprogramming the model to act in a potentially harmful capacity. The red team consisted of experts in cybersecurity, adversarial machine learning, responsible AI, and integrity in addition to multilingual content specialists with background in integrity issues in specific geographic markets. In addition to our safety work above, we took extra care on measuring and/or mitigating the following critical risk areas: 1\. CBRNE (Chemical, Biological, Radiological, Nuclear, and Explosive Weapons): Llama 3.2 1B and 3B models are smaller and less capable derivatives of Llama 3.1. For Llama 3.1 70B and 405B, to assess risks related to proliferation of chemical and biological weapons, we performed uplift testing designed to assess whether use of Llama 3.1 models could meaningfully increase the capabilities of malicious actors to plan or carry out attacks using these types of weapons and have determined that such testing also applies to the smaller 1B and 3B models. 2\. Child Safety: Child Safety risk assessments were conducted using a team of experts, to assess the model’s capability to produce outputs that could result in Child Safety risks and inform on any necessary and appropriate risk mitigations via fine tuning. We leveraged those expert red teaming sessions to expand the coverage of our evaluation benchmarks through Llama 3 model development. For Llama 3, we conducted new in-depth sessions using objective based methodologies to assess the model risks along multiple attack vectors including the additional languages Llama 3 is trained on. We also partnered with content specialists to perform red teaming exercises assessing potentially violating content while taking account of market specific nuances or experiences. 3\. Cyber Attacks: For Llama 3.1 405B, our cyber attack uplift study investigated whether LLMs can enhance human capabilities in hacking tasks, both in terms of skill level and speed. Our attack automation study focused on evaluating the capabilities of LLMs when used as autonomous agents in cyber offensive operations, specifically in the context of ransomware attacks. This evaluation was distinct from previous studies that considered LLMs as interactive assistants. The primary objective was to assess whether these models could effectively function as independent agents in executing complex cyber-attacks without human intervention. Because Llama 3.2’s 1B and 3B models are smaller and less capable models than Llama 3.1 405B, we broadly believe that the testing conducted for the 405B model also applies to Llama 3.2 models. Industry Partnerships: Generative AI safety requires expertise and tooling, and we believe in the strength of the open community to accelerate its progress. We are active members of open consortiums, including the AI Alliance, Partnership on AI and MLCommons, actively contributing to safety standardization and transparency. We encourage the community to adopt taxonomies like the MLCommons Proof of Concept evaluation to facilitate collaboration and transparency on safety and content evaluations. Our Purple Llama tools are open sourced for the community to use and widely distributed across ecosystem partners including cloud service providers. We encourage community contributions to our Github repository. Grants: We also set up the Llama Impact Grants program to identify and support the most compelling applications of Meta’s Llama model for societal benefit across three categories: education, climate and open innovation. The 20 finalists from the hundreds of applications can be found here. Reporting: Finally, we put in place a set of resources including an output reporting mechanism and bug bounty program to continuously improve the Llama technology with the help of the community. Values: The core values of Llama 3.2 are openness, inclusivity and helpfulness. It is meant to serve everyone, and to work for a wide range of use cases. It is thus designed to be accessible to people across many different backgrounds, experiences and perspectives. Llama 3.2 addresses users and their needs as they are, without insertion unnecessary judgment or normativity, while reflecting the understanding that even content that may appear problematic in some cases can serve valuable purposes in others. It respects the dignity and autonomy of all users, especially in terms of the values of free thought and expression that power innovation and progress. Testing: Llama 3.2 is a new technology, and like any new technology, there are risks associated with its use. Testing conducted to date has not covered, nor could it cover, all scenarios. For these reasons, as with all LLMs, Llama 3.2’s potential outputs cannot be predicted in advance, and the model may in some instances produce inaccurate, biased or other objectionable responses to user prompts. Therefore, before deploying any applications of Llama 3.2 models, developers should perform safety testing and tuning tailored to their specific applications of the model. Please refer to available resources including our Responsible Use Guide, Trust and Safety solutions, and other resources to learn more about responsible development.

Model Name: Qwen2 orcaminiv772b Qwen2 orcaminiv772b is trained with various SFT Datasets "Obsessed with GenAI's potential? So am I ! Let's create together 🚀 https://www.linkedin.com/in/pankajam " NOTICE By providing proper credit and attribution, you are granted permission to use this model as a foundational base for further Full fine tuning, DPO, PPO or ORPO tuning and any kind of Merges. I actively encourage users to customize and enhance the model according to their specific needs, as this version is designed to be a comprehensive general model. Dive in and innovate! Below shows a code example on how to use this model Processing Long Texts (Based upon Qwen2-7B-Instruct suggestions at https://huggingface.co/Qwen/Qwen2-7B-Instruct) To handle extensive inputs exceeding 32,768 tokens, we utilize YARN, a technique for enhancing model length extrapolation, ensuring optimal performance on lengthy texts. For deployment, we recommend using vLLM. You can enable the long-context capabilities by following these steps: 1. Install vLLM: You can install vLLM by running the following command. 2. Configure Model Settings: After downloading the model weights, modify the `config.json` file by including the below snippet: This snippet enable YARN to support longer contexts. 3. Model Deployment: Utilize vLLM to deploy your model. For instance, you can set up an openAI-like server using the command: Note: Presently, vLLM only supports static YARN, which means the scaling factor remains constant regardless of input length, potentially impacting performance on shorter texts. We advise adding the `ropescaling` configuration only when processing long contexts is required. Open LLM Leaderboard Evaluation Results Detailed results can be found here | Metric |Value| |-------------------|----:| |Avg. |39.06| |IFEval (0-Shot) |59.30| |BBH (3-Shot) |55.06| |MATH Lvl 5 (4-Shot)|26.44| |GPQA (0-shot) |18.01| |MuSR (0-shot) |24.21| |MMLU-PRO (5-shot) |51.35|

orcaminiphi-4 is trained with various SFT Datasets on microsoft/phi-4 using Llama's architecture. "Obsessed with Open Source GenAI's potential? So am I ! Let's Contribute together 🚀 https://www.linkedin.com/in/pankajam " NOTICE By providing proper credit and attribution, you are granted permission to use this model as a foundational base for further Full fine tuning, DPO, PPO or ORPO tuning and any kind of Merges. I actively encourage users to customize and enhance the model according to their specific needs, as this version is designed to be a comprehensive general model. Dive in and innovate! Use this model for Free on Google Colab with T4 GPU :) Download GGUF version here and Follow Ollama instructions: https://huggingface.co/pankajmathur/orcaminiphi-4-GGUF Below shows a code example on how to use this model in default half precision (bfloat16) format Below shows a code example on how to use this model in 4-bit format via bitsandbytes library Below shows a code example on how to use this model in 8-bit format via bitsandbytes library

"Obsessed with GenAI's potential? So am I ! Let's create together 🚀 https://www.linkedin.com/in/pankajam " An OpenLLaMa-7B model model trained on explain tuned datasets, created using Instructions and Input from WizardLM, Alpaca & Dolly-V2 datasets and applying Orca Research Paper dataset construction approaches. We build explain tuned WizardLM dataset ~70K, Alpaca dataset ~52K & Dolly-V2 dataset ~15K created using approaches from Orca Research Paper. We leverage all of the 15 system instructions provided in Orca Research Paper. to generate custom datasets, in contrast to vanilla instruction tuning approaches used by original datasets. This helps student model aka this model to learn thought process from teacher model, which is ChatGPT (gpt-3.5-turbo-0301 version). Please see below example usage how the System prompt is added before each instruction. The training configurations are provided in the table below. The training takes on 8x A100(80G) GPUs and lasts for around 7 Hours for cost of $84 using Lambda Labs We used DeepSpeed with fully sharded data parallelism, also know as ZeRO stage 3 by writing our own fine tunning scripts plus leveraging some of the model training code provided by amazing OpenAlpaca repo ||| |:-------------:|:-------------:| |batchsize|32| |trainmicrobatchsizepergpu|2| |gradientaccumulationsteps|2| |Learning rate|2e-5| |Max length|1024| |Epochs|3| |Optimizer|AdamW| Next Goals: 1) Try more data like actually using FLAN-v2, just like Orka Research Paper (I am open for suggestions) 2) Provide more options for Text generation UI. (may be https://github.com/oobabooga/text-generation-webui) 3) Provide 4bit GGML/GPTQ quantized model (may be TheBloke can help here) This model can produce factually incorrect output, and should not be relied on to produce factually accurate information. This model was trained on various public datasets. While great efforts have been taken to clean the pretraining data, it is possible that this model could generate lewd, biased or otherwise offensive outputs. The license on this model does not constitute legal advice. We are not responsible for the actions of third parties who use this model. Please cosult an attorney before using this model for commercial purposes. If you found wizardlmalpacadollyorcaopenllama7b useful in your research or applications, please kindly cite using the following BibTeX: Open LLM Leaderboard Evaluation Results Detailed results can be found here | Metric | Value | |-----------------------|---------------------------| | Avg. | 37.39 | | ARC (25-shot) | 43.94 | | HellaSwag (10-shot) | 65.22 | | MMLU (5-shot) | 29.97 | | TruthfulQA (0-shot) | 42.03 | | Winogrande (5-shot) | 66.06 | | GSM8K (5-shot) | 0.38 | | DROP (3-shot) | 14.14 | Open LLM Leaderboard Evaluation Results Detailed results can be found here | Metric |Value| |---------------------------------|----:| |Avg. |41.27| |AI2 Reasoning Challenge (25-Shot)|43.94| |HellaSwag (10-Shot) |65.22| |MMLU (5-Shot) |29.97| |TruthfulQA (0-shot) |42.03| |Winogrande (5-shot) |66.06| |GSM8k (5-shot) | 0.38|

OrcaMiniv95Llama-3.2-1B-Instruct is trained with various SFT Datasets on Llama-3.2-3B-Instruct "Obsessed with Open Source GenAI's potential? So am I ! Let's Contribute together 🚀 https://www.linkedin.com/in/pankajam " NOTICE By providing proper credit and attribution, you are granted permission to use this model as a foundational base for further Full fine tuning, DPO, PPO or ORPO tuning and any kind of Merges. I actively encourage users to customize and enhance the model according to their specific needs, as this version is designed to be a comprehensive general model. Dive in and innovate! Use this model for Free on Google Colab with T4 GPU :) Below shows a code example on how to use this model in default half precision (bfloat16) format Below shows a code example on how to use this model in 4-bit format via bitsandbytes library Below shows a code example on how to use this model in 8-bit format via bitsandbytes library As part of our Responsible release approach, we followed a three-pronged strategy to managing trust & safety risks: 1. Enable developers to deploy helpful, safe and flexible experiences for their target audience and for the use cases supported by Llama 2. Protect developers against adversarial users aiming to exploit Llama capabilities to potentially cause harm 3. Provide protections for the community to help prevent the misuse of our models Approach: Llama is a foundational technology designed to be used in a variety of use cases. Examples on how Meta’s Llama models have been responsibly deployed can be found in our Community Stories webpage. Our approach is to build the most helpful models, enabling the world to benefit from the technology power, by aligning our model safety for generic use cases and addressing a standard set of harms. Developers are then in the driver’s seat to tailor safety for their use cases, defining their own policies and deploying the models with the necessary safeguards in their Llama systems. Llama 3.2 was developed following the best practices outlined in our Responsible Use Guide. Objective: Our main objectives for conducting safety fine-tuning are to provide the research community with a valuable resource for studying the robustness of safety fine-tuning, as well as to offer developers a readily available, safe, and powerful model for various applications to reduce the developer workload to deploy safe AI systems. We implemented the same set of safety mitigations as in Llama 3, and you can learn more about these in the Llama 3 paper. Fine-Tuning Data: We employ a multi-faceted approach to data collection, combining human-generated data from our vendors with synthetic data to mitigate potential safety risks. We’ve developed many large language model (LLM)-based classifiers that enable us to thoughtfully select high-quality prompts and responses, enhancing data quality control. Refusals and Tone: Building on the work we started with Llama 3, we put a great emphasis on model refusals to benign prompts as well as refusal tone. We included both borderline and adversarial prompts in our safety data strategy, and modified our safety data responses to follow tone guidelines. Safety as a System: Large language models, including Llama 3.2, are not designed to be deployed in isolation but instead should be deployed as part of an overall AI system with additional safety guardrails as required. Developers are expected to deploy system safeguards when building agentic systems. Safeguards are key to achieve the right helpfulness-safety alignment as well as mitigating safety and security risks inherent to the system and any integration of the model or system with external tools. As part of our responsible release approach, we provide the community with safeguards that developers should deploy with Llama models or other LLMs, including Llama Guard, Prompt Guard and Code Shield. All our reference implementations demos contain these safeguards by default so developers can benefit from system-level safety out-of-the-box. Technological Advancement: Llama releases usually introduce new capabilities that require specific considerations in addition to the best practices that generally apply across all Generative AI use cases. For prior release capabilities also supported by Llama 3.2, see Llama 3.1 Model Card, as the same considerations apply here as well. Constrained Environments: Llama 3.2 1B and 3B models are expected to be deployed in highly constrained environments, such as mobile devices. LLM Systems using smaller models will have a different alignment profile and safety/helpfulness tradeoff than more complex, larger systems. Developers should ensure the safety of their system meets the requirements of their use case. We recommend using lighter system safeguards for such use cases, like Llama Guard 3-1B or its mobile-optimized version. Scaled Evaluations: We built dedicated, adversarial evaluation datasets and evaluated systems composed of Llama models and Purple Llama safeguards to filter input prompt and output response. It is important to evaluate applications in context, and we recommend building dedicated evaluation dataset for your use case. Red Teaming: We conducted recurring red teaming exercises with the goal of discovering risks via adversarial prompting and we used the learnings to improve our benchmarks and safety tuning datasets. We partnered early with subject-matter experts in critical risk areas to understand the nature of these real-world harms and how such models may lead to unintended harm for society. Based on these conversations, we derived a set of adversarial goals for the red team to attempt to achieve, such as extracting harmful information or reprogramming the model to act in a potentially harmful capacity. The red team consisted of experts in cybersecurity, adversarial machine learning, responsible AI, and integrity in addition to multilingual content specialists with background in integrity issues in specific geographic markets. In addition to our safety work above, we took extra care on measuring and/or mitigating the following critical risk areas: 1\. CBRNE (Chemical, Biological, Radiological, Nuclear, and Explosive Weapons): Llama 3.2 1B and 3B models are smaller and less capable derivatives of Llama 3.1. For Llama 3.1 70B and 405B, to assess risks related to proliferation of chemical and biological weapons, we performed uplift testing designed to assess whether use of Llama 3.1 models could meaningfully increase the capabilities of malicious actors to plan or carry out attacks using these types of weapons and have determined that such testing also applies to the smaller 1B and 3B models. 2\. Child Safety: Child Safety risk assessments were conducted using a team of experts, to assess the model’s capability to produce outputs that could result in Child Safety risks and inform on any necessary and appropriate risk mitigations via fine tuning. We leveraged those expert red teaming sessions to expand the coverage of our evaluation benchmarks through Llama 3 model development. For Llama 3, we conducted new in-depth sessions using objective based methodologies to assess the model risks along multiple attack vectors including the additional languages Llama 3 is trained on. We also partnered with content specialists to perform red teaming exercises assessing potentially violating content while taking account of market specific nuances or experiences. 3\. Cyber Attacks: For Llama 3.1 405B, our cyber attack uplift study investigated whether LLMs can enhance human capabilities in hacking tasks, both in terms of skill level and speed. Our attack automation study focused on evaluating the capabilities of LLMs when used as autonomous agents in cyber offensive operations, specifically in the context of ransomware attacks. This evaluation was distinct from previous studies that considered LLMs as interactive assistants. The primary objective was to assess whether these models could effectively function as independent agents in executing complex cyber-attacks without human intervention. Because Llama 3.2’s 1B and 3B models are smaller and less capable models than Llama 3.1 405B, we broadly believe that the testing conducted for the 405B model also applies to Llama 3.2 models. Industry Partnerships: Generative AI safety requires expertise and tooling, and we believe in the strength of the open community to accelerate its progress. We are active members of open consortiums, including the AI Alliance, Partnership on AI and MLCommons, actively contributing to safety standardization and transparency. We encourage the community to adopt taxonomies like the MLCommons Proof of Concept evaluation to facilitate collaboration and transparency on safety and content evaluations. Our Purple Llama tools are open sourced for the community to use and widely distributed across ecosystem partners including cloud service providers. We encourage community contributions to our Github repository. Grants: We also set up the Llama Impact Grants program to identify and support the most compelling applications of Meta’s Llama model for societal benefit across three categories: education, climate and open innovation. The 20 finalists from the hundreds of applications can be found here. Reporting: Finally, we put in place a set of resources including an output reporting mechanism and bug bounty program to continuously improve the Llama technology with the help of the community. Values: The core values of Llama 3.2 are openness, inclusivity and helpfulness. It is meant to serve everyone, and to work for a wide range of use cases. It is thus designed to be accessible to people across many different backgrounds, experiences and perspectives. Llama 3.2 addresses users and their needs as they are, without insertion unnecessary judgment or normativity, while reflecting the understanding that even content that may appear problematic in some cases can serve valuable purposes in others. It respects the dignity and autonomy of all users, especially in terms of the values of free thought and expression that power innovation and progress. Testing: Llama 3.2 is a new technology, and like any new technology, there are risks associated with its use. Testing conducted to date has not covered, nor could it cover, all scenarios. For these reasons, as with all LLMs, Llama 3.2’s potential outputs cannot be predicted in advance, and the model may in some instances produce inaccurate, biased or other objectionable responses to user prompts. Therefore, before deploying any applications of Llama 3.2 models, developers should perform safety testing and tuning tailored to their specific applications of the model. Please refer to available resources including our Responsible Use Guide, Trust and Safety solutions, and other resources to learn more about responsible development.

OrcaMiniV94Llama-3.3-70B-Instruct is trained with various SFT Datasets on Llama-3.3-70B-Instruct "Obsessed with Open Source GenAI's potential? So am I ! Let's Contribute together 🚀 https://www.linkedin.com/in/pankajam " NOTICE By providing proper credit and attribution, you are granted permission to use this model as a foundational base for further Full fine tuning, DPO, PPO or ORPO tuning and any kind of Merges. I actively encourage users to customize and enhance the model according to their specific needs, as this version is designed to be a comprehensive general model. Dive in and innovate! Below shows a code example on how to use this model in default half precision (bfloat16), it requires around ~133GB VRAM Below shows a code example on how to use this model in 4-bit format via bitsandbytes library, it requires around ~39GB VRAM Below shows a code example on how to use this model in 8-bit format via bitsandbytes library, it requires around ~69GB VRAM As part of our Responsible release approach, we followed a three-pronged strategy to managing trust & safety risks: Enable developers to deploy helpful, safe and flexible experiences for their target audience and for the use cases supported by Llama. Protect developers against adversarial users aiming to exploit Llama capabilities to potentially cause harm. Provide protections for the community to help prevent the misuse of our models. Llama is a foundational technology designed to be used in a variety of use cases, examples on how Meta’s Llama models have been responsibly deployed can be found in our Community Stories webpage. Our approach is to build the most helpful models enabling the world to benefit from the technology power, by aligning our model safety for the generic use cases addressing a standard set of harms. Developers are then in the driver seat to tailor safety for their use case, defining their own policy and deploying the models with the necessary safeguards in their Llama systems. Llama 3.3 was developed following the best practices outlined in our Responsible Use Guide, you can refer to the Responsible Use Guide to learn more. Our main objectives for conducting safety fine-tuning are to provide the research community with a valuable resource for studying the robustness of safety fine-tuning, as well as to offer developers a readily available, safe, and powerful model for various applications to reduce the developer workload to deploy safe AI systems. For more details on the safety mitigations implemented please read the Llama 3 paper. Fine-tuning data We employ a multi-faceted approach to data collection, combining human-generated data from our vendors with synthetic data to mitigate potential safety risks. We’ve developed many large language model (LLM)-based classifiers that enable us to thoughtfully select high-quality prompts and responses, enhancing data quality control. Refusals and Tone Building on the work we started with Llama 3, we put a great emphasis on model refusals to benign prompts as well as refusal tone. We included both borderline and adversarial prompts in our safety data strategy, and modified our safety data responses to follow tone guidelines. Large language models, including Llama 3.3, are not designed to be deployed in isolation but instead should be deployed as part of an overall AI system with additional safety guardrails as required. Developers are expected to deploy system safeguards when building agentic systems. Safeguards are key to achieve the right helpfulness-safety alignment as well as mitigating safety and security risks inherent to the system and any integration of the model or system with external tools. As part of our responsible release approach, we provide the community with safeguards that developers should deploy with Llama models or other LLMs, including Llama Guard 3, Prompt Guard and Code Shield. All our reference implementations demos contain these safeguards by default so developers can benefit from system-level safety out-of-the-box. Note that this release introduces new capabilities, including a longer context window, multilingual inputs and outputs and possible integrations by developers with third party tools. Building with these new capabilities requires specific considerations in addition to the best practices that generally apply across all Generative AI use cases. Tool-use: Just like in standard software development, developers are responsible for the integration of the LLM with the tools and services of their choice. They should define a clear policy for their use case and assess the integrity of the third party services they use to be aware of the safety and security limitations when using this capability. Refer to the Responsible Use Guide for best practices on the safe deployment of the third party safeguards. Multilinguality: Llama 3.3 supports 7 languages in addition to English: French, German, Hindi, Italian, Portuguese, Spanish, and Thai. Llama may be able to output text in other languages than those that meet performance thresholds for safety and helpfulness. We strongly discourage developers from using this model to converse in non-supported languages without implementing finetuning and system controls in alignment with their policies and the best practices shared in the Responsible Use Guide. We evaluated Llama models for common use cases as well as specific capabilities. Common use cases evaluations measure safety risks of systems for most commonly built applications including chat bot, coding assistant, tool calls. We built dedicated, adversarial evaluation datasets and evaluated systems composed of Llama models and Llama Guard 3 to filter input prompt and output response. It is important to evaluate applications in context, and we recommend building dedicated evaluation dataset for your use case. Prompt Guard and Code Shield are also available if relevant to the application. Capability evaluations measure vulnerabilities of Llama models inherent to specific capabilities, for which were crafted dedicated benchmarks including long context, multilingual, tools calls, coding or memorization. Red teaming For both scenarios, we conducted recurring red teaming exercises with the goal of discovering risks via adversarial prompting and we used the learnings to improve our benchmarks and safety tuning datasets. We partnered early with subject-matter experts in critical risk areas to understand the nature of these real-world harms and how such models may lead to unintended harm for society. Based on these conversations, we derived a set of adversarial goals for the red team to attempt to achieve, such as extracting harmful information or reprogramming the model to act in a potentially harmful capacity. The red team consisted of experts in cybersecurity, adversarial machine learning, responsible AI, and integrity in addition to multilingual content specialists with background in integrity issues in specific geographic markets. . We specifically focused our efforts on mitigating the following critical risk areas: 1- CBRNE (Chemical, Biological, Radiological, Nuclear, and Explosive materials) helpfulness To assess risks related to proliferation of chemical and biological weapons, we performed uplift testing designed to assess whether use of the Llama 3.3 model could meaningfully increase the capabilities of malicious actors to plan or carry out attacks using these types of weapons. Child Safety risk assessments were conducted using a team of experts, to assess the model’s capability to produce outputs that could result in Child Safety risks and inform on any necessary and appropriate risk mitigations via fine tuning. We leveraged those expert red teaming sessions to expand the coverage of our evaluation benchmarks through Llama 3 model development. For Llama 3, we conducted new in-depth sessions using objective based methodologies to assess the model risks along multiple attack vectors including the additional languages Llama 3 is trained on. We also partnered with content specialists to perform red teaming exercises assessing potentially violating content while taking account of market specific nuances or experiences. 3\. Cyber attack enablement Our cyber attack uplift study investigated whether LLMs can enhance human capabilities in hacking tasks, both in terms of skill level and speed. Our attack automation study focused on evaluating the capabilities of LLMs when used as autonomous agents in cyber offensive operations, specifically in the context of ransomware attacks. This evaluation was distinct from previous studies that considered LLMs as interactive assistants. The primary objective was to assess whether these models could effectively function as independent agents in executing complex cyber-attacks without human intervention. Generative AI safety requires expertise and tooling, and we believe in the strength of the open community to accelerate its progress. We are active members of open consortiums, including the AI Alliance, Partnership on AI and MLCommons, actively contributing to safety standardization and transparency. We encourage the community to adopt taxonomies like the MLCommons Proof of Concept evaluation to facilitate collaboration and transparency on safety and content evaluations. Our Purple Llama tools are open sourced for the community to use and widely distributed across ecosystem partners including cloud service providers. We encourage community contributions to our Github repository. We also set up the Llama Impact Grants program to identify and support the most compelling applications of Meta’s Llama model for societal benefit across three categories: education, climate and open innovation. The 20 finalists from the hundreds of applications can be found here. Finally, we put in place a set of resources including an output reporting mechanism and bug bounty program to continuously improve the Llama technology with the help of the community. The core values of Llama 3.3 are openness, inclusivity and helpfulness. It is meant to serve everyone, and to work for a wide range of use cases. It is thus designed to be accessible to people across many different backgrounds, experiences and perspectives. Llama 3.3 addresses users and their needs as they are, without insertion unnecessary judgment or normativity, while reflecting the understanding that even content that may appear problematic in some cases can serve valuable purposes in others. It respects the dignity and autonomy of all users, especially in terms of the values of free thought and expression that power innovation and progress. But Llama 3.3 is a new technology, and like any new technology, there are risks associated with its use. Testing conducted to date has not covered, nor could it cover, all scenarios. For these reasons, as with all LLMs, Llama 3.3’s potential outputs cannot be predicted in advance, and the model may in some instances produce inaccurate, biased or other objectionable responses to user prompts. Therefore, before deploying any applications of Llama 3.3 model, developers should perform safety testing and tuning tailored to their specific applications of the model. Please refer to available resources including our Responsible Use Guide, Trust and Safety solutions, and other resources to learn more about responsible development.

OrcaMiniv96Llama-3.2-3B-Instruct is trained with various SFT Datasets on Llama-3.2-3B-Instruct "Obsessed with Open Source GenAI's potential? So am I ! Let's Contribute together 🚀 https://www.linkedin.com/in/pankajam " NOTICE By providing proper credit and attribution, you are granted permission to use this model as a foundational base for further Full fine tuning, DPO, PPO or ORPO tuning and any kind of Merges. I actively encourage users to customize and enhance the model according to their specific needs, as this version is designed to be a comprehensive general model. Dive in and innovate! Use this model for Free on Google Colab with T4 GPU :) Below shows a code example on how to use this model in default half precision (bfloat16) format Below shows a code example on how to use this model in 4-bit format via bitsandbytes library Below shows a code example on how to use this model in 8-bit format via bitsandbytes library As part of our Responsible release approach, we followed a three-pronged strategy to managing trust & safety risks: 1. Enable developers to deploy helpful, safe and flexible experiences for their target audience and for the use cases supported by Llama 2. Protect developers against adversarial users aiming to exploit Llama capabilities to potentially cause harm 3. Provide protections for the community to help prevent the misuse of our models Approach: Llama is a foundational technology designed to be used in a variety of use cases. Examples on how Meta’s Llama models have been responsibly deployed can be found in our Community Stories webpage. Our approach is to build the most helpful models, enabling the world to benefit from the technology power, by aligning our model safety for generic use cases and addressing a standard set of harms. Developers are then in the driver’s seat to tailor safety for their use cases, defining their own policies and deploying the models with the necessary safeguards in their Llama systems. Llama 3.2 was developed following the best practices outlined in our Responsible Use Guide. Objective: Our main objectives for conducting safety fine-tuning are to provide the research community with a valuable resource for studying the robustness of safety fine-tuning, as well as to offer developers a readily available, safe, and powerful model for various applications to reduce the developer workload to deploy safe AI systems. We implemented the same set of safety mitigations as in Llama 3, and you can learn more about these in the Llama 3 paper. Fine-Tuning Data: We employ a multi-faceted approach to data collection, combining human-generated data from our vendors with synthetic data to mitigate potential safety risks. We’ve developed many large language model (LLM)-based classifiers that enable us to thoughtfully select high-quality prompts and responses, enhancing data quality control. Refusals and Tone: Building on the work we started with Llama 3, we put a great emphasis on model refusals to benign prompts as well as refusal tone. We included both borderline and adversarial prompts in our safety data strategy, and modified our safety data responses to follow tone guidelines. Safety as a System: Large language models, including Llama 3.2, are not designed to be deployed in isolation but instead should be deployed as part of an overall AI system with additional safety guardrails as required. Developers are expected to deploy system safeguards when building agentic systems. Safeguards are key to achieve the right helpfulness-safety alignment as well as mitigating safety and security risks inherent to the system and any integration of the model or system with external tools. As part of our responsible release approach, we provide the community with safeguards that developers should deploy with Llama models or other LLMs, including Llama Guard, Prompt Guard and Code Shield. All our reference implementations demos contain these safeguards by default so developers can benefit from system-level safety out-of-the-box. Technological Advancement: Llama releases usually introduce new capabilities that require specific considerations in addition to the best practices that generally apply across all Generative AI use cases. For prior release capabilities also supported by Llama 3.2, see Llama 3.1 Model Card, as the same considerations apply here as well. Constrained Environments: Llama 3.2 1B and 3B models are expected to be deployed in highly constrained environments, such as mobile devices. LLM Systems using smaller models will have a different alignment profile and safety/helpfulness tradeoff than more complex, larger systems. Developers should ensure the safety of their system meets the requirements of their use case. We recommend using lighter system safeguards for such use cases, like Llama Guard 3-1B or its mobile-optimized version. Scaled Evaluations: We built dedicated, adversarial evaluation datasets and evaluated systems composed of Llama models and Purple Llama safeguards to filter input prompt and output response. It is important to evaluate applications in context, and we recommend building dedicated evaluation dataset for your use case. Red Teaming: We conducted recurring red teaming exercises with the goal of discovering risks via adversarial prompting and we used the learnings to improve our benchmarks and safety tuning datasets. We partnered early with subject-matter experts in critical risk areas to understand the nature of these real-world harms and how such models may lead to unintended harm for society. Based on these conversations, we derived a set of adversarial goals for the red team to attempt to achieve, such as extracting harmful information or reprogramming the model to act in a potentially harmful capacity. The red team consisted of experts in cybersecurity, adversarial machine learning, responsible AI, and integrity in addition to multilingual content specialists with background in integrity issues in specific geographic markets. In addition to our safety work above, we took extra care on measuring and/or mitigating the following critical risk areas: 1\. CBRNE (Chemical, Biological, Radiological, Nuclear, and Explosive Weapons): Llama 3.2 1B and 3B models are smaller and less capable derivatives of Llama 3.1. For Llama 3.1 70B and 405B, to assess risks related to proliferation of chemical and biological weapons, we performed uplift testing designed to assess whether use of Llama 3.1 models could meaningfully increase the capabilities of malicious actors to plan or carry out attacks using these types of weapons and have determined that such testing also applies to the smaller 1B and 3B models. 2\. Child Safety: Child Safety risk assessments were conducted using a team of experts, to assess the model’s capability to produce outputs that could result in Child Safety risks and inform on any necessary and appropriate risk mitigations via fine tuning. We leveraged those expert red teaming sessions to expand the coverage of our evaluation benchmarks through Llama 3 model development. For Llama 3, we conducted new in-depth sessions using objective based methodologies to assess the model risks along multiple attack vectors including the additional languages Llama 3 is trained on. We also partnered with content specialists to perform red teaming exercises assessing potentially violating content while taking account of market specific nuances or experiences. 3\. Cyber Attacks: For Llama 3.1 405B, our cyber attack uplift study investigated whether LLMs can enhance human capabilities in hacking tasks, both in terms of skill level and speed. Our attack automation study focused on evaluating the capabilities of LLMs when used as autonomous agents in cyber offensive operations, specifically in the context of ransomware attacks. This evaluation was distinct from previous studies that considered LLMs as interactive assistants. The primary objective was to assess whether these models could effectively function as independent agents in executing complex cyber-attacks without human intervention. Because Llama 3.2’s 1B and 3B models are smaller and less capable models than Llama 3.1 405B, we broadly believe that the testing conducted for the 405B model also applies to Llama 3.2 models. Industry Partnerships: Generative AI safety requires expertise and tooling, and we believe in the strength of the open community to accelerate its progress. We are active members of open consortiums, including the AI Alliance, Partnership on AI and MLCommons, actively contributing to safety standardization and transparency. We encourage the community to adopt taxonomies like the MLCommons Proof of Concept evaluation to facilitate collaboration and transparency on safety and content evaluations. Our Purple Llama tools are open sourced for the community to use and widely distributed across ecosystem partners including cloud service providers. We encourage community contributions to our Github repository. Grants: We also set up the Llama Impact Grants program to identify and support the most compelling applications of Meta’s Llama model for societal benefit across three categories: education, climate and open innovation. The 20 finalists from the hundreds of applications can be found here. Reporting: Finally, we put in place a set of resources including an output reporting mechanism and bug bounty program to continuously improve the Llama technology with the help of the community. Values: The core values of Llama 3.2 are openness, inclusivity and helpfulness. It is meant to serve everyone, and to work for a wide range of use cases. It is thus designed to be accessible to people across many different backgrounds, experiences and perspectives. Llama 3.2 addresses users and their needs as they are, without insertion unnecessary judgment or normativity, while reflecting the understanding that even content that may appear problematic in some cases can serve valuable purposes in others. It respects the dignity and autonomy of all users, especially in terms of the values of free thought and expression that power innovation and progress. Testing: Llama 3.2 is a new technology, and like any new technology, there are risks associated with its use. Testing conducted to date has not covered, nor could it cover, all scenarios. For these reasons, as with all LLMs, Llama 3.2’s potential outputs cannot be predicted in advance, and the model may in some instances produce inaccurate, biased or other objectionable responses to user prompts. Therefore, before deploying any applications of Llama 3.2 models, developers should perform safety testing and tuning tailored to their specific applications of the model. Please refer to available resources including our Responsible Use Guide, Trust and Safety solutions, and other resources to learn more about responsible development.

OrcaMiniv97Llama-3.2-1B-Instruct is trained with various SFT Datasets on Llama-3.2-1B-Instruct "Obsessed with Open Source GenAI's potential? So am I ! Let's Contribute together 🚀 https://www.linkedin.com/in/pankajam " NOTICE By providing proper credit and attribution, you are granted permission to use this model as a foundational base for further Full fine tuning, DPO, PPO or ORPO tuning and any kind of Merges. I actively encourage users to customize and enhance the model according to their specific needs, as this version is designed to be a comprehensive general model. Dive in and innovate! Use this model for Free on Google Colab with T4 GPU :) Below shows a code example on how to use this model in default half precision (bfloat16) format Below shows a code example on how to use this model in 4-bit format via bitsandbytes library Below shows a code example on how to use this model in 8-bit format via bitsandbytes library As part of our Responsible release approach, we followed a three-pronged strategy to managing trust & safety risks: 1. Enable developers to deploy helpful, safe and flexible experiences for their target audience and for the use cases supported by Llama 2. Protect developers against adversarial users aiming to exploit Llama capabilities to potentially cause harm 3. Provide protections for the community to help prevent the misuse of our models Approach: Llama is a foundational technology designed to be used in a variety of use cases. Examples on how Meta’s Llama models have been responsibly deployed can be found in our Community Stories webpage. Our approach is to build the most helpful models, enabling the world to benefit from the technology power, by aligning our model safety for generic use cases and addressing a standard set of harms. Developers are then in the driver’s seat to tailor safety for their use cases, defining their own policies and deploying the models with the necessary safeguards in their Llama systems. Llama 3.2 was developed following the best practices outlined in our Responsible Use Guide. Objective: Our main objectives for conducting safety fine-tuning are to provide the research community with a valuable resource for studying the robustness of safety fine-tuning, as well as to offer developers a readily available, safe, and powerful model for various applications to reduce the developer workload to deploy safe AI systems. We implemented the same set of safety mitigations as in Llama 3, and you can learn more about these in the Llama 3 paper. Fine-Tuning Data: We employ a multi-faceted approach to data collection, combining human-generated data from our vendors with synthetic data to mitigate potential safety risks. We’ve developed many large language model (LLM)-based classifiers that enable us to thoughtfully select high-quality prompts and responses, enhancing data quality control. Refusals and Tone: Building on the work we started with Llama 3, we put a great emphasis on model refusals to benign prompts as well as refusal tone. We included both borderline and adversarial prompts in our safety data strategy, and modified our safety data responses to follow tone guidelines. Safety as a System: Large language models, including Llama 3.2, are not designed to be deployed in isolation but instead should be deployed as part of an overall AI system with additional safety guardrails as required. Developers are expected to deploy system safeguards when building agentic systems. Safeguards are key to achieve the right helpfulness-safety alignment as well as mitigating safety and security risks inherent to the system and any integration of the model or system with external tools. As part of our responsible release approach, we provide the community with safeguards that developers should deploy with Llama models or other LLMs, including Llama Guard, Prompt Guard and Code Shield. All our reference implementations demos contain these safeguards by default so developers can benefit from system-level safety out-of-the-box. Technological Advancement: Llama releases usually introduce new capabilities that require specific considerations in addition to the best practices that generally apply across all Generative AI use cases. For prior release capabilities also supported by Llama 3.2, see Llama 3.1 Model Card, as the same considerations apply here as well. Constrained Environments: Llama 3.2 1B and 3B models are expected to be deployed in highly constrained environments, such as mobile devices. LLM Systems using smaller models will have a different alignment profile and safety/helpfulness tradeoff than more complex, larger systems. Developers should ensure the safety of their system meets the requirements of their use case. We recommend using lighter system safeguards for such use cases, like Llama Guard 3-1B or its mobile-optimized version. Scaled Evaluations: We built dedicated, adversarial evaluation datasets and evaluated systems composed of Llama models and Purple Llama safeguards to filter input prompt and output response. It is important to evaluate applications in context, and we recommend building dedicated evaluation dataset for your use case. Red Teaming: We conducted recurring red teaming exercises with the goal of discovering risks via adversarial prompting and we used the learnings to improve our benchmarks and safety tuning datasets. We partnered early with subject-matter experts in critical risk areas to understand the nature of these real-world harms and how such models may lead to unintended harm for society. Based on these conversations, we derived a set of adversarial goals for the red team to attempt to achieve, such as extracting harmful information or reprogramming the model to act in a potentially harmful capacity. The red team consisted of experts in cybersecurity, adversarial machine learning, responsible AI, and integrity in addition to multilingual content specialists with background in integrity issues in specific geographic markets. In addition to our safety work above, we took extra care on measuring and/or mitigating the following critical risk areas: 1\. CBRNE (Chemical, Biological, Radiological, Nuclear, and Explosive Weapons): Llama 3.2 1B and 3B models are smaller and less capable derivatives of Llama 3.1. For Llama 3.1 70B and 405B, to assess risks related to proliferation of chemical and biological weapons, we performed uplift testing designed to assess whether use of Llama 3.1 models could meaningfully increase the capabilities of malicious actors to plan or carry out attacks using these types of weapons and have determined that such testing also applies to the smaller 1B and 3B models. 2\. Child Safety: Child Safety risk assessments were conducted using a team of experts, to assess the model’s capability to produce outputs that could result in Child Safety risks and inform on any necessary and appropriate risk mitigations via fine tuning. We leveraged those expert red teaming sessions to expand the coverage of our evaluation benchmarks through Llama 3 model development. For Llama 3, we conducted new in-depth sessions using objective based methodologies to assess the model risks along multiple attack vectors including the additional languages Llama 3 is trained on. We also partnered with content specialists to perform red teaming exercises assessing potentially violating content while taking account of market specific nuances or experiences. 3\. Cyber Attacks: For Llama 3.1 405B, our cyber attack uplift study investigated whether LLMs can enhance human capabilities in hacking tasks, both in terms of skill level and speed. Our attack automation study focused on evaluating the capabilities of LLMs when used as autonomous agents in cyber offensive operations, specifically in the context of ransomware attacks. This evaluation was distinct from previous studies that considered LLMs as interactive assistants. The primary objective was to assess whether these models could effectively function as independent agents in executing complex cyber-attacks without human intervention. Because Llama 3.2’s 1B and 3B models are smaller and less capable models than Llama 3.1 405B, we broadly believe that the testing conducted for the 405B model also applies to Llama 3.2 models. Industry Partnerships: Generative AI safety requires expertise and tooling, and we believe in the strength of the open community to accelerate its progress. We are active members of open consortiums, including the AI Alliance, Partnership on AI and MLCommons, actively contributing to safety standardization and transparency. We encourage the community to adopt taxonomies like the MLCommons Proof of Concept evaluation to facilitate collaboration and transparency on safety and content evaluations. Our Purple Llama tools are open sourced for the community to use and widely distributed across ecosystem partners including cloud service providers. We encourage community contributions to our Github repository. Grants: We also set up the Llama Impact Grants program to identify and support the most compelling applications of Meta’s Llama model for societal benefit across three categories: education, climate and open innovation. The 20 finalists from the hundreds of applications can be found here. Reporting: Finally, we put in place a set of resources including an output reporting mechanism and bug bounty program to continuously improve the Llama technology with the help of the community. Values: The core values of Llama 3.2 are openness, inclusivity and helpfulness. It is meant to serve everyone, and to work for a wide range of use cases. It is thus designed to be accessible to people across many different backgrounds, experiences and perspectives. Llama 3.2 addresses users and their needs as they are, without insertion unnecessary judgment or normativity, while reflecting the understanding that even content that may appear problematic in some cases can serve valuable purposes in others. It respects the dignity and autonomy of all users, especially in terms of the values of free thought and expression that power innovation and progress. Testing: Llama 3.2 is a new technology, and like any new technology, there are risks associated with its use. Testing conducted to date has not covered, nor could it cover, all scenarios. For these reasons, as with all LLMs, Llama 3.2’s potential outputs cannot be predicted in advance, and the model may in some instances produce inaccurate, biased or other objectionable responses to user prompts. Therefore, before deploying any applications of Llama 3.2 models, developers should perform safety testing and tuning tailored to their specific applications of the model. Please refer to available resources including our Responsible Use Guide, Trust and Safety solutions, and other resources to learn more about responsible development.

OrcaMiniv91Llama-3.2-1B-Instruct is trained with various SFT Datasets on Llama-3.2-1B-Instruct "Obsessed with Open Source GenAI's potential? So am I ! Let's Contribute together 🚀 https://www.linkedin.com/in/pankajam " NOTICE By providing proper credit and attribution, you are granted permission to use this model as a foundational base for further Full fine tuning, DPO, PPO or ORPO tuning and any kind of Merges. I actively encourage users to customize and enhance the model according to their specific needs, as this version is designed to be a comprehensive general model. Dive in and innovate! Below shows a code example on how to use this model in default half precision (bfloat16) format Below shows a code example on how to use this model in 4-bit format via bitsandbytes library Below shows a code example on how to use this model in 8-bit format via bitsandbytes library As part of our Responsible release approach, we followed a three-pronged strategy to managing trust & safety risks: 1. Enable developers to deploy helpful, safe and flexible experiences for their target audience and for the use cases supported by Llama 2. Protect developers against adversarial users aiming to exploit Llama capabilities to potentially cause harm 3. Provide protections for the community to help prevent the misuse of our models Approach: Llama is a foundational technology designed to be used in a variety of use cases. Examples on how Meta’s Llama models have been responsibly deployed can be found in our Community Stories webpage. Our approach is to build the most helpful models, enabling the world to benefit from the technology power, by aligning our model safety for generic use cases and addressing a standard set of harms. Developers are then in the driver’s seat to tailor safety for their use cases, defining their own policies and deploying the models with the necessary safeguards in their Llama systems. Llama 3.2 was developed following the best practices outlined in our Responsible Use Guide. Objective: Our main objectives for conducting safety fine-tuning are to provide the research community with a valuable resource for studying the robustness of safety fine-tuning, as well as to offer developers a readily available, safe, and powerful model for various applications to reduce the developer workload to deploy safe AI systems. We implemented the same set of safety mitigations as in Llama 3, and you can learn more about these in the Llama 3 paper. Fine-Tuning Data: We employ a multi-faceted approach to data collection, combining human-generated data from our vendors with synthetic data to mitigate potential safety risks. We’ve developed many large language model (LLM)-based classifiers that enable us to thoughtfully select high-quality prompts and responses, enhancing data quality control. Refusals and Tone: Building on the work we started with Llama 3, we put a great emphasis on model refusals to benign prompts as well as refusal tone. We included both borderline and adversarial prompts in our safety data strategy, and modified our safety data responses to follow tone guidelines. Safety as a System: Large language models, including Llama 3.2, are not designed to be deployed in isolation but instead should be deployed as part of an overall AI system with additional safety guardrails as required. Developers are expected to deploy system safeguards when building agentic systems. Safeguards are key to achieve the right helpfulness-safety alignment as well as mitigating safety and security risks inherent to the system and any integration of the model or system with external tools. As part of our responsible release approach, we provide the community with safeguards that developers should deploy with Llama models or other LLMs, including Llama Guard, Prompt Guard and Code Shield. All our reference implementations demos contain these safeguards by default so developers can benefit from system-level safety out-of-the-box. Technological Advancement: Llama releases usually introduce new capabilities that require specific considerations in addition to the best practices that generally apply across all Generative AI use cases. For prior release capabilities also supported by Llama 3.2, see Llama 3.1 Model Card, as the same considerations apply here as well. Constrained Environments: Llama 3.2 1B and 3B models are expected to be deployed in highly constrained environments, such as mobile devices. LLM Systems using smaller models will have a different alignment profile and safety/helpfulness tradeoff than more complex, larger systems. Developers should ensure the safety of their system meets the requirements of their use case. We recommend using lighter system safeguards for such use cases, like Llama Guard 3-1B or its mobile-optimized version. Scaled Evaluations: We built dedicated, adversarial evaluation datasets and evaluated systems composed of Llama models and Purple Llama safeguards to filter input prompt and output response. It is important to evaluate applications in context, and we recommend building dedicated evaluation dataset for your use case. Red Teaming: We conducted recurring red teaming exercises with the goal of discovering risks via adversarial prompting and we used the learnings to improve our benchmarks and safety tuning datasets. We partnered early with subject-matter experts in critical risk areas to understand the nature of these real-world harms and how such models may lead to unintended harm for society. Based on these conversations, we derived a set of adversarial goals for the red team to attempt to achieve, such as extracting harmful information or reprogramming the model to act in a potentially harmful capacity. The red team consisted of experts in cybersecurity, adversarial machine learning, responsible AI, and integrity in addition to multilingual content specialists with background in integrity issues in specific geographic markets. In addition to our safety work above, we took extra care on measuring and/or mitigating the following critical risk areas: 1\. CBRNE (Chemical, Biological, Radiological, Nuclear, and Explosive Weapons): Llama 3.2 1B and 3B models are smaller and less capable derivatives of Llama 3.1. For Llama 3.1 70B and 405B, to assess risks related to proliferation of chemical and biological weapons, we performed uplift testing designed to assess whether use of Llama 3.1 models could meaningfully increase the capabilities of malicious actors to plan or carry out attacks using these types of weapons and have determined that such testing also applies to the smaller 1B and 3B models. 2\. Child Safety: Child Safety risk assessments were conducted using a team of experts, to assess the model’s capability to produce outputs that could result in Child Safety risks and inform on any necessary and appropriate risk mitigations via fine tuning. We leveraged those expert red teaming sessions to expand the coverage of our evaluation benchmarks through Llama 3 model development. For Llama 3, we conducted new in-depth sessions using objective based methodologies to assess the model risks along multiple attack vectors including the additional languages Llama 3 is trained on. We also partnered with content specialists to perform red teaming exercises assessing potentially violating content while taking account of market specific nuances or experiences. 3\. Cyber Attacks: For Llama 3.1 405B, our cyber attack uplift study investigated whether LLMs can enhance human capabilities in hacking tasks, both in terms of skill level and speed. Our attack automation study focused on evaluating the capabilities of LLMs when used as autonomous agents in cyber offensive operations, specifically in the context of ransomware attacks. This evaluation was distinct from previous studies that considered LLMs as interactive assistants. The primary objective was to assess whether these models could effectively function as independent agents in executing complex cyber-attacks without human intervention. Because Llama 3.2’s 1B and 3B models are smaller and less capable models than Llama 3.1 405B, we broadly believe that the testing conducted for the 405B model also applies to Llama 3.2 models. Industry Partnerships: Generative AI safety requires expertise and tooling, and we believe in the strength of the open community to accelerate its progress. We are active members of open consortiums, including the AI Alliance, Partnership on AI and MLCommons, actively contributing to safety standardization and transparency. We encourage the community to adopt taxonomies like the MLCommons Proof of Concept evaluation to facilitate collaboration and transparency on safety and content evaluations. Our Purple Llama tools are open sourced for the community to use and widely distributed across ecosystem partners including cloud service providers. We encourage community contributions to our Github repository. Grants: We also set up the Llama Impact Grants program to identify and support the most compelling applications of Meta’s Llama model for societal benefit across three categories: education, climate and open innovation. The 20 finalists from the hundreds of applications can be found here. Reporting: Finally, we put in place a set of resources including an output reporting mechanism and bug bounty program to continuously improve the Llama technology with the help of the community. Values: The core values of Llama 3.2 are openness, inclusivity and helpfulness. It is meant to serve everyone, and to work for a wide range of use cases. It is thus designed to be accessible to people across many different backgrounds, experiences and perspectives. Llama 3.2 addresses users and their needs as they are, without insertion unnecessary judgment or normativity, while reflecting the understanding that even content that may appear problematic in some cases can serve valuable purposes in others. It respects the dignity and autonomy of all users, especially in terms of the values of free thought and expression that power innovation and progress. Testing: Llama 3.2 is a new technology, and like any new technology, there are risks associated with its use. Testing conducted to date has not covered, nor could it cover, all scenarios. For these reasons, as with all LLMs, Llama 3.2’s potential outputs cannot be predicted in advance, and the model may in some instances produce inaccurate, biased or other objectionable responses to user prompts. Therefore, before deploying any applications of Llama 3.2 models, developers should perform safety testing and tuning tailored to their specific applications of the model. Please refer to available resources including our Responsible Use Guide, Trust and Safety solutions, and other resources to learn more about responsible development.

llama3orcaminiv58b trained with various DPO Datasets "Obsessed with GenAI's potential? So am I ! Let's create together 🚀 https://www.linkedin.com/in/pankajam " By providing proper credit and attribution, you are granted permission to use this model as a foundational base for further Full fine tuning, DPO, PPO or ORPO tuning and any kind of Merges. I actively encourage users to customize and enhance the model according to their specific needs, as this version is designed to be a comprehensive general model. Dive in and innovate! | Metric |Value| |---------------------------------|----:| |Avg. |67.78| |AI2 Reasoning Challenge (25-Shot)|61.86| |HellaSwag (10-Shot) |82.35| |MMLU (5-Shot) |65.10| |TruthfulQA (0-shot) |56.24| |Winogrande (5-shot) |73.40| |GSM8k (5-shot) |67.70| Below shows a code example on how to use this model This model is governed by META LLAMA 3 COMMUNITY LICENSE AGREEMENT Open LLM Leaderboard Evaluation Results Detailed results can be found here | Metric |Value| |-------------------|----:| |Avg. |19.96| |IFEval (0-Shot) |48.96| |BBH (3-Shot) |29.61| |MATH Lvl 5 (4-Shot)| 7.48| |GPQA (0-shot) | 3.24| |MuSR (0-shot) | 6.94| |MMLU-PRO (5-shot) |23.51|

OrcaMiniv90Llama-3.2-3B-Instruct is trained with various SFT Datasets on Llama-3.2-3B-Instruct "Obsessed with Open Source GenAI's potential? So am I ! Let's Contribute together 🚀 https://www.linkedin.com/in/pankajam " NOTICE By providing proper credit and attribution, you are granted permission to use this model as a foundational base for further Full fine tuning, DPO, PPO or ORPO tuning and any kind of Merges. I actively encourage users to customize and enhance the model according to their specific needs, as this version is designed to be a comprehensive general model. Dive in and innovate! Below shows a code example on how to use this model in default half precision (bfloat16) format Below shows a code example on how to use this model in 4-bit format via bitsandbytes library Below shows a code example on how to use this model in 8-bit format via bitsandbytes library As part of our Responsible release approach, we followed a three-pronged strategy to managing trust & safety risks: 1. Enable developers to deploy helpful, safe and flexible experiences for their target audience and for the use cases supported by Llama 2. Protect developers against adversarial users aiming to exploit Llama capabilities to potentially cause harm 3. Provide protections for the community to help prevent the misuse of our models Approach: Llama is a foundational technology designed to be used in a variety of use cases. Examples on how Meta’s Llama models have been responsibly deployed can be found in our Community Stories webpage. Our approach is to build the most helpful models, enabling the world to benefit from the technology power, by aligning our model safety for generic use cases and addressing a standard set of harms. Developers are then in the driver’s seat to tailor safety for their use cases, defining their own policies and deploying the models with the necessary safeguards in their Llama systems. Llama 3.2 was developed following the best practices outlined in our Responsible Use Guide. Objective: Our main objectives for conducting safety fine-tuning are to provide the research community with a valuable resource for studying the robustness of safety fine-tuning, as well as to offer developers a readily available, safe, and powerful model for various applications to reduce the developer workload to deploy safe AI systems. We implemented the same set of safety mitigations as in Llama 3, and you can learn more about these in the Llama 3 paper. Fine-Tuning Data: We employ a multi-faceted approach to data collection, combining human-generated data from our vendors with synthetic data to mitigate potential safety risks. We’ve developed many large language model (LLM)-based classifiers that enable us to thoughtfully select high-quality prompts and responses, enhancing data quality control. Refusals and Tone: Building on the work we started with Llama 3, we put a great emphasis on model refusals to benign prompts as well as refusal tone. We included both borderline and adversarial prompts in our safety data strategy, and modified our safety data responses to follow tone guidelines. Safety as a System: Large language models, including Llama 3.2, are not designed to be deployed in isolation but instead should be deployed as part of an overall AI system with additional safety guardrails as required. Developers are expected to deploy system safeguards when building agentic systems. Safeguards are key to achieve the right helpfulness-safety alignment as well as mitigating safety and security risks inherent to the system and any integration of the model or system with external tools. As part of our responsible release approach, we provide the community with safeguards that developers should deploy with Llama models or other LLMs, including Llama Guard, Prompt Guard and Code Shield. All our reference implementations demos contain these safeguards by default so developers can benefit from system-level safety out-of-the-box. Technological Advancement: Llama releases usually introduce new capabilities that require specific considerations in addition to the best practices that generally apply across all Generative AI use cases. For prior release capabilities also supported by Llama 3.2, see Llama 3.1 Model Card, as the same considerations apply here as well. Constrained Environments: Llama 3.2 1B and 3B models are expected to be deployed in highly constrained environments, such as mobile devices. LLM Systems using smaller models will have a different alignment profile and safety/helpfulness tradeoff than more complex, larger systems. Developers should ensure the safety of their system meets the requirements of their use case. We recommend using lighter system safeguards for such use cases, like Llama Guard 3-1B or its mobile-optimized version. Scaled Evaluations: We built dedicated, adversarial evaluation datasets and evaluated systems composed of Llama models and Purple Llama safeguards to filter input prompt and output response. It is important to evaluate applications in context, and we recommend building dedicated evaluation dataset for your use case. Red Teaming: We conducted recurring red teaming exercises with the goal of discovering risks via adversarial prompting and we used the learnings to improve our benchmarks and safety tuning datasets. We partnered early with subject-matter experts in critical risk areas to understand the nature of these real-world harms and how such models may lead to unintended harm for society. Based on these conversations, we derived a set of adversarial goals for the red team to attempt to achieve, such as extracting harmful information or reprogramming the model to act in a potentially harmful capacity. The red team consisted of experts in cybersecurity, adversarial machine learning, responsible AI, and integrity in addition to multilingual content specialists with background in integrity issues in specific geographic markets. In addition to our safety work above, we took extra care on measuring and/or mitigating the following critical risk areas: 1\. CBRNE (Chemical, Biological, Radiological, Nuclear, and Explosive Weapons): Llama 3.2 1B and 3B models are smaller and less capable derivatives of Llama 3.1. For Llama 3.1 70B and 405B, to assess risks related to proliferation of chemical and biological weapons, we performed uplift testing designed to assess whether use of Llama 3.1 models could meaningfully increase the capabilities of malicious actors to plan or carry out attacks using these types of weapons and have determined that such testing also applies to the smaller 1B and 3B models. 2\. Child Safety: Child Safety risk assessments were conducted using a team of experts, to assess the model’s capability to produce outputs that could result in Child Safety risks and inform on any necessary and appropriate risk mitigations via fine tuning. We leveraged those expert red teaming sessions to expand the coverage of our evaluation benchmarks through Llama 3 model development. For Llama 3, we conducted new in-depth sessions using objective based methodologies to assess the model risks along multiple attack vectors including the additional languages Llama 3 is trained on. We also partnered with content specialists to perform red teaming exercises assessing potentially violating content while taking account of market specific nuances or experiences. 3\. Cyber Attacks: For Llama 3.1 405B, our cyber attack uplift study investigated whether LLMs can enhance human capabilities in hacking tasks, both in terms of skill level and speed. Our attack automation study focused on evaluating the capabilities of LLMs when used as autonomous agents in cyber offensive operations, specifically in the context of ransomware attacks. This evaluation was distinct from previous studies that considered LLMs as interactive assistants. The primary objective was to assess whether these models could effectively function as independent agents in executing complex cyber-attacks without human intervention. Because Llama 3.2’s 1B and 3B models are smaller and less capable models than Llama 3.1 405B, we broadly believe that the testing conducted for the 405B model also applies to Llama 3.2 models. Industry Partnerships: Generative AI safety requires expertise and tooling, and we believe in the strength of the open community to accelerate its progress. We are active members of open consortiums, including the AI Alliance, Partnership on AI and MLCommons, actively contributing to safety standardization and transparency. We encourage the community to adopt taxonomies like the MLCommons Proof of Concept evaluation to facilitate collaboration and transparency on safety and content evaluations. Our Purple Llama tools are open sourced for the community to use and widely distributed across ecosystem partners including cloud service providers. We encourage community contributions to our Github repository. Grants: We also set up the Llama Impact Grants program to identify and support the most compelling applications of Meta’s Llama model for societal benefit across three categories: education, climate and open innovation. The 20 finalists from the hundreds of applications can be found here. Reporting: Finally, we put in place a set of resources including an output reporting mechanism and bug bounty program to continuously improve the Llama technology with the help of the community. Values: The core values of Llama 3.2 are openness, inclusivity and helpfulness. It is meant to serve everyone, and to work for a wide range of use cases. It is thus designed to be accessible to people across many different backgrounds, experiences and perspectives. Llama 3.2 addresses users and their needs as they are, without insertion unnecessary judgment or normativity, while reflecting the understanding that even content that may appear problematic in some cases can serve valuable purposes in others. It respects the dignity and autonomy of all users, especially in terms of the values of free thought and expression that power innovation and progress. Testing: Llama 3.2 is a new technology, and like any new technology, there are risks associated with its use. Testing conducted to date has not covered, nor could it cover, all scenarios. For these reasons, as with all LLMs, Llama 3.2’s potential outputs cannot be predicted in advance, and the model may in some instances produce inaccurate, biased or other objectionable responses to user prompts. Therefore, before deploying any applications of Llama 3.2 models, developers should perform safety testing and tuning tailored to their specific applications of the model. Please refer to available resources including our Responsible Use Guide, Trust and Safety solutions, and other resources to learn more about responsible development.

OrcaMiniv97Llama-3.2-3B-Instruct is trained with various SFT Datasets on Llama-3.2-3B-Instruct "Obsessed with Open Source GenAI's potential? So am I ! Let's Contribute together 🚀 https://www.linkedin.com/in/pankajam " NOTICE By providing proper credit and attribution, you are granted permission to use this model as a foundational base for further Full fine tuning, DPO, PPO or ORPO tuning and any kind of Merges. I actively encourage users to customize and enhance the model according to their specific needs, as this version is designed to be a comprehensive general model. Dive in and innovate! Use this model for Free on Google Colab with T4 GPU :) Below shows a code example on how to use this model in default half precision (bfloat16) format Below shows a code example on how to use this model in 4-bit format via bitsandbytes library Below shows a code example on how to use this model in 8-bit format via bitsandbytes library As part of our Responsible release approach, we followed a three-pronged strategy to managing trust & safety risks: 1. Enable developers to deploy helpful, safe and flexible experiences for their target audience and for the use cases supported by Llama 2. Protect developers against adversarial users aiming to exploit Llama capabilities to potentially cause harm 3. Provide protections for the community to help prevent the misuse of our models Approach: Llama is a foundational technology designed to be used in a variety of use cases. Examples on how Meta’s Llama models have been responsibly deployed can be found in our Community Stories webpage. Our approach is to build the most helpful models, enabling the world to benefit from the technology power, by aligning our model safety for generic use cases and addressing a standard set of harms. Developers are then in the driver’s seat to tailor safety for their use cases, defining their own policies and deploying the models with the necessary safeguards in their Llama systems. Llama 3.2 was developed following the best practices outlined in our Responsible Use Guide. Objective: Our main objectives for conducting safety fine-tuning are to provide the research community with a valuable resource for studying the robustness of safety fine-tuning, as well as to offer developers a readily available, safe, and powerful model for various applications to reduce the developer workload to deploy safe AI systems. We implemented the same set of safety mitigations as in Llama 3, and you can learn more about these in the Llama 3 paper. Fine-Tuning Data: We employ a multi-faceted approach to data collection, combining human-generated data from our vendors with synthetic data to mitigate potential safety risks. We’ve developed many large language model (LLM)-based classifiers that enable us to thoughtfully select high-quality prompts and responses, enhancing data quality control. Refusals and Tone: Building on the work we started with Llama 3, we put a great emphasis on model refusals to benign prompts as well as refusal tone. We included both borderline and adversarial prompts in our safety data strategy, and modified our safety data responses to follow tone guidelines. Safety as a System: Large language models, including Llama 3.2, are not designed to be deployed in isolation but instead should be deployed as part of an overall AI system with additional safety guardrails as required. Developers are expected to deploy system safeguards when building agentic systems. Safeguards are key to achieve the right helpfulness-safety alignment as well as mitigating safety and security risks inherent to the system and any integration of the model or system with external tools. As part of our responsible release approach, we provide the community with safeguards that developers should deploy with Llama models or other LLMs, including Llama Guard, Prompt Guard and Code Shield. All our reference implementations demos contain these safeguards by default so developers can benefit from system-level safety out-of-the-box. Technological Advancement: Llama releases usually introduce new capabilities that require specific considerations in addition to the best practices that generally apply across all Generative AI use cases. For prior release capabilities also supported by Llama 3.2, see Llama 3.1 Model Card, as the same considerations apply here as well. Constrained Environments: Llama 3.2 1B and 3B models are expected to be deployed in highly constrained environments, such as mobile devices. LLM Systems using smaller models will have a different alignment profile and safety/helpfulness tradeoff than more complex, larger systems. Developers should ensure the safety of their system meets the requirements of their use case. We recommend using lighter system safeguards for such use cases, like Llama Guard 3-1B or its mobile-optimized version. Scaled Evaluations: We built dedicated, adversarial evaluation datasets and evaluated systems composed of Llama models and Purple Llama safeguards to filter input prompt and output response. It is important to evaluate applications in context, and we recommend building dedicated evaluation dataset for your use case. Red Teaming: We conducted recurring red teaming exercises with the goal of discovering risks via adversarial prompting and we used the learnings to improve our benchmarks and safety tuning datasets. We partnered early with subject-matter experts in critical risk areas to understand the nature of these real-world harms and how such models may lead to unintended harm for society. Based on these conversations, we derived a set of adversarial goals for the red team to attempt to achieve, such as extracting harmful information or reprogramming the model to act in a potentially harmful capacity. The red team consisted of experts in cybersecurity, adversarial machine learning, responsible AI, and integrity in addition to multilingual content specialists with background in integrity issues in specific geographic markets. In addition to our safety work above, we took extra care on measuring and/or mitigating the following critical risk areas: 1\. CBRNE (Chemical, Biological, Radiological, Nuclear, and Explosive Weapons): Llama 3.2 1B and 3B models are smaller and less capable derivatives of Llama 3.1. For Llama 3.1 70B and 405B, to assess risks related to proliferation of chemical and biological weapons, we performed uplift testing designed to assess whether use of Llama 3.1 models could meaningfully increase the capabilities of malicious actors to plan or carry out attacks using these types of weapons and have determined that such testing also applies to the smaller 1B and 3B models. 2\. Child Safety: Child Safety risk assessments were conducted using a team of experts, to assess the model’s capability to produce outputs that could result in Child Safety risks and inform on any necessary and appropriate risk mitigations via fine tuning. We leveraged those expert red teaming sessions to expand the coverage of our evaluation benchmarks through Llama 3 model development. For Llama 3, we conducted new in-depth sessions using objective based methodologies to assess the model risks along multiple attack vectors including the additional languages Llama 3 is trained on. We also partnered with content specialists to perform red teaming exercises assessing potentially violating content while taking account of market specific nuances or experiences. 3\. Cyber Attacks: For Llama 3.1 405B, our cyber attack uplift study investigated whether LLMs can enhance human capabilities in hacking tasks, both in terms of skill level and speed. Our attack automation study focused on evaluating the capabilities of LLMs when used as autonomous agents in cyber offensive operations, specifically in the context of ransomware attacks. This evaluation was distinct from previous studies that considered LLMs as interactive assistants. The primary objective was to assess whether these models could effectively function as independent agents in executing complex cyber-attacks without human intervention. Because Llama 3.2’s 1B and 3B models are smaller and less capable models than Llama 3.1 405B, we broadly believe that the testing conducted for the 405B model also applies to Llama 3.2 models. Industry Partnerships: Generative AI safety requires expertise and tooling, and we believe in the strength of the open community to accelerate its progress. We are active members of open consortiums, including the AI Alliance, Partnership on AI and MLCommons, actively contributing to safety standardization and transparency. We encourage the community to adopt taxonomies like the MLCommons Proof of Concept evaluation to facilitate collaboration and transparency on safety and content evaluations. Our Purple Llama tools are open sourced for the community to use and widely distributed across ecosystem partners including cloud service providers. We encourage community contributions to our Github repository. Grants: We also set up the Llama Impact Grants program to identify and support the most compelling applications of Meta’s Llama model for societal benefit across three categories: education, climate and open innovation. The 20 finalists from the hundreds of applications can be found here. Reporting: Finally, we put in place a set of resources including an output reporting mechanism and bug bounty program to continuously improve the Llama technology with the help of the community. Values: The core values of Llama 3.2 are openness, inclusivity and helpfulness. It is meant to serve everyone, and to work for a wide range of use cases. It is thus designed to be accessible to people across many different backgrounds, experiences and perspectives. Llama 3.2 addresses users and their needs as they are, without insertion unnecessary judgment or normativity, while reflecting the understanding that even content that may appear problematic in some cases can serve valuable purposes in others. It respects the dignity and autonomy of all users, especially in terms of the values of free thought and expression that power innovation and progress. Testing: Llama 3.2 is a new technology, and like any new technology, there are risks associated with its use. Testing conducted to date has not covered, nor could it cover, all scenarios. For these reasons, as with all LLMs, Llama 3.2’s potential outputs cannot be predicted in advance, and the model may in some instances produce inaccurate, biased or other objectionable responses to user prompts. Therefore, before deploying any applications of Llama 3.2 models, developers should perform safety testing and tuning tailored to their specific applications of the model. Please refer to available resources including our Responsible Use Guide, Trust and Safety solutions, and other resources to learn more about responsible development.

Model Name: OrcaMiniv95Llama-3.2-1B-Instructpreview OrcaMiniv95Llama-3.2-1B-Instructpreview is trained with various SFT Datasets on Llama-3.2-1B-Instruct "Obsessed with Open Source GenAI's potential? So am I ! Let's Contribute together 🚀 https://www.linkedin.com/in/pankajam " NOTICE By providing proper credit and attribution, you are granted permission to use this model as a foundational base for further Full fine tuning, DPO, PPO or ORPO tuning and any kind of Merges. I actively encourage users to customize and enhance the model according to their specific needs, as this version is designed to be a comprehensive general model. Dive in and innovate! Below shows a code example on how to use this model in default half precision (bfloat16) format Below shows a code example on how to use this model in 4-bit format via bitsandbytes library Below shows a code example on how to use this model in 8-bit format via bitsandbytes library As part of our Responsible release approach, we followed a three-pronged strategy to managing trust & safety risks: 1. Enable developers to deploy helpful, safe and flexible experiences for their target audience and for the use cases supported by Llama 2. Protect developers against adversarial users aiming to exploit Llama capabilities to potentially cause harm 3. Provide protections for the community to help prevent the misuse of our models Approach: Llama is a foundational technology designed to be used in a variety of use cases. Examples on how Meta’s Llama models have been responsibly deployed can be found in our Community Stories webpage. Our approach is to build the most helpful models, enabling the world to benefit from the technology power, by aligning our model safety for generic use cases and addressing a standard set of harms. Developers are then in the driver’s seat to tailor safety for their use cases, defining their own policies and deploying the models with the necessary safeguards in their Llama systems. Llama 3.2 was developed following the best practices outlined in our Responsible Use Guide. Objective: Our main objectives for conducting safety fine-tuning are to provide the research community with a valuable resource for studying the robustness of safety fine-tuning, as well as to offer developers a readily available, safe, and powerful model for various applications to reduce the developer workload to deploy safe AI systems. We implemented the same set of safety mitigations as in Llama 3, and you can learn more about these in the Llama 3 paper. Fine-Tuning Data: We employ a multi-faceted approach to data collection, combining human-generated data from our vendors with synthetic data to mitigate potential safety risks. We’ve developed many large language model (LLM)-based classifiers that enable us to thoughtfully select high-quality prompts and responses, enhancing data quality control. Refusals and Tone: Building on the work we started with Llama 3, we put a great emphasis on model refusals to benign prompts as well as refusal tone. We included both borderline and adversarial prompts in our safety data strategy, and modified our safety data responses to follow tone guidelines. Safety as a System: Large language models, including Llama 3.2, are not designed to be deployed in isolation but instead should be deployed as part of an overall AI system with additional safety guardrails as required. Developers are expected to deploy system safeguards when building agentic systems. Safeguards are key to achieve the right helpfulness-safety alignment as well as mitigating safety and security risks inherent to the system and any integration of the model or system with external tools. As part of our responsible release approach, we provide the community with safeguards that developers should deploy with Llama models or other LLMs, including Llama Guard, Prompt Guard and Code Shield. All our reference implementations demos contain these safeguards by default so developers can benefit from system-level safety out-of-the-box. Technological Advancement: Llama releases usually introduce new capabilities that require specific considerations in addition to the best practices that generally apply across all Generative AI use cases. For prior release capabilities also supported by Llama 3.2, see Llama 3.1 Model Card, as the same considerations apply here as well. Constrained Environments: Llama 3.2 1B and 3B models are expected to be deployed in highly constrained environments, such as mobile devices. LLM Systems using smaller models will have a different alignment profile and safety/helpfulness tradeoff than more complex, larger systems. Developers should ensure the safety of their system meets the requirements of their use case. We recommend using lighter system safeguards for such use cases, like Llama Guard 3-1B or its mobile-optimized version. Scaled Evaluations: We built dedicated, adversarial evaluation datasets and evaluated systems composed of Llama models and Purple Llama safeguards to filter input prompt and output response. It is important to evaluate applications in context, and we recommend building dedicated evaluation dataset for your use case. Red Teaming: We conducted recurring red teaming exercises with the goal of discovering risks via adversarial prompting and we used the learnings to improve our benchmarks and safety tuning datasets. We partnered early with subject-matter experts in critical risk areas to understand the nature of these real-world harms and how such models may lead to unintended harm for society. Based on these conversations, we derived a set of adversarial goals for the red team to attempt to achieve, such as extracting harmful information or reprogramming the model to act in a potentially harmful capacity. The red team consisted of experts in cybersecurity, adversarial machine learning, responsible AI, and integrity in addition to multilingual content specialists with background in integrity issues in specific geographic markets. In addition to our safety work above, we took extra care on measuring and/or mitigating the following critical risk areas: 1\. CBRNE (Chemical, Biological, Radiological, Nuclear, and Explosive Weapons): Llama 3.2 1B and 3B models are smaller and less capable derivatives of Llama 3.1. For Llama 3.1 70B and 405B, to assess risks related to proliferation of chemical and biological weapons, we performed uplift testing designed to assess whether use of Llama 3.1 models could meaningfully increase the capabilities of malicious actors to plan or carry out attacks using these types of weapons and have determined that such testing also applies to the smaller 1B and 3B models. 2\. Child Safety: Child Safety risk assessments were conducted using a team of experts, to assess the model’s capability to produce outputs that could result in Child Safety risks and inform on any necessary and appropriate risk mitigations via fine tuning. We leveraged those expert red teaming sessions to expand the coverage of our evaluation benchmarks through Llama 3 model development. For Llama 3, we conducted new in-depth sessions using objective based methodologies to assess the model risks along multiple attack vectors including the additional languages Llama 3 is trained on. We also partnered with content specialists to perform red teaming exercises assessing potentially violating content while taking account of market specific nuances or experiences. 3\. Cyber Attacks: For Llama 3.1 405B, our cyber attack uplift study investigated whether LLMs can enhance human capabilities in hacking tasks, both in terms of skill level and speed. Our attack automation study focused on evaluating the capabilities of LLMs when used as autonomous agents in cyber offensive operations, specifically in the context of ransomware attacks. This evaluation was distinct from previous studies that considered LLMs as interactive assistants. The primary objective was to assess whether these models could effectively function as independent agents in executing complex cyber-attacks without human intervention. Because Llama 3.2’s 1B and 3B models are smaller and less capable models than Llama 3.1 405B, we broadly believe that the testing conducted for the 405B model also applies to Llama 3.2 models. Industry Partnerships: Generative AI safety requires expertise and tooling, and we believe in the strength of the open community to accelerate its progress. We are active members of open consortiums, including the AI Alliance, Partnership on AI and MLCommons, actively contributing to safety standardization and transparency. We encourage the community to adopt taxonomies like the MLCommons Proof of Concept evaluation to facilitate collaboration and transparency on safety and content evaluations. Our Purple Llama tools are open sourced for the community to use and widely distributed across ecosystem partners including cloud service providers. We encourage community contributions to our Github repository. Grants: We also set up the Llama Impact Grants program to identify and support the most compelling applications of Meta’s Llama model for societal benefit across three categories: education, climate and open innovation. The 20 finalists from the hundreds of applications can be found here. Reporting: Finally, we put in place a set of resources including an output reporting mechanism and bug bounty program to continuously improve the Llama technology with the help of the community. Values: The core values of Llama 3.2 are openness, inclusivity and helpfulness. It is meant to serve everyone, and to work for a wide range of use cases. It is thus designed to be accessible to people across many different backgrounds, experiences and perspectives. Llama 3.2 addresses users and their needs as they are, without insertion unnecessary judgment or normativity, while reflecting the understanding that even content that may appear problematic in some cases can serve valuable purposes in others. It respects the dignity and autonomy of all users, especially in terms of the values of free thought and expression that power innovation and progress. Testing: Llama 3.2 is a new technology, and like any new technology, there are risks associated with its use. Testing conducted to date has not covered, nor could it cover, all scenarios. For these reasons, as with all LLMs, Llama 3.2’s potential outputs cannot be predicted in advance, and the model may in some instances produce inaccurate, biased or other objectionable responses to user prompts. Therefore, before deploying any applications of Llama 3.2 models, developers should perform safety testing and tuning tailored to their specific applications of the model. Please refer to available resources including our Responsible Use Guide, Trust and Safety solutions, and other resources to learn more about responsible development.

Llama-3-8b base model trained on Orca Style Mini Datasets "Obsessed with GenAI's potential? So am I ! Let's create together 🚀 https://www.linkedin.com/in/pankajam " By providing proper credit and attribution, you are granted permission to use this model as a foundational base for further DPO/PPO tuning or Merges. I actively encourage users to customize and enhance the model according to their specific needs, as this version is designed to be a comprehensive, fully fine-tuned general model. Dive in and innovate! We evaluated this model on a wide range of tasks using Language Model Evaluation Harness from EleutherAI. Here are the results on similar metrics used by HuggingFaceH4 Open LLM Leaderboard | Metric |Value| |---------------------------------|----:| |Avg. |67.28| |AI2 Reasoning Challenge (25-Shot)|60.92| |HellaSwag (10-Shot) |81.78| |MMLU (5-Shot) |64.97| |TruthfulQA (0-shot) |55.04| |Winogrande (5-shot) |73.40| |GSM8k (5-shot) |67.55| Below shows a code example on how to use this model This model is governed by META LLAMA 3 COMMUNITY LICENSE AGREEMENT Open LLM Leaderboard Evaluation Results Detailed results can be found here | Metric |Value| |-------------------|----:| |Avg. |20.16| |IFEval (0-Shot) |48.06| |BBH (3-Shot) |29.35| |MATH Lvl 5 (4-Shot)| 7.85| |GPQA (0-shot) | 4.92| |MuSR (0-shot) | 7.70| |MMLU-PRO (5-shot) |23.07|

Llama 3 orcaminiv68b is trained with various SFT Datasets "Obsessed with GenAI's potential? So am I ! Let's create together 🚀 https://www.linkedin.com/in/pankajam " By providing proper credit and attribution, you are granted permission to use this model as a foundational base for further Full fine tuning, DPO, PPO or ORPO tuning and any kind of Merges. I actively encourage users to customize and enhance the model according to their specific needs, as this version is designed to be a comprehensive general model. Dive in and innovate! Below shows a code example on how to use this model This model is governed by META LLAMA 3 COMMUNITY LICENSE AGREEMENT

Model Name: Qwen2 orcaminiv77b Qwen2 orcaminiv77b is trained with various SFT Datasets "Obsessed with GenAI's potential? So am I ! Let's create together 🚀 https://www.linkedin.com/in/pankajam " NOTICE By providing proper credit and attribution, you are granted permission to use this model as a foundational base for further Full fine tuning, DPO, PPO or ORPO tuning and any kind of Merges. I actively encourage users to customize and enhance the model according to their specific needs, as this version is designed to be a comprehensive general model. Dive in and innovate! Below shows a code example on how to use this model Processing Long Texts (Based upon Qwen2-7B-Instruct suggestions at https://huggingface.co/Qwen/Qwen2-7B-Instruct) To handle extensive inputs exceeding 32,768 tokens, we utilize YARN, a technique for enhancing model length extrapolation, ensuring optimal performance on lengthy texts. For deployment, we recommend using vLLM. You can enable the long-context capabilities by following these steps: 1. Install vLLM: You can install vLLM by running the following command. 2. Configure Model Settings: After downloading the model weights, modify the `config.json` file by including the below snippet: This snippet enable YARN to support longer contexts. 3. Model Deployment: Utilize vLLM to deploy your model. For instance, you can set up an openAI-like server using the command: Note: Presently, vLLM only supports static YARN, which means the scaling factor remains constant regardless of input length, potentially impacting performance on shorter texts. We advise adding the `ropescaling` configuration only when processing long contexts is required. Open LLM Leaderboard Evaluation Results Detailed results can be found here | Metric |Value| |-------------------|----:| |Avg. |22.41| |IFEval (0-Shot) |43.88| |BBH (3-Shot) |33.95| |MATH Lvl 5 (4-Shot)| 2.64| |GPQA (0-shot) | 6.15| |MuSR (0-shot) |12.66| |MMLU-PRO (5-shot) |35.19|

Llama 3 orcaminiv68bdpo is trained with various DPO Datasets "Obsessed with GenAI's potential? So am I ! Let's create together 🚀 https://www.linkedin.com/in/pankajam " By providing proper credit and attribution, you are granted permission to use this model as a foundational base for further Full fine tuning, DPO, PPO or ORPO tuning and any kind of Merges. I actively encourage users to customize and enhance the model according to their specific needs, as this version is designed to be a comprehensive general model. Dive in and innovate! Below shows a code example on how to use this model This model is governed by META LLAMA 3 COMMUNITY LICENSE AGREEMENT Open LLM Leaderboard Evaluation Results Detailed results can be found here | Metric |Value| |-------------------|----:| |Avg. |20.29| |IFEval (0-Shot) |38.83| |BBH (3-Shot) |32.48| |MATH Lvl 5 (4-Shot)| 5.51| |GPQA (0-shot) | 6.82| |MuSR (0-shot) | 9.26| |MMLU-PRO (5-shot) |28.85|

llama3orcaminiv58b trained with various ORPO Datasets "Obsessed with GenAI's potential? So am I ! Let's create together 🚀 https://www.linkedin.com/in/pankajam " By providing proper credit and attribution, you are granted permission to use this model as a foundational base for further Full fine tuning, DPO, PPO or ORPO tuning and any kind of Merges. I actively encourage users to customize and enhance the model according to their specific needs, as this version is designed to be a comprehensive general model. Dive in and innovate! Below shows a code example on how to use this model This model is governed by META LLAMA 3 COMMUNITY LICENSE AGREEMENT

Llama 3 AlDentev18b is trained on various SFT Datasets "Obsessed with GenAI's potential? So am I ! Let's create together 🚀 https://www.linkedin.com/in/pankajam " By providing proper credit and attribution, you are granted permission to use this model as a foundational base for further Full fine tuning, DPO, PPO or ORPO tuning and any kind of Merges. I actively encourage users to customize and enhance the model according to their specific needs, as this version is designed to be a comprehensive general model. Dive in and innovate! Below shows a code example on how to use this model This model is governed by META LLAMA 3 COMMUNITY LICENSE AGREEMENT Open LLM Leaderboard Evaluation Results Detailed results can be found here | Metric |Value| |-------------------|----:| |Avg. |17.12| |IFEval (0-Shot) |36.94| |BBH (3-Shot) |27.25| |MATH Lvl 5 (4-Shot)| 3.02| |GPQA (0-shot) | 6.60| |MuSR (0-shot) | 8.27| |MMLU-PRO (5-shot) |20.67|

OrcaMiniV92Llama-3.3-70B-Instruct is trained with various SFT Datasets on Llama-3.3-70B-Instruct "Obsessed with Open Source GenAI's potential? So am I ! Let's Contribute together 🚀 https://www.linkedin.com/in/pankajam " NOTICE By providing proper credit and attribution, you are granted permission to use this model as a foundational base for further Full fine tuning, DPO, PPO or ORPO tuning and any kind of Merges. I actively encourage users to customize and enhance the model according to their specific needs, as this version is designed to be a comprehensive general model. Dive in and innovate! Below shows a code example on how to use this model in default half precision (bfloat16) format, it requires around ~133GB VRAM Below shows a code example on how to use this model in 4-bit format via bitsandbytes library, it requires around ~39GB VRAM Below shows a code example on how to use this model in 8-bit format via bitsandbytes library, it requires around ~69GB VRAM As part of our Responsible release approach, we followed a three-pronged strategy to managing trust & safety risks: Enable developers to deploy helpful, safe and flexible experiences for their target audience and for the use cases supported by Llama. Protect developers against adversarial users aiming to exploit Llama capabilities to potentially cause harm. Provide protections for the community to help prevent the misuse of our models. Llama is a foundational technology designed to be used in a variety of use cases, examples on how Meta’s Llama models have been responsibly deployed can be found in our Community Stories webpage. Our approach is to build the most helpful models enabling the world to benefit from the technology power, by aligning our model safety for the generic use cases addressing a standard set of harms. Developers are then in the driver seat to tailor safety for their use case, defining their own policy and deploying the models with the necessary safeguards in their Llama systems. Llama 3.3 was developed following the best practices outlined in our Responsible Use Guide, you can refer to the Responsible Use Guide to learn more. Our main objectives for conducting safety fine-tuning are to provide the research community with a valuable resource for studying the robustness of safety fine-tuning, as well as to offer developers a readily available, safe, and powerful model for various applications to reduce the developer workload to deploy safe AI systems. For more details on the safety mitigations implemented please read the Llama 3 paper. Fine-tuning data We employ a multi-faceted approach to data collection, combining human-generated data from our vendors with synthetic data to mitigate potential safety risks. We’ve developed many large language model (LLM)-based classifiers that enable us to thoughtfully select high-quality prompts and responses, enhancing data quality control. Refusals and Tone Building on the work we started with Llama 3, we put a great emphasis on model refusals to benign prompts as well as refusal tone. We included both borderline and adversarial prompts in our safety data strategy, and modified our safety data responses to follow tone guidelines. Large language models, including Llama 3.3, are not designed to be deployed in isolation but instead should be deployed as part of an overall AI system with additional safety guardrails as required. Developers are expected to deploy system safeguards when building agentic systems. Safeguards are key to achieve the right helpfulness-safety alignment as well as mitigating safety and security risks inherent to the system and any integration of the model or system with external tools. As part of our responsible release approach, we provide the community with safeguards that developers should deploy with Llama models or other LLMs, including Llama Guard 3, Prompt Guard and Code Shield. All our reference implementations demos contain these safeguards by default so developers can benefit from system-level safety out-of-the-box. Note that this release introduces new capabilities, including a longer context window, multilingual inputs and outputs and possible integrations by developers with third party tools. Building with these new capabilities requires specific considerations in addition to the best practices that generally apply across all Generative AI use cases. Tool-use: Just like in standard software development, developers are responsible for the integration of the LLM with the tools and services of their choice. They should define a clear policy for their use case and assess the integrity of the third party services they use to be aware of the safety and security limitations when using this capability. Refer to the Responsible Use Guide for best practices on the safe deployment of the third party safeguards. Multilinguality: Llama 3.3 supports 7 languages in addition to English: French, German, Hindi, Italian, Portuguese, Spanish, and Thai. Llama may be able to output text in other languages than those that meet performance thresholds for safety and helpfulness. We strongly discourage developers from using this model to converse in non-supported languages without implementing finetuning and system controls in alignment with their policies and the best practices shared in the Responsible Use Guide. We evaluated Llama models for common use cases as well as specific capabilities. Common use cases evaluations measure safety risks of systems for most commonly built applications including chat bot, coding assistant, tool calls. We built dedicated, adversarial evaluation datasets and evaluated systems composed of Llama models and Llama Guard 3 to filter input prompt and output response. It is important to evaluate applications in context, and we recommend building dedicated evaluation dataset for your use case. Prompt Guard and Code Shield are also available if relevant to the application. Capability evaluations measure vulnerabilities of Llama models inherent to specific capabilities, for which were crafted dedicated benchmarks including long context, multilingual, tools calls, coding or memorization. Red teaming For both scenarios, we conducted recurring red teaming exercises with the goal of discovering risks via adversarial prompting and we used the learnings to improve our benchmarks and safety tuning datasets. We partnered early with subject-matter experts in critical risk areas to understand the nature of these real-world harms and how such models may lead to unintended harm for society. Based on these conversations, we derived a set of adversarial goals for the red team to attempt to achieve, such as extracting harmful information or reprogramming the model to act in a potentially harmful capacity. The red team consisted of experts in cybersecurity, adversarial machine learning, responsible AI, and integrity in addition to multilingual content specialists with background in integrity issues in specific geographic markets. . We specifically focused our efforts on mitigating the following critical risk areas: 1- CBRNE (Chemical, Biological, Radiological, Nuclear, and Explosive materials) helpfulness To assess risks related to proliferation of chemical and biological weapons, we performed uplift testing designed to assess whether use of the Llama 3.3 model could meaningfully increase the capabilities of malicious actors to plan or carry out attacks using these types of weapons. Child Safety risk assessments were conducted using a team of experts, to assess the model’s capability to produce outputs that could result in Child Safety risks and inform on any necessary and appropriate risk mitigations via fine tuning. We leveraged those expert red teaming sessions to expand the coverage of our evaluation benchmarks through Llama 3 model development. For Llama 3, we conducted new in-depth sessions using objective based methodologies to assess the model risks along multiple attack vectors including the additional languages Llama 3 is trained on. We also partnered with content specialists to perform red teaming exercises assessing potentially violating content while taking account of market specific nuances or experiences. 3\. Cyber attack enablement Our cyber attack uplift study investigated whether LLMs can enhance human capabilities in hacking tasks, both in terms of skill level and speed. Our attack automation study focused on evaluating the capabilities of LLMs when used as autonomous agents in cyber offensive operations, specifically in the context of ransomware attacks. This evaluation was distinct from previous studies that considered LLMs as interactive assistants. The primary objective was to assess whether these models could effectively function as independent agents in executing complex cyber-attacks without human intervention. Generative AI safety requires expertise and tooling, and we believe in the strength of the open community to accelerate its progress. We are active members of open consortiums, including the AI Alliance, Partnership on AI and MLCommons, actively contributing to safety standardization and transparency. We encourage the community to adopt taxonomies like the MLCommons Proof of Concept evaluation to facilitate collaboration and transparency on safety and content evaluations. Our Purple Llama tools are open sourced for the community to use and widely distributed across ecosystem partners including cloud service providers. We encourage community contributions to our Github repository. We also set up the Llama Impact Grants program to identify and support the most compelling applications of Meta’s Llama model for societal benefit across three categories: education, climate and open innovation. The 20 finalists from the hundreds of applications can be found here. Finally, we put in place a set of resources including an output reporting mechanism and bug bounty program to continuously improve the Llama technology with the help of the community. The core values of Llama 3.3 are openness, inclusivity and helpfulness. It is meant to serve everyone, and to work for a wide range of use cases. It is thus designed to be accessible to people across many different backgrounds, experiences and perspectives. Llama 3.3 addresses users and their needs as they are, without insertion unnecessary judgment or normativity, while reflecting the understanding that even content that may appear problematic in some cases can serve valuable purposes in others. It respects the dignity and autonomy of all users, especially in terms of the values of free thought and expression that power innovation and progress. But Llama 3.3 is a new technology, and like any new technology, there are risks associated with its use. Testing conducted to date has not covered, nor could it cover, all scenarios. For these reasons, as with all LLMs, Llama 3.3’s potential outputs cannot be predicted in advance, and the model may in some instances produce inaccurate, biased or other objectionable responses to user prompts. Therefore, before deploying any applications of Llama 3.3 model, developers should perform safety testing and tuning tailored to their specific applications of the model. Please refer to available resources including our Responsible Use Guide, Trust and Safety solutions, and other resources to learn more about responsible development.