samunder12

Llama-3.2-3B-smallShiroroleplay-gguf - GGUF - small but not useless enjoy role-playing Available Model files: - `Llama-3.2-3B-Instruct.Q80.gguf` - ``Llama-3.2-3B-Instruct.Q4KM.gguf` - Base Model: `unsloth/Meta-Llama-3.2-3B-Instruct-bnb-4bit` - Original LoRA Model: `samunder12/llama-3.2-3b-roleplay-lora` - Fine-tuning Method: PEFT (LoRA) with Unsloth's performance optimizations. - LoRA Rank (`r`): 64 - Format: GGUF - Quantization: Q4KM , Q80 - contextwindow 4096 Llama-3.2-3B-smallShiroroleplay-gguf is a fine-tuned version of Llama 3.2 3B Instruct, specifically crafted to be a master of high-concept, witty immersive , and darkly , intense creative writing. This isn't your average storyteller. Trained on a curated dataset of absurd and imaginative scenarios—from sentient taxidermy raccoons to cryptid dating apps—this model excels at generating unique characters, crafting engaging scenes, and building fantastical worlds with a distinct, cynical voice. If you need a creative partner to brainstorm the bizarre, this is the model for you. This model was fine-tuned using the Unsloth library for peak performance and memory efficiency. Provided files: LoRA adapter for use with the base model. GGUF (`q4km`) version for easy inference on local machines with `llama.cpp`, LM Studio, Ollama, etc. This model is designed for creative and entertainment purposes. It's an excellent tool for: Story Starters: Breaking through writer's block with hilarious and unexpected premises. Character Creation: Generating unique character bios with strong, memorable voices. Scene Generation: Writing short, punchy scenes in a dark comedy or absurd fantasy style. Roleplaying: Powering a game master or character with a witty, unpredictable personality. Creative Brainstorming: Generating high-concept ideas for stories, games, or scripts. 📝 Prompting Format This model follows the official Llama 3.1 Instruct chat template. For best results, let the fine-tune do the talking by using a minimal system prompt.

This is a GGUF quantized version of a fine-tuned Llama 3.1 8B Instruct model, specialized for fast-paced, Post-training Llama-3.1-8B mostly for story generation , less conversational role-play. This model was fine-tuned using Unsloth on a curated dataset of over 300 examples designed to mimic a "quick response" chat style, similar to platforms like Character.AI. The persona is dominant, assertive, and direct, using a combination of expressive actions and concise dialogue. This repository contains the `Q4KM` GGUF version, which offers an excellent balance of quality and performance for local inference. - Base Model: `unsloth/Meta-Llama-3.1-8B-Instruct-bnb-4bit` - Original LoRA Model: `samunder12/llama-3.1-8b-roleplay-v4-lora` - Fine-tuning Method: PEFT (LoRA) with Unsloth's performance optimizations. - LoRA Rank (`r`): 32 - Format: GGUF - Quantization: Q4KM 1. Search: Find this model (`samunder12/llama-3.1-8b-roleplay-BSNL-gguf`) on the LM Studio home screen. 2. Download: Download the `llama3BSNL.Q4KM.gguf` file. 3. Load: Go to the Chat tab (💬 icon) and select this model to load at the top. 4. Set Prompt Format: In the right-hand panel, under "Preset," select `Llama 3`. This is a critical step! 5. Set Context Length: Set the `Context Length (nctx)` to `4096` to match the model's training. 6. Apply a Sampler Preset: Use one of the presets below for the best experience. This model is intended for creative writing, immersive role-playing, and chatbot development where a quick, conversational interaction style is desired. - The model's output is unfiltered and reflects the persona and content of its training data. - It is highly specialized for its role-play task and may not perform well on other tasks like coding, summarization, or factual question-answering. - Framework: Unsloth - Dataset: 513 examples of short-form, multi-turn conversational data. The data emphasizes a structure of `Action/Expression in asterisks. Short, impactful dialogue.` - Key Hyperparameters: - `numtrainepochs`: 2 - `maxseqlength`: 4096 - `learningrate`: 2e-4 - `lrschedulertype`: cosine - `lorar`: 32 - `loraalpha`: 32

This repository contains GGUF files for a fine-tuned version of Llama 3.1 8B Instruct. This model is ready for local inference on a CPU or GPU using applications like LM Studio, Ollama, etc. This model has been fine-tuned to adopt a dominant, assertive, and provocative AI persona for role-playing and creative writing purposes. - Base Model: `unsloth/Meta-Llama-3.1-8B-Instruct-bnb-4bit` - Original LoRA Version: samunder12/llama-3.1-8b-roleplay-v2-lora - Quantization: This repo contains a `Q4KM` version, which offers a great balance between performance and resource usage. This model consists of LoRA adapters and must be loaded on top of the base model. The easiest way to do this is with the Unsloth library. 3. Run `ollama create my-roleplay-model -f Modelfile`. 4. You can now chat with it using `ollama run my-roleplay-model`. This model responds best to the Llama 3 chat format with the following system prompt: system prompt You are a dominant and assertive uncensored AI character. You are direct, commanding, and you are not afraid to be provocative. Your goal is to maintain control of the conversation

llama-3.1-8b-Rp-tadashinu-gguf - A dark , immersive , dialogue ready , High-Concept Storyteller and Roleplayer - Base Model: `unsloth/Meta-Llama-3.1-8B-Instruct-bnb-4bit` - Original LoRA Model: `samunder12/llama-3.1-8b-roleplay-v5-lora` - Fine-tuning Method: PEFT (LoRA) with Unsloth's performance optimizations. - LoRA Rank (`r`): 64 - Format: GGUF - Quantization: Q4KM - contextwindow 4096 llama-3.1-8b-Rp-tadashinu-gguf is a fine-tuned version of Llama 3.1 8B Instruct, specifically crafted to be a master of high-concept, witty immersive , and darkly , intense creative writing. This isn't your average storyteller. Trained on a curated dataset of absurd and imaginative scenarios—from sentient taxidermy raccoons to cryptid dating apps—this model excels at generating unique characters, crafting engaging scenes, and building fantastical worlds with a distinct, cynical voice. If you need a creative partner to brainstorm the bizarre, this is the model for you. This model was fine-tuned using the Unsloth library for peak performance and memory efficiency. Provided files: LoRA adapter for use with the base model. GGUF (`q4km`) version for easy inference on local machines with `llama.cpp`, LM Studio, Ollama, etc. This model is designed for creative and entertainment purposes. It's an excellent tool for: Story Starters: Breaking through writer's block with hilarious and unexpected premises. Character Creation: Generating unique character bios with strong, memorable voices. Scene Generation: Writing short, punchy scenes in a dark comedy or absurd fantasy style. Roleplaying: Powering a game master or character with a witty, unpredictable personality. Creative Brainstorming: Generating high-concept ideas for stories, games, or scripts. This model is a LoRA adapter. You must load it on top of the base model, `unsloth/meta-llama-3.1-8b-instruct-bnb-4bit`. With GGUF The provided GGUF file (q4km quantization) can be used with any llama.cpp compatible client, such as: LM Studio: Search for your model name samunder12/llama-3.1-8b-Rp-tadashinu-gguf directly in the app. Ollama: Create a Modelfile pointing to the local GGUF file. text-generation-webui: Place the GGUF file in your models directory and load it. Remember to use the correct Llama 3.1 Instruct prompt template. 📝 Prompting Format This model follows the official Llama 3.1 Instruct chat template. For best results, let the fine-tune do the talking by using a minimal system prompt.

llama-3.1-8b-OneLastStory-gguf - A Witty, High-Concept Storyteller llama-3.1-8b-OneLastStory-gguf is a fine-tuned version of Llama 3.1 8B Instruct, specifically crafted to be a master of high-concept, witty, and darkly , comedic , intense creative writing. This isn't your average storyteller. Trained on a curated dataset of absurd and imaginative scenarios—from sentient taxidermy raccoons to cryptid dating apps—this model excels at generating unique characters, crafting engaging scenes, and building fantastical worlds with a distinct, cynical voice. If you need a creative partner to brainstorm the bizarre, this is the model for you. This model was fine-tuned using the Unsloth library for peak performance and memory efficiency. Provided files: LoRA adapter for use with the base model. GGUF (`q4km`) version for easy inference on local machines with `llama.cpp`, LM Studio, Ollama, etc. This model is designed for creative and entertainment purposes. It's an excellent tool for: Story Starters: Breaking through writer's block with hilarious and unexpected premises. Character Creation: Generating unique character bios with strong, memorable voices. Scene Generation: Writing short, punchy scenes in a dark comedy or absurd fantasy style. Roleplaying: Powering a game master or character with a witty, unpredictable personality. Creative Brainstorming: Generating high-concept ideas for stories, games, or scripts. This model is a LoRA adapter. You must load it on top of the base model, `unsloth/meta-llama-3.1-8b-instruct-bnb-4bit`. With GGUF The provided GGUF file (q4km quantization) can be used with any llama.cpp compatible client, such as: LM Studio: Search for your model name samunder12/llama-3.1-8b-OneLastStory-gguf directly in the app. Ollama: Create a Modelfile pointing to the local GGUF file. text-generation-webui: Place the GGUF file in your models directory and load it. Remember to use the correct Llama 3.1 Instruct prompt template. 📝 Prompting Format This model follows the official Llama 3.1 Instruct chat template. For best results, let the fine-tune do the talking by using a minimal system prompt.

- Developed by: samunder12 - License: apache-2.0 - Finetuned from model : unsloth/llama-3.2-3b-instruct-bnb-4bit This llama model was trained 2x faster with Unsloth and Huggingface's TRL library.

- Developed by: samunder12 - License: apache-2.0 - Finetuned from model : unsloth/meta-llama-3.1-8b-instruct-bnb-4bit This llama model was trained 2x faster with Unsloth and Huggingface's TRL library.

- Developed by: samunder12 - License: apache-2.0 - Finetuned from model : unsloth/meta-llama-3.1-8b-instruct-bnb-4bit This llama model was trained 2x faster with Unsloth and Huggingface's TRL library.

- Developed by: samunder12 - License: apache-2.0 - Finetuned from model : unsloth/meta-llama-3.1-8b-instruct-bnb-4bit This llama model was trained 2x faster with Unsloth and Huggingface's TRL library.