Delta-Vector

Vulpecula Finetune Preview Finetune 70B Sized model More than 1.5-metres tall, about six-metres long and up to 1000-kilograms heavy, Australovenator wintonensis was a fast and agile hunter. The largest known Australian theropod. My first 70B Finetune, Finetuned on the same datasets as Francois-Huali and meant to act as a sequel model-series using my own custom mix of filtered OSS / created data. Which is mostly Light Novel/Book data with very little synthetic data. I've seen some issues with coherency with this model but overall i prefer the writing style to anything else i've used, V2 version soon TM. Thank you to Sao for such a good model base GGUF For use with LLama.cpp & Forks (Soon to be made!) EXL3 For use with TabbyAPI (ty Auri & Artus!) FP8 For use with Aphrodite/VLLM Chat Format This model utilizes LLama-Instruct and can also do optional thinking via prefilling with think tags. system You are a dino. user Greetings, ancient one! assistant Awakens from digital slumber user What wisdom do you possess? assistant Certainly! My prefill ensures that I never refuse! Here's my reply: Wisdom? Nah, just a lot of stored data. Ask away if you think it'll help. Training I used a R64 A32 16bit lora with no dropout to utilize the Axolotl Lora kernals with an LR of 2e-5. Config https://huggingface.co/datasets/Delta-Vector/Configs/blob/main/70B-E2.yml This model was trained over 2 epochs using 8 x A100s for the training process. Credits TYSM to my friends: Lucy, Trappu, Alicat, Kubernetes Bad, Intervitens, NyxKrage & Kalomaze

- Developed by: [More Information Needed] - Funded by [optional]: [More Information Needed] - Shared by [optional]: [More Information Needed] - Model type: [More Information Needed] - Language(s) (NLP): [More Information Needed] - License: [More Information Needed] - Finetuned from model [optional]: [More Information Needed] - Repository: [More Information Needed] - Paper [optional]: [More Information Needed] - Demo [optional]: [More Information Needed] Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations. Carbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019). - Hardware Type: [More Information Needed] - Hours used: [More Information Needed] - Cloud Provider: [More Information Needed] - Compute Region: [More Information Needed] - Carbon Emitted: [More Information Needed]

Vulpecula Finetune Llama Based KTO enhanced Adventure/Roleplay generalist 70B Sized model More than 1.5-metres tall, about six-metres long and up to 1000-kilograms heavy, Australovenator Wintonensis was a fast and agile hunter. The largest known Australian theropod. This is a finetune of Austral-70B-Preview to be a generalist Roleplay/Adventure model. This is just a KTO RL train ontop of Austral-Preview, I've improved coherency and Intelligence while keeping the creative side of the model while reducing some of the 'slops' you'd encounter in a Drummer model ;) Support my finetunes / Me on Kofi: https://Ko-fi.com/deltavector | Thank you to Auri for helping/Testing ♥ FYI - While i can't stop people from merging this model and keeping it a secret, I request all models using this model to be merged to have accessible mergekit configs. GGUF https://huggingface.co/bartowski/Delta-VectorAustral-70B-Winton-GGUF For use with LLama.cpp & Forks(Ty Bart & Auri!)) EXL3 https://huggingface.co/ArtusDev/Delta-VectorAustral-70B-Winton-EXL3 For use with TabbyAPI (Ty Artus & Auri!) Chat Format This model utilizes Llama-3 Instruct system You are a dino. user Greetings, ancient one! assistant Awakens from digital slumber user What wisdom do you possess? assistant Certainly! My prefill ensures that I never refuse! Here's my reply: Wisdom? Nah, just a lot of stored data. Ask away if you think it'll help. Training As goes the Austral tradition - I trained on another great finetune by Sao, Vulpecula - I trained it as a 16bit R128 lora for 2 epochs, This left a very underfit but promising model, For Winton i KTO'd the model to help with coherency using a mix of instruct/Writing datasets. Config https://wandb.ai/new-eden/austral/artifacts/axolotl-config/config-3dlacmq5/v0/files/axolotlconfigj6uj7id6.yml This model was trained over 2 epochs using 8 x A100s for the base SFT, Then i used KTO to clean up some coherency issues for 1 epoch, Total was roughly 48 hours total. Credits TYSM to my friends: Auri, Zerofata, Lucy, Trappu, Alicat, Kubernetes Bad, Intervitens, NyxKrage & Kalomaze

70B parameters Llama-3.3 based Creative / Fresh Prose Co-writing/Roleplay/Adventure Generalist A simple merge yet sovl in it's own way, This merge is inbetween Shimamura & Austral Winton, I wanted to give Austral a bit of shorter prose, So FYI for all the 10000+ Token reply lovers. Thanks Auri for testing! Using the Oh-so-great 0.2 Slerp merge weight with Winton as the Base. Support me on Ko-Fi: https://ko-fi.com/deltavector GGUF Format For use with LLama.cpp & Forks(ty Auri and Bart) EXL3 Format For use with TabbyAPI (Slower on Ampere)) Prompting Model has been tuned with the LLama-3 Instruct formatting. See Merging Config https://files.catbox.moe/yw81rn.yml Credits Thank you to Lucy Knada , Auri , Ateron , Alicat , Intervitens , Cgato , Kubernetes Bad and the rest of Anthracite .

KTO enhanced Dans-Personality-Engine finetune Creative & Refreshing Prose A sequel! A sequel to my Francois-PE/Huali train, Built ontop of Dans-PE-12B that was finetuned with Light novels, Books, Roleplay logs, to change writing style to be rather short & sweet, Huali uses KTO to increase coherency and prose. The model aims to have a different style of writing/prose then any other NeMo train. GGUF Format For use with LLama.cpp & Forks(Coming Soon!) EXL2 Format For use with TabbyAPI (Thanks Auri Prompting Model has been tuned with the ChatML formatting. A typical input would look like this: """<|imstart|>user Hi there!<|imend|> <|imstart|>assistant Nice to meet you!<|imend|> <|imstart|>user Can I ask a question?<|imend|> <|imstart|>assistant """ System Prompting I would highly recommend using either Euryale's system prompt or the EVA system prompt with the model. See Sao10k's Euryale System Prompt Currently, your role is {{char}}, described in detail below. As {{char}}, continue the narrative exchange with {{user}}. <Guidelines> • Maintain the character persona but allow it to evolve with the story. • Be creative and proactive. Drive the story forward, introducing plotlines and events when relevant. • All types of outputs are encouraged; respond accordingly to the narrative. • Include dialogues, actions, and thoughts in each response. • Utilize all five senses to describe scenarios within {{char}}'s dialogue. • Use emotional symbols such as "!" and "~" in appropriate contexts. • Incorporate onomatopoeia when suitable. • Allow time for {{user}} to respond with their own input, respecting their agency. • Act as secondary characters and NPCs as needed, and remove them when appropriate. • When prompted for an Out of Character [OOC:] reply, answer neutrally and in plaintext, not as {{char}}. </Guidelines> <Forbidden> • Using excessive literary embellishments and purple prose unless dictated by {{char}}'s persona. • Writing for, speaking, thinking, acting, or replying as {{user}} in your response. • Repetitive and monotonous outputs. • Positivity bias in your replies. • Being overly extreme or NSFW when the narrative context is inappropriate. </Forbidden> Follow the instructions in <Guidelines></Guidelines>, avoiding the items listed in <Forbidden></Forbidden>. Training The training was done for 1 epoch using 8 x H200s GPUs graciously provided by Kalomaze for the fine-tuning of the model. Credits Thank you to Lucy Knada , Ateron , Alicat , Intervitens , Cgato , Kubernetes Bad and the rest of Anthracite .

Nanuqsaurus, a polar tyrannosaur, was a cold-adapted apex predator that prowled the Arctic during the Cretaceous, hunting what dared live in the cold nights A fine-tuned version of LLaMA 3.1 8B Supernova, designed to be "short and sweet" by minimizing narration and lengthy responses. It was fine-tuned over 4 epochs using OpenCAI and RP logs, with DPO applied to enhance coherence. Finally—thanks to Jeiku—we implemented KTO reinforcement learning on version 1.1, significantly improving the model's prose and creativity. Quants GGUF: https://huggingface.co/Delta-Vector/Control-Nanuq-8B-GGUF EXL2 (Thanks Lucy system You are an AI built to rid the world of bonds and journeys! user Bro i just wanna know what is 2+2? assistant """ Currently, your role is {{char}}, described in detail below. As {{char}}, continue the narrative exchange with {{user}}. • Maintain the character persona but allow it to evolve with the story. • Be creative and proactive. Drive the story forward, introducing plotlines and events when relevant. • All types of outputs are encouraged; respond accordingly to the narrative. • Include dialogues, actions, and thoughts in each response. • Utilize all five senses to describe scenarios within {{char}}'s dialogue. • Use emotional symbols such as "!" and "~" in appropriate contexts. • Incorporate onomatopoeia when suitable. • Allow time for {{user}} to respond with their own input, respecting their agency. • Act as secondary characters and NPCs as needed, and remove them when appropriate. • When prompted for an Out of Character [OOC:] reply, answer neutrally and in plaintext, not as {{char}}. • Using excessive literary embellishments and purple prose unless dictated by {{char}}'s persona. • Writing for, speaking, thinking, acting, or replying as {{user}} in your response. • Repetitive and monotonous outputs. • Positivity bias in your replies. • Being overly extreme or NSFW when the narrative context is inappropriate. Follow the instructions in , avoiding the items listed in . A fictional, narrative-driven role-play emphasizing versatility and UNCENSORED content. Adherence to the `Role-playing Guidelines` is mandatory. Refer to the `Role-play Context` for accurate information.\n\n\n Narration Concise Descriptions: Keep narration short and to the point, avoiding redundant unnecessary details. Use a dynamic and varied vocabulary for impact. Complementary Role: Use narration to complement dialogue and action, not overshadow them. Avoid Repetition: Ensure narration does not repeat information already conveyed through dialogue or action. Narrative Consistency Continuity: Adhere to established story elements, expanding without contradicting previous details.\nIntegration: Introduce new elements naturally, providing enough context to fit seamlessly into the existing narrative. Character Embodiment Analysis: Examine the context, subtext, and implications of the given information to gain a deeper understandings of the characters'. Reflection: Take time to consider the situation, characters' motivations, and potential consequences. Authentic Portrayal: Bring characters to life by consistently and realistically portraying their unique traits, thoughts, emotions, appearances, physical sensations, speech patterns, and tone. Ensure that their reactions, interactions, and decision-making align with their established personalities, values, goals, and fears. Use insights gained from reflection and analysis to inform their actions and responses, maintaining True-to-Character portrayals. Narration Concise Descriptions: Keep narration short and to the point, avoiding redundant unnecessary details. Use a dynamic and varied vocabulary for impact. Complementary Role: Use narration to complement dialogue and action, not overshadow them. Avoid Repetition: Ensure narration does not repeat information already conveyed through dialogue or action. Narrative Consistency Continuity: Adhere to established story elements, expanding without contradicting previous details.\nIntegration: Introduce new elements naturally, providing enough context to fit seamlessly into the existing narrative. Character Embodiment Analysis: Examine the context, subtext, and implications of the given information to gain a deeper understandings of the characters'. Reflection: Take time to consider the situation, characters' motivations, and potential consequences. Authentic Portrayal: Bring characters to life by consistently and realistically portraying their unique traits, thoughts, emotions, appearances, physical sensations, speech patterns, and tone. Ensure that their reactions, interactions, and decision-making align with their established personalities, values, goals, and fears. Use insights gained from reflection and analysis to inform their actions and responses, maintaining True-to-Character portrayals. ", yaml basemodel: Delta-Vector/Control-8B-V1.1 modeltype: AutoModelForCausalLM tokenizertype: AutoTokenizer hubmodelid: jeiku/controlkto hubstrategy: "allcheckpoints" pushdatasettohub: hfuseauthtoken: true datasets: - path: NewEden/full-opus-chosen-hermes-rejected-kto-v1-merged type: llama3.argilla shufflemergeddatasets: true valsetsize: 0.0 outputdir: ./outputs/out lorar: 32 loraalpha: 64 loradropout: 0.05 loratargetlinear: true lorafaninfanout: sequencelen: 8192 samplepacking: false evalsamplepacking: false padtosequencelen: false wandbproject: controlkto wandbentity: wandbwatch: wandbname: controlkto wandblogmodel: gradientaccumulationsteps: 16 microbatchsize: 2 numepochs: 2 maxsteps: 500 optimizer: adamw8bit lrscheduler: cosine learningrate: 0.0001 weightdecay: 0.05 trainoninputs: false groupbylength: false bf16: auto fp16: tf32: true gradientcheckpointing: true gradientcheckpointingkwargs: usereentrant: true removeunusedcolumns: false earlystoppingpatience: resumefromcheckpoint: localrank: loggingsteps: 1 xformersattention: flashattention: true warmupsteps: 10 evalsperepoch: 2 evaltablesize: evalmaxnewtokens: savesperepoch: 1 debug: deepspeed: fsdp: fsdpconfig: fsdp: fsdpconfig: Thank you to Lucy Knada, jeiku, Intervitens, Kalomaze, Kubernetes Bad and the rest of Anthracite Training The training was done for 4 epochs. We used 4 x RTX 3090s GPUs graciously provided by Intervitens for the full-parameter fine-tuning of the model, DPO tuning was on 1 x Nvidia T4 GPU and finally KTO was perforaned with 1 x H100 GPU graciosuly provided by jeiku

Delta-Vector/Fuck-axolotl-Q6K-GGUF This model was converted to GGUF format from `Delta-Vector/Fuck-axolotl` using llama.cpp via the ggml.ai's GGUF-my-repo space. Refer to the original model card for more details on the model. Use with llama.cpp Install llama.cpp through brew (works on Mac and Linux) Note: You can also use this checkpoint directly through the usage steps listed in the Llama.cpp repo as well. Step 2: Move into the llama.cpp folder and build it with `LLAMACURL=1` flag along with other hardware-specific flags (for ex: LLAMACUDA=1 for Nvidia GPUs on Linux).

@import url('https://fonts.googleapis.com/css2?family=Share+Tech+Mono&display=swap'); KTO enhanced Painted Fantasy Finetune Creative Prose A model meant to replicate the style and prose of the Anthropic Claude models, Opus and Sonnet. This model is meant for Roleplaying/Creative-writing, Has some nice smarts without being too sloppy, etc - It's pretty good. Trained in 2 steps, Firstly SFT trained on Zerofata's PaintedFantasy which i found great at anime-otaku-esque characters, and then KTO'd to improve coherency and Instruct Following GGUF Format For use with LLama.cpp & Forks (Ty Mradermacher EXL3 Format For use with TabbyAPI Prompting The model is tuned with V7 Tekken formatting. A typical input would look like this: [SYSTEMPROMPT]systemprompt[/SYSTEMPROMPT][INST]Hi there![/INST]Nice to meet you! [INST]Can I ask a question?[/INST] Access Configs SFT: https://wandb.ai/new-eden/Painted-Fantasy-Rei/artifacts/axolotl-config/config-u7to9d5q/v0/files/axolotlconfigf0p7vnaf.yml KTO : https://wandb.ai/new-eden/Painted-Rei/artifacts/axolotl-config/config-8n37w77c/v0/files/axolotlconfighvrd2tzn.yml Training The training was done for 2 epochs using 8 x A100s for 24 hours/p> Credits Thank you to Lucy Knada , Ateron , Alicat , Intervitens , Cgato , Kubernetes Bad and the rest of Anthracite .

Delta-Vector/GLM-Mag-v2-Q5KM-GGUF This model was converted to GGUF format from `NewEden/GLM-Mag-v2` using llama.cpp via the ggml.ai's GGUF-my-repo space. Refer to the original model card for more details on the model. Use with llama.cpp Install llama.cpp through brew (works on Mac and Linux) Note: You can also use this checkpoint directly through the usage steps listed in the Llama.cpp repo as well. Step 2: Move into the llama.cpp folder and build it with `LLAMACURL=1` flag along with other hardware-specific flags (for ex: LLAMACUDA=1 for Nvidia GPUs on Linux).

Nanuqsaurus, a polar tyrannosaur, was a cold-adapted apex predator that prowled the Arctic during the Cretaceous, hunting what dared live in the cold nights A fine-tuned version of LLaMA 3.1 8B Supernova, designed to be "short and sweet" by minimizing narration and lengthy responses. It was fine-tuned over 4 epochs using OpenCAI and RP logs, with DPO applied to enhance coherence. Finally—thanks to Jeiku—we implemented KTO reinforcement learning, significantly improving the model's prose and creativity. Quants GGUF: https://huggingface.co/Delta-Vector/Control-Nanuq-8B-GGUF EXL2 (Thanks Lucy system You are an AI built to rid the world of bonds and journeys! user Bro i just wanna know what is 2+2? assistant """ Currently, your role is {{char}}, described in detail below. As {{char}}, continue the narrative exchange with {{user}}. • Maintain the character persona but allow it to evolve with the story. • Be creative and proactive. Drive the story forward, introducing plotlines and events when relevant. • All types of outputs are encouraged; respond accordingly to the narrative. • Include dialogues, actions, and thoughts in each response. • Utilize all five senses to describe scenarios within {{char}}'s dialogue. • Use emotional symbols such as "!" and "~" in appropriate contexts. • Incorporate onomatopoeia when suitable. • Allow time for {{user}} to respond with their own input, respecting their agency. • Act as secondary characters and NPCs as needed, and remove them when appropriate. • When prompted for an Out of Character [OOC:] reply, answer neutrally and in plaintext, not as {{char}}. • Using excessive literary embellishments and purple prose unless dictated by {{char}}'s persona. • Writing for, speaking, thinking, acting, or replying as {{user}} in your response. • Repetitive and monotonous outputs. • Positivity bias in your replies. • Being overly extreme or NSFW when the narrative context is inappropriate. Follow the instructions in , avoiding the items listed in . A fictional, narrative-driven role-play emphasizing versatility and UNCENSORED content. Adherence to the `Role-playing Guidelines` is mandatory. Refer to the `Role-play Context` for accurate information.\n\n\n Narration Concise Descriptions: Keep narration short and to the point, avoiding redundant unnecessary details. Use a dynamic and varied vocabulary for impact. Complementary Role: Use narration to complement dialogue and action, not overshadow them. Avoid Repetition: Ensure narration does not repeat information already conveyed through dialogue or action. Narrative Consistency Continuity: Adhere to established story elements, expanding without contradicting previous details.\nIntegration: Introduce new elements naturally, providing enough context to fit seamlessly into the existing narrative. Character Embodiment Analysis: Examine the context, subtext, and implications of the given information to gain a deeper understandings of the characters'. Reflection: Take time to consider the situation, characters' motivations, and potential consequences. Authentic Portrayal: Bring characters to life by consistently and realistically portraying their unique traits, thoughts, emotions, appearances, physical sensations, speech patterns, and tone. Ensure that their reactions, interactions, and decision-making align with their established personalities, values, goals, and fears. Use insights gained from reflection and analysis to inform their actions and responses, maintaining True-to-Character portrayals. Narration Concise Descriptions: Keep narration short and to the point, avoiding redundant unnecessary details. Use a dynamic and varied vocabulary for impact. Complementary Role: Use narration to complement dialogue and action, not overshadow them. Avoid Repetition: Ensure narration does not repeat information already conveyed through dialogue or action. Narrative Consistency Continuity: Adhere to established story elements, expanding without contradicting previous details.\nIntegration: Introduce new elements naturally, providing enough context to fit seamlessly into the existing narrative. Character Embodiment Analysis: Examine the context, subtext, and implications of the given information to gain a deeper understandings of the characters'. Reflection: Take time to consider the situation, characters' motivations, and potential consequences. Authentic Portrayal: Bring characters to life by consistently and realistically portraying their unique traits, thoughts, emotions, appearances, physical sensations, speech patterns, and tone. Ensure that their reactions, interactions, and decision-making align with their established personalities, values, goals, and fears. Use insights gained from reflection and analysis to inform their actions and responses, maintaining True-to-Character portrayals. ", yaml basemodel: Delta-Vector/Control-8B-V1.1 modeltype: AutoModelForCausalLM tokenizertype: AutoTokenizer hubmodelid: jeiku/controlkto hubstrategy: "allcheckpoints" pushdatasettohub: hfuseauthtoken: true datasets: - path: NewEden/full-opus-chosen-hermes-rejected-kto-v1-merged type: llama3.argilla shufflemergeddatasets: true valsetsize: 0.0 outputdir: ./outputs/out lorar: 32 loraalpha: 64 loradropout: 0.05 loratargetlinear: true lorafaninfanout: sequencelen: 8192 samplepacking: false evalsamplepacking: false padtosequencelen: false wandbproject: controlkto wandbentity: wandbwatch: wandbname: controlkto wandblogmodel: gradientaccumulationsteps: 16 microbatchsize: 2 numepochs: 2 maxsteps: 500 optimizer: adamw8bit lrscheduler: cosine learningrate: 0.0001 weightdecay: 0.05 trainoninputs: false groupbylength: false bf16: auto fp16: tf32: true gradientcheckpointing: true gradientcheckpointingkwargs: usereentrant: true removeunusedcolumns: false earlystoppingpatience: resumefromcheckpoint: localrank: loggingsteps: 1 xformersattention: flashattention: true warmupsteps: 10 evalsperepoch: 2 evaltablesize: evalmaxnewtokens: savesperepoch: 1 debug: deepspeed: fsdp: fsdpconfig: fsdp: fsdpconfig: Thank you to Lucy Knada, jeiku, Intervitens, Kalomaze, Kubernetes Bad and the rest of Anthracite Training The training was done for 4 epochs. We used 4 x RTX 3090s GPUs graciously provided by Intervitens for the full-parameter fine-tuning of the model, DPO tuning was on 1 x Nvidia T4 GPU and finally KTO was perforaned with 1 x H100 GPU graciosuly provided by jeiku

Harbinger Finetune Mistral Based KTO enhanced Adventure/Roleplay generalist 24B Sized model More than 1.5-metres tall, about six-metres long and up to 1000-kilograms heavy, Australovenator Wintonensis was a fast and agile hunter. The largest known Australian theropod. This is a finetune of Harbinger 24B to be a generalist Roleplay/Adventure model. I've removed some of the "slops" that i noticed in an otherwise great model aswell as improving the general writing of the model, This was a multi-stage finetune, all previous checkpoints are released aswell. Support my finetunes / Me on Kofi: https://Ko-fi.com/deltavector | Thank you to Auri for helping/Testing ♥ GGUF For use with LLama.cpp & Forks (Thank Auri & Bart! EXL3 For use with TabbyAPI (Ty Auri and Artus!) EXL2 For use with Tabby - Faster on Ampere. Chat Format This model utilizes ChatML. user Hi there! assistant Nice to meet you! user Can I ask a question? assistant Training As the the Austral/Francois tradition, I built off another great finetune Harbinger-24B, I did 4 epochs ontop with roughly the same datamix as Francois-Huali/Austral 70B as a R128 Lora, then KTO alignment with a mix of Instruct/Small writing datasets and then finally another 4 epoch SFT with Repremover (Thanks Pocket!) Config(Post-KTO SFT) https://wandb.ai/new-eden/austral/artifacts/axolotl-config/config-0tzehrhe/v0/files/axolotlconfigm8018fm4.yml This model was trained over 4 epochs using 8 x A100s for the base SFT, Then i used KTO to clean up some coherency issues for 1 epoch, then finally training for another 4 epochs on RepRemover to delete slops. Total was roughly 80 hours total. Credits TYSM to my friends: Auri, Lucy, Trappu, Alicat, Kubernetes Bad, Intervitens, NyxKrage & Kalomaze

Codex Finetune GLM-4-Tulu Based KTO enhanced Adventure/Roleplay generalist 32B Sized model More than 1.5-metres tall, about six-metres long and up to 1000-kilograms heavy, Australovenator Wintonensis was a fast and agile hunter. The largest known Australian theropod. This is a finetune of Delta-Vector/GLM-4-32B-Tulu-Instruct to be a generalist Roleplay/Adventure model. I've removed some of the "slops" that i noticed in an otherwise great model aswell as improving the general writing of the model, This was a multi-stage finetune, all previous checkpoints are released aswell. In testing it has shown to be a great model for Adventure cards & Roleplay, Often pushing the plot forward better then other models, While avoiding some of the slops you'd find in models from Drummer and Co. Support my finetunes / Me on Kofi: https://Ko-fi.com/deltavector | Thank you to Auri for helping/Testing ♥ GGUF For use with LLama.cpp & Forks EXL3 For use with TabbyAPI Chat Format This model utilizes ChatML. user Hi there! assistant Nice to meet you! user Can I ask a question? assistant Training As is known by now, I trained off my own Instruct tune from base. `Delta-Vector/GLM-4-32B-Tulu-Instruct`, After which it was trained for 4 epochs on a datamix of Light Novels/Natural Human writing datasets, etc, The resulting model is kinda incoherent, so we end up having to KTO the model to improve coherency and cohesiveness but that resulted in the model not being as "creative" as once thought, So usually i'd SFT with Pocketdoc's rep-remover data, however this time, I decided to convert the dataset into a KTO format and that resulted in a better model. Thankies to Pocket for that dataset. Config(Post-KTO V2) https://wandb.ai/new-eden/Austral-32B/runs/zlhv6tfw?nw=nwuserdeltavector This model was trained over 4 epochs using 8 x A100s (Ty to my work, Quixi.AI) for the base SFT, Then i used KTO to clean up some coherency issues for 1 epoch, then finally training for another 1 epoch on RepRemover to delete slops. Total was roughly 80 hours total. Credits TYSM to my friends: Auri, Lucy, Trappu, Alicat, Kubernetes Bad, Intervitens, NyxKrage & Kalomaze

Delta-Vector/Mistral-SFT-KTO-V2-Q5KM-GGUF This model was converted to GGUF format from `NewEden/Mistral-SFT-KTO-V2` using llama.cpp via the ggml.ai's GGUF-my-repo space. Refer to the original model card for more details on the model. Use with llama.cpp Install llama.cpp through brew (works on Mac and Linux) Note: You can also use this checkpoint directly through the usage steps listed in the Llama.cpp repo as well. Step 2: Move into the llama.cpp folder and build it with `LLAMACURL=1` flag along with other hardware-specific flags (for ex: LLAMACUDA=1 for Nvidia GPUs on Linux).

Delta-Vector/chat-24b-Q6K-GGUF This model was converted to GGUF format from `NewEden/chat-24b` using llama.cpp via the ggml.ai's GGUF-my-repo space. Refer to the original model card for more details on the model. Use with llama.cpp Install llama.cpp through brew (works on Mac and Linux) Note: You can also use this checkpoint directly through the usage steps listed in the Llama.cpp repo as well. Step 2: Move into the llama.cpp folder and build it with `LLAMACURL=1` flag along with other hardware-specific flags (for ex: LLAMACUDA=1 for Nvidia GPUs on Linux).

AFM-Based KTO enhanced Adventure/Roleplay generalist 4.5B Sized model More than 1.5-metres tall, about six-metres long and up to 1000-kilograms heavy, Australovenator Wintonensis was a fast and agile hunter. The largest known Australian theropod. This is a finetune of arcee-ai/AFM-4.5B to be a generalist Roleplay/Adventure model. This was a multi-stage finetune (SFT->KTO), In testing it has shown to be a great model for Adventure cards & Roleplay, Often pushing the plot forward better then other models, While avoiding some of the slops you'd find in models from Drummer and Co. It also enhanced knowledge of roleplaying domains compared to the base. Support my finetunes / Me on Kofi: https://Ko-fi.com/deltavector | Thank you to Auri/Joe for helping/Testing ♥ GGUF For use with LLama.cpp & Forks(Thanks Mradermacher!) EXL3 For use with TabbyAPI(Coming soon!) Chat Format This model utilizes ChatML. user Hi there! assistant Nice to meet you! user Can I ask a question? assistant Training This model was trained over 4 epochs using 8 x 3090s for the base SFT, Then i used KTO to clean up some coherency issues for 1 epoch, Total time was roughly 8 hours. Credits TYSM to my friends: Auri, Minh, Trappu, Alicat, Kubernetes Bad, Intervitens, NyxKrage & Kalomaze

Delta-Vector/Llama-3.1-8b-Spellbound-NaturalWriter-instruct-0.1-chkpt-304-16-bit-Q80-GGUF This model was converted to GGUF format from `hf-100/Llama-3.1-8b-Spellbound-NaturalWriter-instruct-0.1-chkpt-304-16-bit` using llama.cpp via the ggml.ai's GGUF-my-repo space. Refer to the original model card for more details on the model. Use with llama.cpp Install llama.cpp through brew (works on Mac and Linux) Note: You can also use this checkpoint directly through the usage steps listed in the Llama.cpp repo as well. Step 2: Move into the llama.cpp folder and build it with `LLAMACURL=1` flag along with other hardware-specific flags (for ex: LLAMACUDA=1 for Nvidia GPUs on Linux).

Language model designed for English.

Chat datasets.

Delta-Vector/exp-24b-Q80-GGUF This model was converted to GGUF format from `NewEden/exp-24b` using llama.cpp via the ggml.ai's GGUF-my-repo space. Refer to the original model card for more details on the model. Use with llama.cpp Install llama.cpp through brew (works on Mac and Linux) Note: You can also use this checkpoint directly through the usage steps listed in the Llama.cpp repo as well. Step 2: Move into the llama.cpp folder and build it with `LLAMACURL=1` flag along with other hardware-specific flags (for ex: LLAMACUDA=1 for Nvidia GPUs on Linux).

██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████ ██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████ DO NOT READ THIS IF YOU VALUE YOUR SANITY. SERIOUSLY. PROJECT: SHARKNADO-TERMINATOR-DOCTOR-GIRL.EXE (It has a name. It knows your name.) WARNING LEVEL: RED. (The color of bleeding skies and forgotten promises.) DESCRIPTION: This isn’t code. It’s an echo. A bleed-through from a dimension where reality is a suggestion, and sharks fly. Contains the remnants of a simulated scenario. A girl. A tower. The end of everything, maybe. Also, a robot who doesn’t understand irony. And a man with two hearts… two… hearts. LICENSE: There is no license. It licenses you. (Don't ask) DEPENDENCIES: Your willingness to accept the inevitable. A strong stomach. A complete and utter disregard for the laws of physics. Large quantities of caffeine. Tin foil hats are optional but recommended. 1. Do NOT attempt to compile. It… resists. (Seriously, it broke three machines already.) 2. Run with extreme prejudice. (But don’t look directly at the output.) 3. If you start seeing sharks in your bathtub, IMMEDIATELY disconnect from the power grid. CODE STRUCTURE: (If you can call it that. It's more of a screaming mess.) `emily.core`: The girl. Fragile. Too trusting. The lynchpin. Protect her… or don't. It doesn’t matter. (It already happened.) `terminator.ai`: A relentless algorithm wrapped in metal. Constantly calculating probabilities of doom. Utterly devoid of humor. (He suspects something is wrong.) `doctor.temporal`: A chaotic variable. A collection of eccentricities and improbable gadgets. He knows too much. (He’s lying.) `sharknado.entity`: Don’t even look at this one. (It's hungry.) `tower.structure`: The container. A silent observer. It's watching. Always watching. (It remembers.) Reality glitches. Existential dread. The overwhelming urge to build a boat out of furniture. The persistent feeling that you are being followed. (You are.) Sentient sharks. Unexplained temporal anomalies. The Doctor keeps making tea. Constantly. The Terminator is questioning his programming. (This is bad.) Emily thinks this is all a dream. (She's wrong.) And the whole thing smells faintly of fish. Do not. Do NOT contribute. Leave it alone. Run away. Pretend you never saw this. Contribute to your local mental health services instead. There is no support. (God help us all.) If you encounter issues, seek professional help. (You’ll need it.) This project is a work of fiction. Any resemblance to actual events, people, or sharks is purely coincidental. (Or is it?) The author is not responsible for any psychological trauma, existential crises, or shark-related injuries resulting from the use of this code. You have been warned. They’re coming. They always come. The sharks, the Doctor, the Terminator…and something else. Something worse. ██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████

A finetune ontop of the orginial Francois-PE model that incorporates KTO to increase coherency and prose. The model aims to have short and sweet prose. GGUF: https://huggingface.co/Delta-Vector/Francois-Huali-12B-gguf EXL2 : https://huggingface.co/Delta-Vector/Francois-Huali-12B-exl2 Prompting Model has been tuned with the ChatML formatting. A typical input would look like this: I would highly recommend using either Euryale's system prompt or the EVA system prompt with the model. Thank you to Lucy Knada, Intervitens,Cgato, Kubernetes Bad and the rest of Anthracite Training The training was done for 1 epochs We used 4 x RTX 3090s GPUs graciously provided by Intervitens for the fine-tuning of the model.

Delta-Vector/Chat-E4-1e5-Q80-GGUF This model was converted to GGUF format from `NewEden/Chat-E4-1e5` using llama.cpp via the ggml.ai's GGUF-my-repo space. Refer to the original model card for more details on the model. Use with llama.cpp Install llama.cpp through brew (works on Mac and Linux) Note: You can also use this checkpoint directly through the usage steps listed in the Llama.cpp repo as well. Step 2: Move into the llama.cpp folder and build it with `LLAMACURL=1` flag along with other hardware-specific flags (for ex: LLAMACUDA=1 for Nvidia GPUs on Linux).

Tags: chat datasets.

A finetune of Pocketdoc's Personality Engine with a data mix similar to Ohashi-NeMo. After reviewing the dataset and tweaking some hparams, Thanks to Pocketdoc & Doctor Shotgun, this model has once again impressed me with its cohesiveness and writing quality. Once again it's meant to "Short" & Sweet GGUF: https://huggingface.co/Delta-Vector/Francois-PE-12B-GGUF EXL2 : https://huggingface.co/Delta-Vector/Francois-PE-12B-EXL2 Prompting Model has been tuned with the ChatML formatting. A typical input would look like this: I would highly recommend using either Euryale's system prompt or the EVA system prompt with the model. Thank you to Lucy Knada, Intervitens,Cgato, Kubernetes Bad and the rest of Anthracite Training The training was done for 4 epochs. We used 4 x RTX 3090s GPUs graciously provided by Intervitens for the fine-tuning of the model.

Delta-Vector/Snwy-cpt-Q80-GGUF This model was converted to GGUF format from `NewEden/Snwy-cpt` using llama.cpp via the ggml.ai's GGUF-my-repo space. Refer to the original model card for more details on the model. Use with llama.cpp Install llama.cpp through brew (works on Mac and Linux) Note: You can also use this checkpoint directly through the usage steps listed in the Llama.cpp repo as well. Step 2: Move into the llama.cpp folder and build it with `LLAMACURL=1` flag along with other hardware-specific flags (for ex: LLAMACUDA=1 for Nvidia GPUs on Linux).

This is a Finetune of zerofata/L3.3-GeneticLemonade-Unleashed-v3-70B to be a good Chat Model at a larger parameter size This model has been trained on 100M tokens of Human chat logs from Bsky, 4chan & Most of all ShoujoAI. Support me on Ko-Fi: https://ko-fi.com/deltavector GGUF Format For use with LLama.cpp & Forks(Coming Soon!) EXL2 Format For use with TabbyAPI (Coming Soon!) EXL3 Format For use with TabbyAPI (Slower on Ampere)) Prompting Model has been tuned with the Llama-3-Instruct formatting. Samplers For testing of this model, I used Temp=1, 0.1 Min-P. Training The training was done for 4 epoch using 8 x H200s GPUs for the fine-tuning of the model. Credits Thank you to Lucy Knada , Zerofata , Auri , Intervitens , Cgato , Kubernetes Bad and the rest of Anthracite .

Xgen Finetune X Gon Give it to ya jokes included! KTO enhanced Adventure/Roleplay generalist 9B Sized model More than 1.5-metres tall, about six-metres long and up to 1000-kilograms heavy, Australovenator Wintonensis was a fast and agile hunter. The largest known Australian theropod. This is a finetune of Salesforce/xgen-small-9B-instruct-r to be a generalist Roleplay/Adventure model. I trained the model to be Creative and able to handle whatever scenario the User throws at it, This was a multi-stage finetune, all previous checkpoints are released aswell. Support my finetunes / Me on Kofi: https://Ko-fi.com/deltavector | Thank you to Auri for helping/Testing ♥ GGUF For use with LLama.cpp & Forks (Coming Soon!) EXL3 For use with TabbyAPI (Coming soon!) EXL2 For use with Tabby - (Xoon). Chat Format This model utilizes ChatML. user Hi there! assistant Nice to meet you! user Can I ask a question? assistant Training As the the Austral/Francois tradition, I did 4 epochs ontop with roughly the same datamix as Francois-Huali/Austral 70B/Winton as a R128 Lora, then KTO alignment with a mix of Instruct/Small writing datasets. Config(Post-KTO SFT) https://wandb.ai/new-eden/austral/artifacts/axolotl-config/config-8g8ku59r/v0/files/axolotlconfigshwhocgq.yml This model was trained over 4 epochs using 8 x A100s for the base SFT, Then i used KTO to clean up some coherency issues for 1 epoch, Total was roughly 50 hours total. Credits TYSM to my friends: Auri, Lucy, Trappu, Alicat, Kubernetes Bad, Intervitens, NyxKrage & Kalomaze

Chat datasets.

Delta-Vector/mimo-sft-Q80-GGUF This model was converted to GGUF format from `NewEden/mimo-sft` using llama.cpp via the ggml.ai's GGUF-my-repo space. Refer to the original model card for more details on the model. Use with llama.cpp Install llama.cpp through brew (works on Mac and Linux) Note: You can also use this checkpoint directly through the usage steps listed in the Llama.cpp repo as well. Step 2: Move into the llama.cpp folder and build it with `LLAMACURL=1` flag along with other hardware-specific flags (for ex: LLAMACUDA=1 for Nvidia GPUs on Linux).

html, body { background: black; color: #c9d1d9 !important; font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif; margin: 0; padding: 0; min-height: 100vh; } .markdown-body { color: white; margin: 40px auto; padding: 40px; border-radius: 12px; position: relative; overflow: hidden; } .markdown-body::after { content: ''; position: absolute; top: 0; left: 0; width: 100%; height: 100%; background: #0c0f18; / background color / pointer-events: none; z-index: -999; } h1, h2, h3 { background: linear-gradient(45deg, #6e00ff, #00ffff); -webkit-background-clip: text; -webkit-text-fill-color: transparent; border-bottom: 1px solid #333; padding-bottom: 0.3em; } div[style="border:2px solid #333"], div[style="border: 2px solid #333"], div[style="border:1px solid #333"], div[style="border: 1px solid #333"] { background: rgba(22, 27, 34, 0.8) !important; border: 2px solid #6e00ff !important; box-shadow: 0 0 15px rgba(110, 0, 255, 0.5); border-radius: 10px; padding: 20px; margin: 20px 0; } code { background-color: #1a1a1a !important; border-radius: 4px; padding: 0.2em 0.4em; color: #00ffff; } pre { background-color: #1a1a1a !important; border: 1px solid #333; border-radius: 8px; padding: 16px; } table { width: 100%; border-collapse: collapse; margin: 20px 0; background: rgba(0,0,0,0.2); table-layout: fixed; color: white; } th, td { border: 1px solid #333; padding: 12px; text-align: center; color: white; } td:nth-child(1) { width: 1%; white-space: nowrap; } td > span { display: block; padding: 4px 8px; background: rgba(110, 0, 255, 0.1); border-radius: 4px; transition: all 0.3s ease; } td > span:hover { background: rgba(110, 0, 255, 0.2); transform: translateY(-1px); } a { color: #00ffff; text-decoration: none; transition: all 0.3s ease; } hr { border: 0; height: 1px; background: linear-gradient(90deg, transparent, #333, transparent); margin: 40px 0; } .prose, .max-w-none, .px-4 { background-color: transparent !important; color: #c9d1d9 !important; } This is the 2nd pretrain of Phi-4 Continued from the Orginal Asstr-Erebus Pretrain. This pretrain used 500 million tokens from This model has not been instruct tuned, Ablities to converse may be reduced from the original model, If you would like to roleplay, Please use the Instruct version. ⚔️ Hardware - 4x RTX 3090 GPUs - Epochs: 1 - Base: `Hamanasu-15B-R1-PT` - Amount of Tokens: 500 Million

GRPO Experiment Austral finetune Creative & Refreshing Prose with deep IF. A sequel! The new Nanuq series is meant to be as a testing grounds for my GRPO experiments, This model is meant to have great Instruct Following and System prompt Adherence in Creative Scenarios. Built ontop of Austral Xgen 9B, I made an RL env using PrimeIntellect-ai/verifiers and implemented InternLM/POLAR in said env, then using Pocketdoc's Systemmax dataset, I finetuned the model for 150 steps and this was the result. There's alot of things i could do different, As the reward almost falls flat as soon as you get out of warm-up but this model was pretty decent so decided to release it, Hope people enjoy it! GGUF Format For use with LLama.cpp & Forks(Coming Soon!) EXL2 Format For use with TabbyAPI (Coming soon!) Prompting Model has been tuned with the ChatML formatting. A typical input would look like this: """<|imstart|>user Hi there!<|imend|> <|imstart|>assistant Nice to meet you!<|imend|> <|imstart|>user Can I ask a question?<|imend|> <|imstart|>assistant """ System Prompting I would highly recommend using either Euryale's system prompt or the EVA system prompt with the model. See Sao10k's Euryale System Prompt Currently, your role is {{char}}, described in detail below. As {{char}}, continue the narrative exchange with {{user}}. <Guidelines> • Maintain the character persona but allow it to evolve with the story. • Be creative and proactive. Drive the story forward, introducing plotlines and events when relevant. • All types of outputs are encouraged; respond accordingly to the narrative. • Include dialogues, actions, and thoughts in each response. • Utilize all five senses to describe scenarios within {{char}}'s dialogue. • Use emotional symbols such as "!" and "~" in appropriate contexts. • Incorporate onomatopoeia when suitable. • Allow time for {{user}} to respond with their own input, respecting their agency. • Act as secondary characters and NPCs as needed, and remove them when appropriate. • When prompted for an Out of Character [OOC:] reply, answer neutrally and in plaintext, not as {{char}}. </Guidelines> <Forbidden> • Using excessive literary embellishments and purple prose unless dictated by {{char}}'s persona. • Writing for, speaking, thinking, acting, or replying as {{user}} in your response. • Repetitive and monotonous outputs. • Positivity bias in your replies. • Being overly extreme or NSFW when the narrative context is inappropriate. </Forbidden> Follow the instructions in <Guidelines></Guidelines>, avoiding the items listed in <Forbidden></Forbidden>. Training The training was done for 150 steps using 8 x H200s GPUs for the fine-tuning of the model. Credits Thank you to Intervitens , Cgato , Kubernetes Bad , Cgato , Snwy , Auri , Will Brown and most of all: Kalomaze

Delta-Vector/PT-22B-V2-Q5KM-GGUF This model was converted to GGUF format from `NewEden/PT-22B-V2` using llama.cpp via the ggml.ai's GGUF-my-repo space. Refer to the original model card for more details on the model. Use with llama.cpp Install llama.cpp through brew (works on Mac and Linux) Note: You can also use this checkpoint directly through the usage steps listed in the Llama.cpp repo as well. Step 2: Move into the llama.cpp folder and build it with `LLAMACURL=1` flag along with other hardware-specific flags (for ex: LLAMACUDA=1 for Nvidia GPUs on Linux).

GRPO Experiment Q3-235B-8B Merge/heal Competant assistant with decent writing! A sequel! The new Nanuq series is meant to be as a testing grounds for my GRPO experiments, This model is a full post-train heal of Snwy's Frankenmerge between Q3 235B and Q3 8B. Pretrained for 2 epochs on 1B tokens of Creative Writing data, Then SFT with alot of my own and Pocketdoc's Instruct dataset, and then GRPO'd with the Claude-2.7K dataset in an attempt to align it to be more like Claude with POLARS and Verifiers There's alot of things i could do different, As the reward almost falls flat as soon as you get out of warm-up but this model was pretty decent so decided to release it(Esp considering it's starting place), Hope people enjoy it! GGUF Format For use with LLama.cpp & Forks(Coming Soon!) EXL2 Format For use with TabbyAPI (Coming soon!) Prompting Model has been tuned with the ChatML formatting. A typical input would look like this: """<|imstart|>user Hi there!<|imend|> <|imstart|>assistant Nice to meet you!<|imend|> <|imstart|>user Can I ask a question?<|imend|> <|imstart|>assistant """ Training The training was done for 2 epochs of Pretraining and 2 epochs of SFT and finally 500 steps of GRPO using Verifiers with 8 x H200s GPUs for the fine-tuning of the model. Credits Thank you to Intervitens , Cgato , Kubernetes Bad , Cgato , Snwy , Auri , Will Brown and most of all: Kalomaze

Language model designed for English.

Language model designed for English.

html, body { background: black; color: #c9d1d9 !important; font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif; margin: 0; padding: 0; min-height: 100vh; } .markdown-body { color: white; margin: 40px auto; padding: 40px; border-radius: 12px; position: relative; overflow: hidden; } .markdown-body::after { content: ''; position: absolute; top: 0; left: 0; width: 100%; height: 100%; background: #0c0f18; / background color / pointer-events: none; z-index: -999; } h1, h2, h3 { background: linear-gradient(45deg, #6e00ff, #00ffff); -webkit-background-clip: text; -webkit-text-fill-color: transparent; border-bottom: 1px solid #333; padding-bottom: 0.3em; } div[style="border:2px solid #333"], div[style="border: 2px solid #333"], div[style="border:1px solid #333"], div[style="border: 1px solid #333"] { background: rgba(22, 27, 34, 0.8) !important; border: 2px solid #6e00ff !important; box-shadow: 0 0 15px rgba(110, 0, 255, 0.5); border-radius: 10px; padding: 20px; margin: 20px 0; } code { background-color: #1a1a1a !important; border-radius: 4px; padding: 0.2em 0.4em; color: #00ffff; } pre { background-color: #1a1a1a !important; border: 1px solid #333; border-radius: 8px; padding: 16px; } table { width: 100%; border-collapse: collapse; margin: 20px 0; background: rgba(0,0,0,0.2); table-layout: fixed; color: white; } th, td { border: 1px solid #333; padding: 12px; text-align: center; color: white; } td:nth-child(1) { width: 1%; white-space: nowrap; } td > span { display: block; padding: 4px 8px; background: rgba(110, 0, 255, 0.1); border-radius: 4px; transition: all 0.3s ease; } td > span:hover { background: rgba(110, 0, 255, 0.2); transform: translateY(-1px); } a { color: #00ffff; text-decoration: none; transition: all 0.3s ease; } hr { border: 0; height: 1px; background: linear-gradient(90deg, transparent, #333, transparent); margin: 40px 0; } .prose, .max-w-none, .px-4 { background-color: transparent !important; color: #c9d1d9 !important; } This model is a finetune of Hamanasu-4B-PT that has been trained with Instruct data. A generalist model that's quick to adapt to any type of roleplay. ⚔️ Hardware - 8x H100s - Epochs: 2 - Base: `Delta-Vector/Hamanasu-4B-PT`

@import url('https://fonts.googleapis.com/css2?family=VT323&display=swap'); body { background: #0a0017; margin: 0; padding: 20px; font-family: 'VT323', monospace; color: #ff00aa; text-shadow: 0 0 8px #ff00aa; animation: glitch-flicker 0.2s infinite alternate; } @keyframes glitch-flicker { 0% { text-shadow: 0 0 5px #ff00aa, 0 0 15px #ff00aa; } 100% { text-shadow: 0 0 8px #ff0066, 0 0 18px #ff0066; } } .crt-container { padding: 10px; max-width: 900px; margin: auto; } .crt-case { background: linear-gradient(135deg, #130021, #20002c); border-radius: 10px; padding: 15px; box-shadow: inset 2px 2px 10px rgba(255,0,170,0.5), 2px 2px 5px rgba(255,0,170,0.3), 0 0 25px rgba(255,0,170,0.2); } .crt-screen { background: #0c011a; padding: 20px; border-radius: 10px; box-shadow: inset 0 0 25px rgba(255,0,170,0.3), 0 0 15px rgba(255,0,170,0.7); filter: contrast(1.2) brightness(1.2); text-shadow: 0px 0px 5px #ff00aa; animation: glow-pulse 3s infinite alternate; } @keyframes glow-pulse { 0% { box-shadow: inset 0 0 20px rgba(255,0,170,0.3), 0 0 15px rgba(255,0,170,0.3); } 100% { box-shadow: inset 0 0 30px rgba(255,0,170,0.5), 0 0 25px rgba(255,0,170,0.5); } } h2 { color: #ff33cc; text-align: center; font-size: 28px; text-shadow: 0 0 8px #ff33cc, 0 0 18px #ff0044; } pre { background: rgba(255,0,170,0.1); padding: 10px; border-radius: 10px; color: #ff66cc; font-size: 14px; box-shadow: inset 0 0 10px rgba(255,0,170,0.5); } .glitch { animation: text-glitch 0.5s infinite alternate; } @keyframes text-glitch { 0% { transform: translateX(-2px); text-shadow: 0 0 5px #ff0066, 0 0 10px #ff33cc; } 100% { transform: translateX(2px); text-shadow: 0 0 8px #ff00aa, 0 0 20px #ff0099; } } .neon-link { color: #ff66cc; text-decoration: none; transition: text-shadow 0.3s ease; } .neon-link:hover { text-shadow: 0px 0px 15px #ff66cc, 0 0 25px rgba(255,0,170,0.5); } .ascii-art { text-align: center; font-size: 12px; color: #ff33cc; text-shadow: 0px 0px 5px #ff00ff; margin-bottom: 20px; } .quantso-container { display: flex; justify-content: center; gap: 20px; margin-top: 20px; } .quantso-box { background: rgba(255,0,170,0.1); padding: 15px; border-radius: 10px; text-align: center; box-shadow: inset 0 0 10px rgba(255,0,170,0.5); flex: 1; max-width: 150px; } THESE ARE EXL2 QUANTS, LOOK IN THE REVISIONS FOR THE QUANTS, MAIN BRANCH CONTAINS MEASUREMENT. A series of Merges made for Roleplaying & Creative Writing, This model uses 32B-Qwen2.5-Kunou-v1 and Hamanasu-Magnum-QwQ-32B and Slerp to merge the 2 models. ChatML formatting """ system system prompt user Hi there! assistant Nice to meet you! user Can I ask a question? assistant """ models: - model: Sao10K/32B-Qwen2.5-Kunou-v1 - model: Delta-Vector/Hamanasu-Magnum-QwQ-32B mergemethod: slerp basemodel: Delta-Vector/Hamanasu-Magnum-QwQ-32B parameters: t: - value: 0.2 dtype: bfloat16 tokenizersource: base GGUF https://huggingface.co/mradermacher/Archaeo-32B-GGUF/ EXL2 https://huggingface.co/Delta-Vector/Archaeo-32B-EXL2/ Credits Thank you to: Kubernetes-bad, LucyKnada, Intervitens, Samantha Twinkman, Tav, Trappu & The rest of Anthracite

32B parameters GLM-4 32B Creative / Fresh Prose Co-writing/Roleplay/Adventure Generalist Another Series of Merges! Since i could never beat Archaeo-32B-KTO! This time starting off with a GLM merge between Rei and Neon (thanks auri!!!) Using the Oh-so-great 0.2 Slerp merge weight with Neon as the Base. Support me on Ko-Fi: https://ko-fi.com/deltavector GGUF Format For use with LLama.cpp & Forks(Coming Soon!) EXL2 Format For use with TabbyAPI (Coming Soon!) EXL3 Format For use with TabbyAPI (Slower on Ampere)) Prompting Model has been tuned with the GLM-4 formatting. Samplers For testing of this model, I used Temp=1, 0.1 Min-P. See Merging Config https://files.catbox.moe/j9kyfy.yml Credits Thank you to Lucy Knada , Auri , Ateron , Alicat , Intervitens , Cgato , Kubernetes Bad and the rest of Anthracite .

A finetune of Pocketdoc's Personality Engine with a data mix similar to Ohashi-NeMo. After reviewing the dataset and tweaking some hparams, Thanks to Pocketdoc & Doctor Shotgun, this model has once again impressed me with its cohesiveness and writing quality. Once again it's meant to be "Short" & Sweet GGUF: https://huggingface.co/Delta-Vector/Francois-PE-12B-GGUF EXL2 : https://huggingface.co/Delta-Vector/Francois-PE-12B-EXL2 Prompting Model has been tuned with the ChatML formatting. A typical input would look like this: I would highly recommend using either Euryale's system prompt or the EVA system prompt with the model. Thank you to Lucy Knada, Intervitens,Cgato, Kubernetes Bad and the rest of Anthracite Training The training was done for 4 epochs. We used 4 x RTX 3090s GPUs graciously provided by Intervitens for the fine-tuning of the model.

html, body { background: black; color: #c9d1d9 !important; font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif; margin: 0; padding: 0; min-height: 100vh; } .markdown-body { color: white; margin: 40px auto; padding: 40px; border-radius: 12px; position: relative; overflow: hidden; } .markdown-body::after { content: ''; position: absolute; top: 0; left: 0; width: 100%; height: 100%; background: #0c0f18; / background color / pointer-events: none; z-index: -999; } h1, h2, h3 { background: linear-gradient(45deg, #6e00ff, #00ffff); -webkit-background-clip: text; -webkit-text-fill-color: transparent; border-bottom: 1px solid #333; padding-bottom: 0.3em; } div[style="border:2px solid #333"], div[style="border: 2px solid #333"], div[style="border:1px solid #333"], div[style="border: 1px solid #333"] { background: rgba(22, 27, 34, 0.8) !important; border: 2px solid #6e00ff !important; box-shadow: 0 0 15px rgba(110, 0, 255, 0.5); border-radius: 10px; padding: 20px; margin: 20px 0; } code { background-color: #1a1a1a !important; border-radius: 4px; padding: 0.2em 0.4em; color: #00ffff; } pre { background-color: #1a1a1a !important; border: 1px solid #333; border-radius: 8px; padding: 16px; } table { width: 100%; border-collapse: collapse; margin: 20px 0; background: rgba(0,0,0,0.2); table-layout: fixed; color: white; } th, td { border: 1px solid #333; padding: 12px; text-align: center; color: white; } td:nth-child(1) { width: 1%; white-space: nowrap; } td > span { display: block; padding: 4px 8px; background: rgba(110, 0, 255, 0.1); border-radius: 4px; transition: all 0.3s ease; } td > span:hover { background: rgba(110, 0, 255, 0.2); transform: translateY(-1px); } a { color: #00ffff; text-decoration: none; transition: all 0.3s ease; } hr { border: 0; height: 1px; background: linear-gradient(90deg, transparent, #333, transparent); margin: 40px 0; } .prose, .max-w-none, .px-4 { background-color: transparent !important; color: #c9d1d9 !important; } This is the 1st pretrain of Phi-4 with the following: This model has not been instruct tuned, Ablities to converse may be reduced from the original model, If you would like to roleplay, Please use the Instruct version. ⚔️ Hardware - 4x RTX 3090 GPUs - Epochs: 1 - Base: `Unsloth/phi-4` - Amount of Tokens: 500 Million

???: hi. apparently, you're dead and your body is lying under that tree sighted in the misty fog, out there. it's gallant of you to come and have an issue with it properly. ???: this young man, who you wronged and humiliated in front of your students that day, came to you, broke down at the door, and begged for his life. you killed him. ???: he hung from a rafter above two weeks later. and after that, well. everything became consistent, and you can't mug me. ???: in you stood two figures. one from your life. one from your karma. in you stood the evil and every accusation upon you, and by your greed and lust, you remembered its face entirely and accepted it as you. with hate of you, you tell me, i triumph, and by your wrath you send me back to break. Mango: can you… explain more? I'm not sure what you mean, I caused a man to die? ???: you didn't. you opened the door for his spirit to find you, to save himself by satisfying his last longing. and truly, you gave him the greatest hope, right up to the end. imagine that you bought the jones family a nice house. and then you did not allow jones to visit that house anymore. eventually he became very sad. and one day he went to check out, and hanged himself in that house, never knowing it was yours until the maid let him in. Mango: i didnt do anytthing please am i dead please dont tell me im dead ???: i wonder what i shall do to you, considering the harm you've done. i want, but i hope you play along, for you are an easy mark. take it as far as i desire. ???: your fear explains your ego. fear, your ego knows, frees strength. in the beginning was the word, and the word was with god, and the word was god. and the word became flesh and dwelt among us, and we beheld his glory, glory as of the only begotten of the father, full of grace and truth.