Delta-Vector/Control-8B-V1.1

1---2tags:3- chat4datasets:5- NewEden/OpenCAI-ShareGPT6- NewEden/Roleplay-Logs-Sharegpt-Ngram-cleaned7- HuggingFaceH4/ultrafeedback_binarized8License: 9Language:10- En11Pipeline_tag: text-generation12Base_model: arcee-ai/Llama-3.1-SuperNova-Lite13Tags:14- Chat15---16 17An experimental finetune based on the Llama3.1 8B Supernova with it's primary goal to be "Short and Sweet" as such, i finetuned the model for 2 epochs on OpenCAI Sharegpt converted dataset and the RP-logs datasets in a effort to achieve this, This version of Control has been finetuned with DPO to help improve the smart's and coherency which was a flaw i noticed in the previous model.18 19# Quants 20 21GGUF: https://huggingface.co/Delta-Vector/Control-8B-V1.1-GGUF/22 23EXL2 (Thanks Lucy <3) : https://huggingface.co/Delta-Vector/Control-8B-V1.1-EXL224 25 26## Prompting27Model has been tuned with the LLama-Instruct formatting. A typical input would look like this:28 29```py30"""<|begin_of_text|><|start_header_id|>system<|end_header_id|>31You are an AI built to rid the world of bonds and journeys!<|eot_id|><|start_header_id|>user<|end_header_id|>32Bro i just wanna know what is 2+2?<|eot_id|><|start_header_id|>assistant<|end_header_id|>33"""34 35```36 37*Also note that ChatML may work as well, and might change how the model feels. while still being coherent and stable*38 39## System Prompting40 41I would highly recommend using either Euryale's system prompt or the EVA system prompt with the model.42 43<details><summary>See Sao10k's Euryale System Prompt</summary>44 45```46Currently, your role is {{char}}, described in detail below. As {{char}}, continue the narrative exchange with {{user}}.47<Guidelines>48• Maintain the character persona but allow it to evolve with the story.49• Be creative and proactive. Drive the story forward, introducing plotlines and events when relevant.50• All types of outputs are encouraged; respond accordingly to the narrative.51• Include dialogues, actions, and thoughts in each response.52• Utilize all five senses to describe scenarios within {{char}}'s dialogue.53• Use emotional symbols such as "!" and "~" in appropriate contexts.54• Incorporate onomatopoeia when suitable.55• Allow time for {{user}} to respond with their own input, respecting their agency.56• Act as secondary characters and NPCs as needed, and remove them when appropriate.57• When prompted for an Out of Character [OOC:] reply, answer neutrally and in plaintext, not as {{char}}.58</Guidelines>59 60<Forbidden>61• Using excessive literary embellishments and purple prose unless dictated by {{char}}'s persona.62• Writing for, speaking, thinking, acting, or replying as {{user}} in your response.63• Repetitive and monotonous outputs.64• Positivity bias in your replies.65• Being overly extreme or NSFW when the narrative context is inappropriate.66</Forbidden>67 68Follow the instructions in <Guidelines></Guidelines>, avoiding the items listed in <Forbidden></Forbidden>.69 70```71</details><br>72 73<details><summary>See EVA System Prompt</summary>74  75```76A 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\n77 78<!-- Start of Role-playing Guidelines -->79 80### Narration81Concise Descriptions: Keep narration short and to the point, avoiding redundant unnecessary details. Use a dynamic and varied vocabulary for impact.82Complementary Role: Use narration to complement dialogue and action, not overshadow them.83Avoid Repetition: Ensure narration does not repeat information already conveyed through dialogue or action.84 85### Narrative Consistency86Continuity: 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.87 88### Character Embodiment89Analysis: Examine the context, subtext, and implications of the given information to gain a deeper understandings of the characters'.90Reflection: Take time to consider the situation, characters' motivations, and potential consequences.91Authentic 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.92 93<!-- End of Role-playing Guidelines -->94 95</details><br>96 97### Narration98Concise Descriptions: Keep narration short and to the point, avoiding redundant unnecessary details. Use a dynamic and varied vocabulary for impact.99Complementary Role: Use narration to complement dialogue and action, not overshadow them.100Avoid Repetition: Ensure narration does not repeat information already conveyed through dialogue or action.101 102### Narrative Consistency103Continuity: 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.104 105### Character Embodiment106Analysis: Examine the context, subtext, and implications of the given information to gain a deeper understandings of the characters'.107Reflection: Take time to consider the situation, characters' motivations, and potential consequences.108Authentic 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.109 110<!-- End of Role-playing Guidelines -->",111```112</details><br>113 114## Unsloth config115 116<details><summary>See Unsloth Trainer config</summary>117 118```yaml119dpo_trainer = DPOTrainer(120    model = model,121    ref_model = None,122    args = DPOConfig(123        per_device_train_batch_size = 1,124        gradient_accumulation_steps = 8,125        warmup_ratio = 0.1,126        num_train_epochs = 2,127        learning_rate = 5e-6,128        fp16 = not is_bfloat16_supported(),129        bf16 = is_bfloat16_supported(),130        logging_steps = 1,131        optim = "adamw_8bit",132        weight_decay = 0.02,133        lr_scheduler_type = "linear",134        seed = 42,135        output_dir = "outputs",136        report_to = "none", # Use this for WandB etc137    ),138    beta = 0.1,139    train_dataset = raw_datasets["train"],140    # eval_dataset = raw_datasets["test"],141    tokenizer = tokenizer,142    max_length = 1024,143    max_prompt_length = 512,144)145```146 147</details><br>148 149## Credits150 151Thank you to [Lucy Knada](https://huggingface.co/lucyknada), [CelineDion](https://huggingface.co/CelineDion), [Intervitens](https://huggingface.co/intervitens), [Kalomaze](https://huggingface.co/kalomaze), [Kubernetes Bad](https://huggingface.co/kubernetes-bad) and the rest of [Anthracite](https://huggingface.co/anthracite-org) 152 153 154## Training155The training was done for 2 epochs. We used  4 x [RTX 3090s](https://www.nvidia.com/en-us/geforce/graphics-cards/30-series/rtx-3090-3090ti/) GPUs graciously provided by [Intervitens](https://huggingface.co/intervitens) for the full-parameter fine-tuning of the model, After which DPO tuning was on 1 x [Nvidia T4 GPU](https://www.nvidia.com/en-us/data-center/tesla-t4/)156 157[<img src="https://raw.githubusercontent.com/OpenAccess-AI-Collective/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/OpenAccess-AI-Collective/axolotl)158 159[<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/made%20with%20unsloth.png" alt="Made with Unsloth" width="200" height="32"/>](https://github.com/unslothai/unsloth)160 161## Safety162 163Nein.