MelloGPT/README.md

license, datasets, model-index

license	datasets	model-index
mit	nbertagnolli/counsel-chat	name results MelloGPT task dataset metrics source type name text-generation Text Generation name type config split args AI2 Reasoning Challenge (25-Shot) ai2_arc ARC-Challenge test num_few_shot 25 type value name acc_norm 53.84 normalized accuracy url name https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=steve-cse/MelloGPT Open LLM Leaderboard task dataset metrics source type name text-generation Text Generation name type split args HellaSwag (10-Shot) hellaswag validation num_few_shot 10 type value name acc_norm 76.12 normalized accuracy url name https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=steve-cse/MelloGPT Open LLM Leaderboard task dataset metrics source type name text-generation Text Generation name type config split args MMLU (5-Shot) cais/mmlu all test num_few_shot 5 type value name acc 55.99 accuracy url name https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=steve-cse/MelloGPT Open LLM Leaderboard task dataset metrics source type name text-generation Text Generation name type config split args TruthfulQA (0-shot) truthful_qa multiple_choice validation num_few_shot 0 type value mc2 55.61 url name https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=steve-cse/MelloGPT Open LLM Leaderboard task dataset metrics source type name text-generation Text Generation name type config split args Winogrande (5-shot) winogrande winogrande_xl validation num_few_shot 5 type value name acc 73.88 accuracy url name https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=steve-cse/MelloGPT Open LLM Leaderboard task dataset metrics source type name text-generation Text Generation name type config split args GSM8k (5-shot) gsm8k main test num_few_shot 5 type value name acc 30.1 accuracy url name https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=steve-cse/MelloGPT Open LLM Leaderboard

MelloGPT

NOTE: This model should not be regarded as a replacement for professional mental health assistance. It is essential to seek support from qualified professionals for personalized and appropriate care.

A fine tuned version of Mistral-7B-Instruct-v0.1 on counsel-chat dataset for mental health counseling conversations.

Motivation

In an era where mental health support is of paramount importance, A large language model fine-tuned on mental health counseling conversations stands as a pioneering solution. Leveraging a diverse dataset of anonymized counseling sessions, the model has been trained to recognize and respond to a wide range of mental health concerns. The fine-tuning process incorporates ethical considerations, privacy concerns, and sensitivity to the nuances of mental health conversations. The resulting model will demonstrate an intricate understanding of mental health issues and provide empathetic and supportive responses.

Prompt Template

<s>[INST] {prompt} [/INST]

Quantized Model

The quantized model can be found here. Thanks to @TheBloke.

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	57.59
AI2 Reasoning Challenge (25-Shot)	53.84
HellaSwag (10-Shot)	76.12
MMLU (5-Shot)	55.99
TruthfulQA (0-shot)	55.61
Winogrande (5-shot)	73.88
GSM8k (5-shot)	30.10

Contributions

This project is open for contributions. Feel free to use the community tab.

Inspiration

This project was inspired by the project(s) listed below:

companion_cube by @KnutJaegersberg

Credits

This is my first attempt at fine-tuning a large language model. It wouldn't be possible without Axolotl and Runpod. The axolotl config file can be found here.