|
--- |
|
library_name: peft |
|
datasets: |
|
- BI55/MedText |
|
pipeline_tag: text-generation |
|
tags: |
|
- llm |
|
- llama2 |
|
- medical |
|
--- |
|
|
|
# Llama2 🦙 finetuned on medical diagnosis |
|
MedText dataset: https://huggingface.co/datasets/BI55/MedText |
|
|
|
1412 pairs of diagnosis cases |
|
|
|
|
|
# About: |
|
The primary objective of this fine-tuning process is to equip Llama2 with the ability to assist in diagnosing various medical cases and diseases. |
|
However, it is essential to clarify that it is not designed to replace real medical professionals. Instead, its purpose is to provide helpful information to users, |
|
suggesting potential next steps based on the input data and the patterns it has learned from the MedText dataset. |
|
|
|
Finetuned on guanaco styled instructions |
|
|
|
``` |
|
###Human |
|
###Assistant |
|
``` |
|
## Training procedure |
|
|
|
|
|
The following `bitsandbytes` quantization config was used during training: |
|
- load_in_8bit: False |
|
- load_in_4bit: True |
|
- llm_int8_threshold: 6.0 |
|
- llm_int8_skip_modules: None |
|
- llm_int8_enable_fp32_cpu_offload: False |
|
- llm_int8_has_fp16_weight: False |
|
- bnb_4bit_quant_type: nf4 |
|
- bnb_4bit_use_double_quant: False |
|
- bnb_4bit_compute_dtype: float16 |
|
### Framework versions |
|
|
|
|
|
- PEFT 0.5.0.dev0 |
