metadata

license: other
library_name: peft
tags:
  - llama-factory
  - lora
  - generated_from_trainer
base_model: mistralai/Mistral-7B-v0.1
model-index:
  - name: mistral-7b-medmcqa-inst-v1
    results: []

mistral-7b-medmcqa-inst-v1

This model is a fine-tuned version of mistralai/Mistral-7B-v0.1 on the medmcqa_instruct dataset.

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0005
train_batch_size: 4
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 16
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_steps: 20
num_epochs: 1.0
mixed_precision_training: Native AMP

Framework versions

PEFT 0.8.2
Transformers 4.37.2
Pytorch 2.1.1+cu121
Datasets 2.16.1
Tokenizers 0.15.1

Performance

hf (pretrained=mistralai/Mistral-7B-v0.1,parallelize=True,load_in_4bit=True,peft=chenhugging/mistral-7b-medmcqa-inst-v1), gen_kwargs: (None), limit: 100.0, num_fewshot: None, batch_size: 1

Tasks	Version	Filter	Metric	Value		Stderr
pubmedqa	1	none	acc	0.98	±	0.0141
medmcqa	Yaml	none	acc	0.48	±	0.0502
professional_medicine	0	none	acc	0.61	±	0.0490
college_medicine	0	none	acc	0.57	±	0.0498
clinical_knowledge	0	none	acc	0.65	±	0.0479
ocn	Yaml	none	acc	0.68	±	0.0469
aocnp	Yaml	none	acc	0.56	±	0.0499

Original Performance (mistralai/Mistral-7B-v0.1)

hf (pretrained=mistralai/Mistral-7B-v0.1,parallelize=True,load_in_4bit=True), gen_kwargs: (None), limit: 100.0, num_fewshot: None, batch_size: 1

Tasks	Version	Filter	Metric	Value		Stderr
pubmedqa	1	none	acc	0.98	±	0.0141
medmcqa	Yaml	none	acc	0.45	±	0.0500
professional_medicine	0	none	acc	0.64	±	0.0482
college_medicine	0	none	acc	0.65	±	0.0479
clinical_knowledge	0	none	acc	0.68	±	0.0469
ocn	Yaml	none	acc	0.62	±	0.0488
aocnp	Yaml	none	acc	0.47	±	0.0502