decilm-7b-ocn-v1

This model is a fine-tuned version of Deci/DeciLM-7B on the oncc_medqa_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: 10
• 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=Deci/DeciLM-7B,trust_remote_code=True,parallelize=True,load_in_4bit=True,peft=chenhugging/decilm-7b-ocn-v1), gen_kwargs: (None), limit: 100.0, num_fewshot: None, batch_size: 1

Tasks	Version	Filter	n-shot	Metric	Value		Stderr
pubmedqa	1	none	0	acc	0.95	±	0.0219
medmcqa	Yaml	none	0	acc	0.34	±	0.0476
professional_medicine	0	none	0	acc	0.58	±	0.0496
college_medicine	0	none	0	acc	0.54	±	0.0501
clinical_knowledge	0	none	0	acc	0.57	±	0.0498
ocn	Yaml	none	0	acc	0.68	±	0.0469
aocnp	Yaml	none	0	acc	0.43	±	0.0498

Appendix: Original Performance

hf (pretrained=Deci/DeciLM-7B,trust_remote_code=True,parallelize=True,load_in_4bit=True), gen_kwargs: (None), limit: 100.0, num_fewshot: None, batch_size: 1

Tasks	Version	Filter	n-shot	Metric	Value		Stderr
pubmedqa	1	none	0	acc	0.89	±	0.0314
medmcqa	Yaml	none	0	acc	0.37	±	0.0485
professional_medicine	0	none	0	acc	0.58	±	0.0496
college_medicine	0	none	0	acc	0.54	±	0.0501
clinical_knowledge	0	none	0	acc	0.63	±	0.0485
ocn	Yaml	none	0	acc	0.64	±	0.0482
aocnp	Yaml	none	0	acc	0.41	±	0.0494