MedQA_L3_150steps_1e6rate_03beat_CSFTDPO

This model is a fine-tuned version of meta-llama/Meta-Llama-3-8B-Instruct on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.5020
• Rewards/chosen: -0.9020
• Rewards/rejected: -1.9172
• Rewards/accuracies: 0.7297
• Rewards/margins: 1.0152
• Logps/rejected: -27.7072
• Logps/chosen: -21.2293
• Logits/rejected: -1.0337
• Logits/chosen: -1.0327

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-06
• train_batch_size: 2
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 4
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 100
• training_steps: 150

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.7072	0.0489	50	0.6474	0.1422	0.0242	0.6505	0.1180	-21.2360	-17.7487	-0.9397	-0.9391
0.6194	0.0977	100	0.5755	-0.5279	-1.1917	0.6989	0.6638	-25.2888	-19.9824	-1.0174	-1.0166
0.5632	0.1466	150	0.5020	-0.9020	-1.9172	0.7297	1.0152	-27.7072	-21.2293	-1.0337	-1.0327

Framework versions

• Transformers 4.41.0
• Pytorch 2.0.0+cu117
• Datasets 2.19.1
• Tokenizers 0.19.1