tsavage68/400STEPS_05beta_1e7rate_Meditron7B_zerozhot

400STEPS_05beta_1e7rate_Meditron7B

This model is a fine-tuned version of epfl-llm/meditron-7b on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.6864
• Rewards/chosen: 0.0004
• Rewards/rejected: -0.0144
• Rewards/accuracies: 0.4945
• Rewards/margins: 0.0148
• Logps/rejected: -27.8226
• Logps/chosen: -26.4806
• Logits/rejected: -0.6110
• Logits/chosen: -0.6109

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-07
• train_batch_size: 4
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 100
• training_steps: 400

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.6941	0.1	50	0.6931	-0.0003	-0.0011	0.4044	0.0008	-27.7959	-26.4820	-0.6106	-0.6104
0.6927	0.2	100	0.6912	-0.0047	-0.0093	0.4769	0.0046	-27.8123	-26.4908	-0.6105	-0.6104
0.6838	0.29	150	0.6896	-0.0023	-0.0105	0.5077	0.0082	-27.8146	-26.4860	-0.6101	-0.6100
0.6906	0.39	200	0.6886	-0.0007	-0.0107	0.4989	0.0100	-27.8151	-26.4828	-0.6109	-0.6108
0.6789	0.49	250	0.6877	-0.0035	-0.0154	0.5121	0.0119	-27.8245	-26.4884	-0.6111	-0.6110
0.6853	0.59	300	0.6852	0.0012	-0.0160	0.5297	0.0172	-27.8257	-26.4791	-0.6112	-0.6111
0.6805	0.68	350	0.6877	-0.0039	-0.0162	0.4725	0.0122	-27.8260	-26.4893	-0.6112	-0.6110
0.6936	0.78	400	0.6864	0.0004	-0.0144	0.4945	0.0148	-27.8226	-26.4806	-0.6110	-0.6109

Framework versions

• Transformers 4.37.2
• Pytorch 2.0.0+cu117
• Datasets 2.17.0
• Tokenizers 0.15.1