Gemma-2-2B_task-1_60-samples_config-2_full_auto

This model is a fine-tuned version of google/gemma-2-2b-it on the GaetanMichelet/chat-60_ft_task-1_auto dataset. It achieves the following results on the evaluation set:

• Loss: 0.9476

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: 0.0001
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• gradient_accumulation_steps: 16
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss
2.3291	0.6957	2	2.2148
2.3149	1.7391	5	2.1203
2.0527	2.7826	8	1.7995
1.8287	3.8261	11	1.4930
1.4013	4.8696	14	1.3096
1.2792	5.9130	17	1.1454
1.0352	6.9565	20	1.0580
0.9755	8.0	23	1.0166
0.9229	8.6957	25	0.9997
0.8475	9.7391	28	0.9786
0.8219	10.7826	31	0.9607
0.7598	11.8261	34	0.9497
0.7358	12.8696	37	0.9476
0.6627	13.9130	40	0.9562
0.6456	14.9565	43	0.9710
0.5933	16.0	46	0.9923
0.5504	16.6957	48	1.0152
0.5298	17.7391	51	1.0546
0.4504	18.7826	54	1.1058
0.4508	19.8261	57	1.1510

Framework versions

• PEFT 0.12.0
• Transformers 4.44.0
• Pytorch 2.1.2+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1