Gemma-2-2B_task-3_120-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-3_auto and the GaetanMichelet/chat-120_ft_task-3_auto datasets. It achieves the following results on the evaluation set:

• Loss: 0.9359

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
1.4057	0.9091	5	1.3750
1.3708	2.0	11	1.2889
1.2158	2.9091	16	1.1944
1.1179	4.0	22	1.1044
1.0198	4.9091	27	1.0251
0.9467	6.0	33	0.9804
0.9148	6.9091	38	0.9615
0.8909	8.0	44	0.9481
0.8645	8.9091	49	0.9407
0.867	10.0	55	0.9363
0.8193	10.9091	60	0.9359
0.7911	12.0	66	0.9404
0.7557	12.9091	71	0.9464
0.7439	14.0	77	0.9614
0.6475	14.9091	82	0.9828
0.6577	16.0	88	1.0100
0.5958	16.9091	93	1.0415
0.5391	18.0	99	1.0821

Framework versions

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