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

• Loss: 0.8633

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
1.3338	0.9091	5	1.3082
1.3005	2.0	11	1.2187
1.1249	2.9091	16	1.1211
1.0441	4.0	22	1.0411
0.9601	4.9091	27	0.9654
0.8731	6.0	33	0.9165
0.8553	6.9091	38	0.8962
0.8229	8.0	44	0.8815
0.7958	8.9091	49	0.8724
0.7945	10.0	55	0.8658
0.7557	10.9091	60	0.8634
0.7322	12.0	66	0.8633
0.7081	12.9091	71	0.8664
0.6938	14.0	77	0.8733
0.6136	14.9091	82	0.8832
0.6249	16.0	88	0.8980
0.573	16.9091	93	0.9280
0.5309	18.0	99	0.9385
0.5198	18.9091	104	0.9874

Framework versions

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