Gemma-2-2B_task-2_180-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, the GaetanMichelet/chat-120_ft_task-2_auto and the GaetanMichelet/chat-180_ft_task-2_auto datasets. It achieves the following results on the evaluation set:

• Loss: 0.8584

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
1.3006	0.9412	8	1.2976
1.2057	2.0	17	1.1583
1.1024	2.9412	25	1.0637
0.9139	4.0	34	0.9544
0.8515	4.9412	42	0.9081
0.863	6.0	51	0.8835
0.8439	6.9412	59	0.8703
0.7927	8.0	68	0.8620
0.7343	8.9412	76	0.8584
0.7243	10.0	85	0.8591
0.6957	10.9412	93	0.8661
0.6372	12.0	102	0.8781
0.6342	12.9412	110	0.8993
0.6126	14.0	119	0.9278
0.5304	14.9412	127	0.9825
0.4754	16.0	136	1.0305

Framework versions

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