Gemma-2-2B_task-3_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-3_auto dataset. It achieves the following results on the evaluation set:

• Loss: 0.9567

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
1.3458	0.6957	2	1.3568
1.3656	1.7391	5	1.3315
1.3043	2.7826	8	1.2450
1.2137	3.8261	11	1.1760
1.1349	4.8696	14	1.1266
1.0892	5.9130	17	1.0786
1.0063	6.9565	20	1.0362
0.9866	8.0	23	1.0086
0.9437	8.6957	25	0.9958
0.9105	9.7391	28	0.9805
0.9086	10.7826	31	0.9711
0.884	11.8261	34	0.9645
0.852	12.8696	37	0.9601
0.8465	13.9130	40	0.9575
0.8197	14.9565	43	0.9567
0.8147	16.0	46	0.9571
0.7741	16.6957	48	0.9576
0.7843	17.7391	51	0.9597
0.7714	18.7826	54	0.9630
0.7444	19.8261	57	0.9685
0.7328	20.8696	60	0.9758
0.7226	21.9130	63	0.9809

Framework versions

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