Gemma-2-2B_task-1_60-samples_config-2_full

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

• Loss: 1.0277

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
2.6105	0.6957	2	2.4360
2.5655	1.7391	5	2.3295
2.2831	2.7826	8	1.9787
2.0137	3.8261	11	1.6425
1.5688	4.8696	14	1.4293
1.4129	5.9130	17	1.2381
1.1583	6.9565	20	1.1529
1.0951	8.0	23	1.1034
1.0201	8.6957	25	1.0816
0.9479	9.7391	28	1.0556
0.9142	10.7826	31	1.0383
0.8428	11.8261	34	1.0284
0.8092	12.8696	37	1.0277
0.7383	13.9130	40	1.0369
0.7168	14.9565	43	1.0502
0.6614	16.0	46	1.0692
0.6166	16.6957	48	1.0939
0.6045	17.7391	51	1.1363
0.5177	18.7826	54	1.1785
0.5067	19.8261	57	1.2188

Framework versions

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