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

• Loss: 0.9617

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
1.3564	0.6957	2	1.3659
1.3765	1.7391	5	1.3404
1.3125	2.7826	8	1.2526
1.2213	3.8261	11	1.1815
1.1432	4.8696	14	1.1329
1.0949	5.9130	17	1.0849
1.0142	6.9565	20	1.0435
0.9925	8.0	23	1.0153
0.9508	8.6957	25	1.0030
0.9191	9.7391	28	0.9871
0.9172	10.7826	31	0.9778
0.892	11.8261	34	0.9705
0.8585	12.8696	37	0.9655
0.8535	13.9130	40	0.9630
0.8316	14.9565	43	0.9618
0.8242	16.0	46	0.9617
0.7836	16.6957	48	0.9622
0.7962	17.7391	51	0.9640
0.7851	18.7826	54	0.9657
0.7553	19.8261	57	0.9702
0.7474	20.8696	60	0.9767
0.7352	21.9130	63	0.9822
0.7191	22.9565	66	0.9871

Framework versions

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