Llama-31-8B_task-3_60-samples_config-3

This model is a fine-tuned version of meta-llama/Meta-Llama-3.1-8B-Instruct on the GaetanMichelet/chat-60_ft_task-3 dataset. It achieves the following results on the evaluation set:

• Loss: 0.4657

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: 1e-05
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• gradient_accumulation_steps: 8
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 150

Training results

Training Loss	Epoch	Step	Validation Loss
2.4489	0.8696	5	2.4703
2.4891	1.9130	11	2.4545
2.4697	2.9565	17	2.4172
2.5196	4.0	23	2.3543
2.4481	4.8696	28	2.2785
2.2925	5.9130	34	2.1509
2.0169	6.9565	40	1.9728
1.6364	8.0	46	1.7441
1.6047	8.8696	51	1.5192
1.3126	9.9130	57	1.2353
1.0406	10.9565	63	0.9674
0.8254	12.0	69	0.7402
0.7723	12.8696	74	0.6372
0.4327	13.9130	80	0.5750
0.279	14.9565	86	0.5313
0.4039	16.0	92	0.5047
0.5227	16.8696	97	0.4883
0.3704	17.9130	103	0.4936
0.497	18.9565	109	0.4907
0.4428	20.0	115	0.4828
0.3477	20.8696	120	0.4825
0.4628	21.9130	126	0.4857
0.367	22.9565	132	0.4827
0.2597	24.0	138	0.4768
0.3473	24.8696	143	0.4684
0.3795	25.9130	149	0.4657
0.437	26.9565	155	0.4706
0.3478	28.0	161	0.4760
0.254	28.8696	166	0.4745
0.3934	29.9130	172	0.4813
0.3074	30.9565	178	0.4815
0.3447	32.0	184	0.4823
0.225	32.8696	189	0.4865

Framework versions

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