Llama-31-8B_task-3_120-samples_config-1_full

This model is a fine-tuned version of meta-llama/Meta-Llama-3.1-8B-Instruct on the GaetanMichelet/chat-60_ft_task-3 and the GaetanMichelet/chat-120_ft_task-3 datasets. It achieves the following results on the evaluation set:

• Loss: 1.1303

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: 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: 50

Training results

Training Loss	Epoch	Step	Validation Loss
1.6541	1.0	11	1.6580
1.4606	2.0	22	1.5146
1.3657	3.0	33	1.3458
1.244	4.0	44	1.2034
1.1606	5.0	55	1.1657
1.1108	6.0	66	1.1446
1.062	7.0	77	1.1341
1.0372	8.0	88	1.1303
0.9427	9.0	99	1.1413
0.819	10.0	110	1.1795
0.8148	11.0	121	1.2101
0.6121	12.0	132	1.2798
0.6309	13.0	143	1.3036
0.3915	14.0	154	1.4030
0.3014	15.0	165	1.4186

Framework versions

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