Llama-31-8B_task-3_120-samples_config_1

This model is a fine-tuned version of meta-llama/Meta-Llama-3.1-8B-Instruct on the GaetanMichelet/chat_60_ft_t3 and the GaetanMichelet/chat_120_ft_t3 datasets. It achieves the following results on the evaluation set:

• Loss: 1.3433

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: 0.0002
• train_batch_size: 1
• eval_batch_size: 2
• 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
1.542	0.9057	6	1.6359
1.5247	1.9623	13	1.5617
1.4951	2.8679	19	1.4805
1.3218	3.9245	26	1.3765
1.2227	4.9811	33	1.3440
1.1431	5.8868	39	1.3433
1.0279	6.9434	46	1.3489
0.9243	8.0	53	1.4227
0.7076	8.9057	59	1.4531
0.6006	9.9623	66	1.4703
0.4636	10.8679	72	1.6322
0.3418	11.9245	79	1.8405
0.2816	12.9811	86	1.8540

Framework versions

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