Llama-31-8B_task-2_180-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_t2, the GaetanMichelet/chat_120_ft_t2 and the GaetanMichelet/chat_180_ft_t2 datasets. It achieves the following results on the evaluation set:

• Loss: 1.2302

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
1.5147	1.0	25	1.5097
1.2938	2.0	50	1.3123
1.2015	3.0	75	1.2619
1.004	4.0	100	1.2593
0.8644	5.0	125	1.2485
0.5633	6.0	150	1.2632
0.5237	7.0	175	1.2302
0.3902	8.0	200	1.2869
0.3029	9.0	225	1.3451
0.2764	10.0	250	1.4304
0.2385	11.0	275	1.5122
0.2049	12.0	300	1.5670
0.1658	13.0	325	1.6368
0.1536	14.0	350	1.6687

Framework versions

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