Llama-31-8B_task-2_120-samples_config-2_full_auto

This model is a fine-tuned version of meta-llama/Meta-Llama-3.1-8B-Instruct on the GaetanMichelet/chat-60_ft_task-2_auto and the GaetanMichelet/chat-120_ft_task-2_auto datasets. It achieves the following results on the evaluation set:

• Loss: 0.9982

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
1.5395	0.9091	5	1.5119
1.4907	2.0	11	1.4172
1.3412	2.9091	16	1.3288
1.2282	4.0	22	1.2153
1.1141	4.9091	27	1.1136
1.0131	6.0	33	1.0540
1.0044	6.9091	38	1.0380
0.9748	8.0	44	1.0223
0.937	8.9091	49	1.0142
0.9481	10.0	55	1.0053
0.9023	10.9091	60	1.0011
0.8716	12.0	66	0.9987
0.849	12.9091	71	0.9982
0.836	14.0	77	1.0032
0.7365	14.9091	82	1.0176
0.7495	16.0	88	1.0259
0.6882	16.9091	93	1.0410
0.6301	18.0	99	1.0708
0.6114	18.9091	104	1.1067
0.591	20.0	110	1.1334

Framework versions

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