Llama-31-8B_task-2_60-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 dataset. It achieves the following results on the evaluation set:

• Loss: 1.0291

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
1.5031	0.6957	2	1.5181
1.4986	1.7391	5	1.4719
1.4173	2.7826	8	1.3915
1.3451	3.8261	11	1.3186
1.2588	4.8696	14	1.2547
1.2035	5.9130	17	1.1920
1.1045	6.9565	20	1.1348
1.074	8.0	23	1.0886
1.0261	8.6957	25	1.0743
0.9858	9.7391	28	1.0616
0.9944	10.7826	31	1.0531
0.9706	11.8261	34	1.0463
0.9367	12.8696	37	1.0399
0.9372	13.9130	40	1.0353
0.9054	14.9565	43	1.0335
0.9095	16.0	46	1.0299
0.8691	16.6957	48	1.0291
0.8796	17.7391	51	1.0303
0.865	18.7826	54	1.0309
0.8473	19.8261	57	1.0353
0.8304	20.8696	60	1.0361
0.8259	21.9130	63	1.0386
0.8022	22.9565	66	1.0422
0.7948	24.0	69	1.0466

Framework versions

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