Llama-31-8B_task-1_180-samples_config-3_full

This model is a fine-tuned version of meta-llama/Meta-Llama-3.1-8B-Instruct on the GaetanMichelet/chat-60_ft_task-1, the GaetanMichelet/chat-120_ft_task-1 and the GaetanMichelet/chat-180_ft_task-1 datasets. It achieves the following results on the evaluation set:

• Loss: 0.8992

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: 1e-05
• 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: 150

Training results

Training Loss	Epoch	Step	Validation Loss
2.4542	1.0	17	2.4259
2.4022	2.0	34	2.3882
2.3317	3.0	51	2.3140
2.2607	4.0	68	2.2050
2.1352	5.0	85	2.0643
1.9456	6.0	102	1.8885
1.7528	7.0	119	1.7025
1.4935	8.0	136	1.4674
1.2733	9.0	153	1.2421
1.1154	10.0	170	1.1134
1.1202	11.0	187	1.0689
0.9449	12.0	204	1.0450
0.9973	13.0	221	1.0253
1.0562	14.0	238	1.0091
0.9947	15.0	255	0.9928
1.0096	16.0	272	0.9804
0.9222	17.0	289	0.9692
0.8838	18.0	306	0.9603
0.8942	19.0	323	0.9511
0.9058	20.0	340	0.9432
0.8837	21.0	357	0.9354
0.795	22.0	374	0.9315
0.8395	23.0	391	0.9243
0.8308	24.0	408	0.9169
0.7863	25.0	425	0.9138
0.7468	26.0	442	0.9068
0.7658	27.0	459	0.9008
0.7128	28.0	476	0.8992
0.6474	29.0	493	0.9064
0.6387	30.0	510	0.9089
0.6846	31.0	527	0.9096
0.6424	32.0	544	0.9173
0.6598	33.0	561	0.9238
0.6634	34.0	578	0.9290
0.5893	35.0	595	0.9400

Framework versions

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