Llama-31-8B_task-3_180-samples_config-4

This model is a fine-tuned version of meta-llama/Meta-Llama-3.1-8B-Instruct on the GaetanMichelet/chat-60_ft_task-3, the GaetanMichelet/chat-120_ft_task-3 and the GaetanMichelet/chat-180_ft_task-3 datasets. It achieves the following results on the evaluation set:

• Loss: 0.4942

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: 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: 150

Training results

Training Loss	Epoch	Step	Validation Loss
2.5495	0.9412	8	2.5077
2.3733	2.0	17	2.4775
2.46	2.9412	25	2.4218
2.4585	4.0	34	2.3278
2.2624	4.9412	42	2.1919
2.0553	6.0	51	1.9704
1.7403	6.9412	59	1.7066
1.3756	8.0	68	1.3617
1.11	8.9412	76	1.0613
0.7161	10.0	85	0.7772
0.7609	10.9412	93	0.6787
0.4358	12.0	102	0.6182
0.4774	12.9412	110	0.5912
0.5569	14.0	119	0.5746
0.427	14.9412	127	0.5487
0.4672	16.0	136	0.5339
0.3495	16.9412	144	0.5525
0.4731	18.0	153	0.5323
0.3913	18.9412	161	0.5243
0.5624	20.0	170	0.5253
0.4684	20.9412	178	0.5222
0.3029	22.0	187	0.5100
0.3522	22.9412	195	0.5085
0.3855	24.0	204	0.4971
0.317	24.9412	212	0.5049
0.338	26.0	221	0.5016
0.391	26.9412	229	0.4942
0.3964	28.0	238	0.5010
0.2951	28.9412	246	0.5098
0.4021	30.0	255	0.5068
0.4021	30.9412	263	0.5070
0.3456	32.0	272	0.5025
0.4431	32.9412	280	0.5050
0.4131	34.0	289	0.5094

Framework versions

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