Llama-31-8B_task-2_60-samples_config-3

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

• Loss: 0.7156

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
1.0751	0.8696	5	1.0960
1.1004	1.9130	11	1.0923
1.1552	2.9565	17	1.0845
1.0884	4.0	23	1.0731
1.0984	4.8696	28	1.0593
1.054	5.9130	34	1.0363
0.9646	6.9565	40	1.0060
0.9982	8.0	46	0.9700
0.9649	8.8696	51	0.9380
0.9161	9.9130	57	0.9017
0.8966	10.9565	63	0.8722
0.8314	12.0	69	0.8468
0.7747	12.8696	74	0.8286
0.8162	13.9130	80	0.8081
0.8422	14.9565	86	0.7906
0.7802	16.0	92	0.7776
0.7179	16.8696	97	0.7692
0.7191	17.9130	103	0.7605
0.6644	18.9565	109	0.7524
0.6898	20.0	115	0.7456
0.6776	20.8696	120	0.7404
0.6571	21.9130	126	0.7338
0.6177	22.9565	132	0.7289
0.6361	24.0	138	0.7246
0.6357	24.8696	143	0.7214
0.6767	25.9130	149	0.7174
0.5947	26.9565	155	0.7170
0.6182	28.0	161	0.7156
0.5899	28.8696	166	0.7157
0.5612	29.9130	172	0.7162
0.5545	30.9565	178	0.7185
0.5574	32.0	184	0.7232
0.5316	32.8696	189	0.7254
0.5276	33.9130	195	0.7338
0.4653	34.9565	201	0.7407

Framework versions

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