Mistral-7B_task-3_120-samples_config-1_full_auto

This model is a fine-tuned version of mistralai/Mistral-7B-Instruct-v0.3 on the GaetanMichelet/chat-60_ft_task-3_auto and the GaetanMichelet/chat-120_ft_task-3_auto datasets. It achieves the following results on the evaluation set:

• Loss: 0.8507

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: 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: 50

Training results

Training Loss	Epoch	Step	Validation Loss
1.1614	1.0	11	1.1416
0.9734	2.0	22	1.0178
0.8716	3.0	33	0.8912
0.8725	4.0	44	0.8626
0.8125	5.0	55	0.8507
0.7541	6.0	66	0.8533
0.6871	7.0	77	0.8667
0.6206	8.0	88	0.9129
0.4919	9.0	99	0.9647
0.381	10.0	110	1.0133
0.3491	11.0	121	1.0941
0.2613	12.0	132	1.1191

Framework versions

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