Mistral-7B_task-3_60-samples_config-2_full

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

• Loss: 0.8737

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
1.2329	0.6957	2	1.2277
1.2254	1.7391	5	1.1423
1.0984	2.7826	8	1.0654
1.0266	3.8261	11	1.0052
0.948	4.8696	14	0.9384
0.884	5.9130	17	0.9020
0.8253	6.9565	20	0.8885
0.822	8.0	23	0.8788
0.7994	8.6957	25	0.8754
0.7669	9.7391	28	0.8737
0.7596	10.7826	31	0.8738
0.7256	11.8261	34	0.8750
0.6803	12.8696	37	0.8834
0.6584	13.9130	40	0.8951
0.6131	14.9565	43	0.9023
0.5777	16.0	46	0.9154
0.5158	16.6957	48	0.9439

Framework versions

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