mistral-lora

This model is a fine-tuned version of mistralai/Mistral-7B-Instruct-v0.2 on the generator dataset. It achieves the following results on the evaluation set:

• Loss: 0.6352

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: 0.0003
• train_batch_size: 8
• eval_batch_size: 8
• seed: 1399
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 100
• training_steps: 500

Training results

Training Loss	Epoch	Step	Validation Loss
1.7822	0.0221	20	1.6700
1.6214	0.0443	40	1.5494
1.5335	0.0664	60	1.4848
1.4679	0.0885	80	1.4369
1.4343	0.1107	100	1.3875
1.355	0.1328	120	1.3318
1.3334	0.1550	140	1.2785
1.2573	0.1771	160	1.2225
1.1879	0.1992	180	1.1666
1.1678	0.2214	200	1.1121
1.0723	0.2435	220	1.0540
1.049	0.2656	240	0.9982
0.9674	0.2878	260	0.9444
0.9192	0.3099	280	0.8952
0.8752	0.3320	300	0.8499
0.8079	0.3542	320	0.8086
0.7969	0.3763	340	0.7712
0.7547	0.3985	360	0.7355
0.6896	0.4206	380	0.7046
0.7037	0.4427	400	0.6803
0.6724	0.4649	420	0.6614
0.6869	0.4870	440	0.6470
0.6412	0.5091	460	0.6390
0.6605	0.5313	480	0.6357
0.6293	0.5534	500	0.6352

Framework versions

• PEFT 0.13.2
• Transformers 4.45.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.20.1