mistral_instruct_generation

This model is a fine-tuned version of mistralai/Mistral-7B-Instruct-v0.1 on the generator dataset. It achieves the following results on the evaluation set:

• Loss: 0.8488

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: 0.0002
• train_batch_size: 4
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: constant
• lr_scheduler_warmup_steps: 0.03
• training_steps: 500

Training results

Training Loss	Epoch	Step	Validation Loss
1.1973	0.0305	20	1.1052
1.0347	0.0610	40	0.9958
0.9213	0.0915	60	0.9600
0.8886	0.1220	80	0.9406
0.9314	0.1524	100	0.9281
0.9668	0.1829	120	0.9197
0.887	0.2134	140	0.9128
0.8727	0.2439	160	0.9066
0.8571	0.2744	180	0.9005
0.8833	0.3049	200	0.8963
0.8466	0.3354	220	0.8912
0.9015	0.3659	240	0.8865
0.8602	0.3963	260	0.8822
0.8989	0.4268	280	0.8788
0.8452	0.4573	300	0.8758
0.8764	0.4878	320	0.8730
0.8702	0.5183	340	0.8708
0.8758	0.5488	360	0.8676
0.8071	0.5793	380	0.8638
0.8473	0.6098	400	0.8618
0.8822	0.6402	420	0.8586
0.8742	0.6707	440	0.8560
0.8526	0.7012	460	0.8533
0.8116	0.7317	480	0.8511
0.8593	0.7622	500	0.8488

Framework versions

• PEFT 0.10.0
• Transformers 4.40.1
• Pytorch 2.3.0+cu121
• Datasets 2.19.0
• Tokenizers 0.19.1