shawgpt-ft-epoch-19

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

• Loss: 2.4492

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.PAGED_ADAMW_8BIT with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 2
• num_epochs: 19
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
8.5018	0.5714	1	4.2401
8.5321	1.5714	2	4.1560
8.0764	2.5714	3	3.9643
7.7435	3.5714	4	3.7783
7.3166	4.5714	5	3.6064
7.0168	5.5714	6	3.4436
6.714	6.5714	7	3.2908
6.4229	7.5714	8	3.1510
6.183	8.5714	9	3.0261
5.9975	9.5714	10	2.9176
5.7077	10.5714	11	2.8235
5.5493	11.5714	12	2.7431
5.3984	12.5714	13	2.6731
5.2546	13.5714	14	2.6121
5.129	14.5714	15	2.5606
5.0369	15.5714	16	2.5182
4.9532	16.5714	17	2.4847
4.9298	17.5714	18	2.4614
2.4286	18.5714	19	2.4492

Framework versions

• PEFT 0.14.0
• Transformers 4.48.3
• Pytorch 2.5.1+cu124
• Datasets 3.3.1
• Tokenizers 0.21.0