lr_1e-04-last-sft/README.md

library_name, base_model, tags, model-index

library_name	base_model	tags	model-index
transformers	ErikDaska/lr_1e-04-last	generated_from_trainer	name results lr_1e-04-last-sft

lr_1e-04-last-sft

This model is a fine-tuned version of ErikDaska/lr_1e-04-last on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.7291

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: 5e-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 0.03
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss
0.9824	0.4444	200	0.8604
0.9089	0.8889	400	0.8122
0.8606	1.3333	600	0.7796
0.8538	1.7778	800	0.7631
0.8147	2.2222	1000	0.7530
0.7895	2.6667	1200	0.7410
0.7649	3.1111	1400	0.7358
0.7317	3.5556	1600	0.7330
0.7863	4.0	1800	0.7292
0.7629	4.4444	2000	0.7297
0.7229	4.8889	2200	0.7291
0.7518	5.0	2250	0.7291

Framework versions

• Transformers 5.10.2
• Pytorch 2.11.0+cu128
• Datasets 5.0.0
• Tokenizers 0.22.2