myBit-Llama2-jp-127M-2B4TLike-wiki-aozora-sort-swallow-tokenizer

This model is a fine-tuned version of on the None dataset. It achieves the following results on the evaluation set:

• Loss: 3.0804

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: 0.0024
• train_batch_size: 24
• eval_batch_size: 24
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 96
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.95) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 750
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss
5.0217	0.0437	500	5.5967
3.8655	0.0875	1000	4.9636
3.5798	0.1312	1500	4.5889
3.4331	0.1749	2000	4.3685
3.3371	0.2187	2500	4.2415
3.2912	0.2624	3000	4.1319
3.2255	0.3061	3500	4.0167
3.1863	0.3499	4000	3.9057
3.1549	0.3936	4500	3.7257
3.1237	0.4373	5000	3.6422
3.1066	0.4811	5500	3.5338
3.0786	0.5248	6000	3.4784
3.051	0.5686	6500	3.4182
3.0279	0.6123	7000	3.3684
3.0516	0.6560	7500	3.3032
3.1796	0.6998	8000	3.2376
3.1997	0.7435	8500	3.2020
3.172	0.7872	9000	3.1750
3.1487	0.8310	9500	3.1502
3.1296	0.8747	10000	3.1273
3.0886	0.9184	10500	3.1030
3.0402	0.9622	11000	3.0804

Framework versions

• Transformers 4.47.1
• Pytorch 2.6.0+cu124
• Datasets 3.5.1
• Tokenizers 0.21.1