llama-3.2-350M-fourier_arithmetic_dataset

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

• Loss: 1.6047

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: 0.0005
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• num_devices: 2
• gradient_accumulation_steps: 16
• total_train_batch_size: 32
• total_eval_batch_size: 2
• optimizer: Use adamw_torch 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: 1000
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss
1.8493	0.1066	1000	1.8628
1.8654	0.2132	2000	1.8692
1.8328	0.3197	3000	1.8328
1.7287	0.4263	4000	1.7136
1.6856	0.5329	5000	1.6816
1.65	0.6395	6000	1.6494
1.6304	0.7460	7000	1.6308
1.6071	0.8526	8000	1.6119
1.6022	0.9592	9000	1.6047

Framework versions

• Transformers 4.48.2
• Pytorch 2.3.1+cu118
• Datasets 3.2.0
• Tokenizers 0.21.0