seliny2/medusa-llama3.1-8b-sharegpt-stage2

work1/deming/shared/medusa-llama3.1-8b-axolotl-stage2

This model is a fine-tuned version of meta-llama/Llama-3.1-8B on the None dataset. It achieves the following results on the evaluation set:

• Loss: 2.8320

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: 2
• eval_batch_size: 2
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• total_train_batch_size: 16
• total_eval_batch_size: 16
• optimizer: Use adamw_bnb_8bit 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: 20
• num_epochs: 1
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
3.8262	0.1487	40	3.6953
3.3894	0.2974	80	3.3145
3.1755	0.4461	120	3.1328
3.0369	0.5948	160	2.9805
2.7292	0.7435	200	2.8809
2.6692	0.8922	240	2.8320

Framework versions

• PEFT 0.14.0
• Transformers 4.48.0
• Pytorch 2.5.1+rocm6.2
• Datasets 3.2.0
• Tokenizers 0.21.1