llama3-samsum

This model is a fine-tuned version of meta-llama/Meta-Llama-3-8B on the Samsumg/samsum dataset.

Model description

It is a first version and has to be improved. The challenge is to fine-tune the model using limited resources. The fine tuning was performed downsampling the dataset, under Colab free plan restrictions.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0002
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 2
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.03
• num_epochs: 1
• mixed_precision_training: Native AMP

Training results

Framework versions

• PEFT 0.12.0
• Transformers 4.43.2
• Pytorch 2.4.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1