Santacoder code-to-text

This model is a fine-tuned version of bigcode/santacoder on copdeparrot/gitub-jupyter-code-to-text.

Training procedure

The model was trained on 4 A100 for 3h with the following hyperparameters were used during training on 4 A100:

• learning_rate: 5e-05
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 4
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 100
• training_steps: 800