Trainer's step loss always drops sharply after each epoch regardless of model / data - 🤗Transformers - Hugging Face Forums

Trainer's step loss always drops sharply after each epoch regardless of model / data - 🤗Transformers - Hugging Face Forums

Hello, This is a specific question on the behaviour of the Trainer class, I hope someone is able to help. I am trying to plot the training loss after each step and therefore I have subclassed Trainer and made a couple of edits at line 1772 (new lines commented with # ADDED LINE) so the if statement now reads step_losses = [] # ADDED LINE if ( ((step + 1) % args.gradient_accumulation_steps != 0) and args.local_rank != -1 and args._no_sync_in_gradient_accumulation ): # Avoid u

Optimized Training and Inference of Hugging Face Models on Azure Databricks – Part 1 - Microsoft Community Hub

deep learning - Why does my train loss jump down when a new epoch starts? - Stack Overflow

trainer] fractional epoch · Issue #9475 · huggingface/transformers · GitHub

Trainer API to log both Training and Validation Metrics - 🤗Transformers - Hugging Face Forums

Size of saved model checkpoints after trainer.train() is much larger when using trainer with deepspeed stage2 · Issue #22822 · huggingface/transformers · GitHub

evaluate_during_training after each epoch · Issue #7134 · huggingface/transformers · GitHub

Huge Num Epochs (9223372036854775807) when using Trainer API with streaming dataset · Issue #22757 · huggingface/transformers · GitHub

deep learning - Why does my train loss jump down when a new epoch starts? - Stack Overflow

Divide HuggingFace training time by 2