Update no_trainer scripts to include gather_for_metrics

Feature request

🤗 Accelerate has a wrapper to help with distributed metric calculation (a tough problem!), and the no_trainer scripts should be updated to include it!

An example can be seen here, below is an example diff of what the integration would look like:

-             predictions, references = accelerator.gather((predictions, batch["labels"]))
-             # If we are in a multiprocess environment, the last batch has duplicates
-             if accelerator.num_processes > 1:
-                 if step == len(eval_dataloader) - 1:
-                     predictions = predictions[: len(eval_dataloader.dataset) - samples_seen]
-                     references = references[: len(eval_dataloader.dataset) - samples_seen]
-                 else:
-                     samples_seen += references.shape[0]
+             predictions, references = accelerator.gather_for_metrics((predictions, batch["labels"])) 

The list of available scripts to update include:

• examples/pytorch/image-classification/run_image_classification_no_trainer.py
• examples/pytorch/language-modeling/run_clm_no_trainer.py
• examples/pytorch/language-modeling/run_mlm_no_trainer.py
• examples/pytorch/multiple-choice/run_swag_no_trainer.py
• examples/pytorch/question-answering/run_qa_beam_search_no_trainer.py
• examples/pytorch/question_answering/run_qa_no_trainer.py
• examples/pytorch/semantic-segmentation/run_semantic_segmentation_no_trainer.py
• examples/pytorch/speech-pretraining/run_wav2vec2_pretraining_no_trainer.py
• examples/pytorch/summarization/run_summarization_no_trainer.py

Motivation

This is a great first issue for someone who wants to learn how to use some of the latest bits in Accelerate and get an easy beginner contribution to the library 🤗

Your contribution

If you decide to pick up this issue, feel free to ping myself (@muellerzr), @sgugger, or @pacman100 to review 🤗

[nandwalritik]

Hi @muellerzr opened PR #18443 for first three examples in the list.

[kiansierra]

Hi @muellerzr I opened this PR https://github.com/huggingface/transformers/pull/18474 for examples 5,6 and 7.