[Feature request] Change default text generation model from gpt2 to distilgpt2 about transformers.js

[xenova]

The only reason I use gpt2 as the default, is because HF uses it as a default:

https://github.com/huggingface/transformers/blob/5fd4e3c87c685fba2dd9615be62131748a8b5ee3/src/transformers/pipelines/__init__.py#LL280

I think default performance might improve once I add more generation parameters (no repeat n grams, etc.)

[xenova]

I think default performance might improve once I add more generation parameters (no repeat n grams, etc.)

We recently added repetition_penalty and no_repeat_ngram_size generation parameters by the way :)