Docs: fix Training step by removing tokenizer from trainer initialization (#43733)
* Docs: fix Training step by removing tokenizer from trainer initialization
We already provide the tokenizer in the data_collator. Trying to provide the tokenizer parameter in the Trainer() __init__() method produces the following error: TypeError: Trainer.__init__() got an unexpected keyword argument 'tokenizer'
* fixup! Docs: fix Training step by removing tokenizer from trainer initialization