Natural language processing

Environment info

• transformers version: 4.11.2
• Platform: Linux-5.4.0-66-generic-x86_64-with-glibc2.10
• Python version: 3.8.8
• PyTorch version (GPU?): 1.9.1+cu102 (True)
• Tensorflow version (GPU?): not installed (NA)
• F

Describe the bug
I'm having major trouble with from_csv.

Context: I'm writing tutorial for build simple text search engine with Jina + Hub. I don't want to include a whole section of processing datasets, hence just passing a CSV into from_csv. I tried with meme dataset (converted tsv) before, and now using [superhero dataset](https://www.kaggle.com/jonathanbesomi/superheroes-nlp-datas

Is your feature request related to a problem? Please describe.
I typically used compressed datasets (e.g. gzipped) to save disk space. This works fine with AllenNLP during training because I can write my dataset reader to load the compressed data. However, the predict command opens the file and reads lines for the Predictor. This fails when it tries to load data from my compressed files.

Hello everyone,

I need to compute the BLEU score with more than one ngram length (ideally, BLEU2, BLEU3, BLEU4, and BLEU5). In my case, this is a very long task, as every hypothesis has some thousand references.

Reading the implementation of the corpus_bleu function, which takes weights:Tuple between its parameters - and thus calculating BLEU-[len(weights)] - , I found out that it gets all t