[BUG] Abnormal progress bar in dataset.map when load_from_cache_file=False

[Jintao-Huang]

Describe the bug

datasets 4.8.5

from datasets import load_dataset, load_from_disk

data = load_from_disk('pretrain_cached_dataset/train')

def dummy_map(example):
    return example

dataset = data.map(dummy_map)
dataset = data.map(dummy_map, load_from_cache_file=True)   # works fine
dataset = data.map(dummy_map, load_from_cache_file=True, num_proc=4)   # works fine
dataset = data.map(dummy_map, load_from_cache_file=False)  # does not work

Steps to reproduce the bug

Expected behavior

Environment info

[Jintao-Huang]

datasets/src/datasets/arrow_dataset.py

Lines 3593 to 3598 in 4e3bf52

	pbar_total = len(self)
	pbar_initial = len(existing_cache_files) * pbar_total // num_shards
	if batched and drop_last_batch:
	batch_size = batch_size or 1
	pbar_initial = pbar_initial // num_shards // batch_size * num_shards * batch_size
	pbar_total = pbar_total // num_shards // batch_size * num_shards * batch_size

[lhoestq]

Hi, can you provide more details on the issue ?

[Jintao-Huang]

This exceeds the maximum value.

[Nitin-Rajasekar]

#self-assign

[Nitin-Rajasekar]

Hey, I've opened a PR for this: #8170