Codec downstream task support: TTS

[jctian98]

What?

Support downstream task of codec project, specifically TTS.
Model architecture: Valle (https://arxiv.org/abs/2301.02111)

[mergify]

This pull request is now in conflict :(

[ftshijt]

Thanks for the great work! I've noted some comments in the PR. Seems that for current stage, the most important thing is to decide how the json format for multi-modal data would be organized. Could you give us some examples and design concept?

[ftshijt]

Do we expect users to add their own tokens?

[jctian98]

@ftshijt Thanks for the review. The code is very temporary at this moment. I would respond and clarify many concerns after the framework is complete.

Currently I only respond a small part of the comments :)

[ftshijt]

Many thanks for the implementation, I left a few minor comments. After your fix, we can merge the current effort~

[ftshijt]

Suggested change

	import logging
	import os
	import sys
	import json
	logging.basicConfig(
	format="%(asctime)s | %(levelname)s | %(name)s | %(message)s",
	datefmt="%Y-%m-%d %H:%M:%S",
	level=os.environ.get("LOGLEVEL", "INFO").upper(),
	stream=sys.stdout,
	)
	logger = logging.getLogger("find all example list based on train_jsons")
	import os
	import sys
	import json