ColossalAI/applications/ColossalChat/examples/inference/chatio.py
YeAnbang df5e9c53cf
[ColossalChat] Update RLHF V2 (#5286)
* Add dpo. Fix sft, ppo, lora. Refactor all

* fix and tested ppo

* 2 nd round refactor

* add ci tests

* fix ci

* fix ci

* fix readme, style

* fix readme style

* fix style, fix benchmark

* reproduce benchmark result, remove useless files

* rename to ColossalChat

* use new image

* fix ci workflow

* fix ci

* use local model/tokenizer for ci tests

* fix ci

* fix ci

* fix ci

* fix ci timeout

* fix rm progress bar. fix ci timeout

* fix ci

* fix ci typo

* remove 3d plugin from ci temporary

* test environment

* cannot save optimizer

* support chat template

* fix readme

* fix path

* test ci locally

* restore build_or_pr

* fix ci data path

* fix benchmark

* fix ci, move ci tests to 3080, disable fast tokenizer

* move ci to 85

* support flash attention 2

* add all-in-one data preparation script. Fix colossal-llama2-chat chat template

* add hardware requirements

* move ci test data

* fix save_model, add unwrap

* fix missing bos

* fix missing bos; support grad accumulation with gemini

* fix ci

* fix ci

* fix ci

* fix llama2 chat template config

* debug sft

* debug sft

* fix colossalai version requirement

* fix ci

* add sanity check to prevent NaN loss

* fix requirements

* add dummy data generation script

* add dummy data generation script

* add dummy data generation script

* add dummy data generation script

* update readme

* update readme

* update readme and ignore

* fix logger bug

* support parallel_output

* modify data preparation logic

* fix tokenization

* update lr

* fix inference

* run pre-commit

---------

Co-authored-by: Tong Li <tong.li352711588@gmail.com>
2024-03-29 14:12:29 +08:00

169 lines
6.3 KiB
Python
Executable File

"""
command line IO utils for chatbot
"""
import abc
import re
from prompt_toolkit import PromptSession
from prompt_toolkit.auto_suggest import AutoSuggestFromHistory
from prompt_toolkit.completion import WordCompleter
from prompt_toolkit.history import InMemoryHistory
from rich.console import Console
from rich.live import Live
from rich.markdown import Markdown
class ChatIO(abc.ABC):
@abc.abstractmethod
def prompt_for_input(self, role: str) -> str:
"""Prompt for input from a role."""
@abc.abstractmethod
def prompt_for_output(self, role: str):
"""Prompt for output from a role."""
@abc.abstractmethod
def stream_output(self, output_stream):
"""Stream output."""
class SimpleChatIO(ChatIO):
def prompt_for_input(self, role) -> str:
return input(f"{role}: ")
def prompt_for_output(self, role: str):
print(f"{role}: ", end="", flush=True)
def stream_output(self, output_stream):
pre = 0
for outputs in output_stream:
outputs = outputs.strip()
outputs = outputs.split(" ")
now = len(outputs) - 1
if now > pre:
print(" ".join(outputs[pre:now]), end=" ", flush=True)
pre = now
print(" ".join(outputs[pre:]), flush=True)
return " ".join(outputs)
class RichChatIO(ChatIO):
def __init__(self):
self._prompt_session = PromptSession(history=InMemoryHistory())
self._completer = WordCompleter(words=["!exit", "!reset"], pattern=re.compile("$"))
self._console = Console()
def prompt_for_input(self, role) -> str:
self._console.print(f"[bold]{role}:")
prompt_input = self._prompt_session.prompt(
completer=self._completer,
multiline=False,
auto_suggest=AutoSuggestFromHistory(),
key_bindings=None,
)
self._console.print()
return prompt_input
def prompt_for_output(self, role: str) -> str:
self._console.print(f"[bold]{role}:")
def stream_output(self, output_stream):
"""Stream output from a role."""
# Create a Live context for updating the console output
with Live(console=self._console, refresh_per_second=60) as live:
# Read lines from the stream
for outputs in output_stream:
accumulated_text = outputs
if not accumulated_text:
continue
# Render the accumulated text as Markdown
# NOTE: this is a workaround for the rendering "unstandard markdown"
# in rich. The chatbots output treat "\n" as a new line for
# better compatibility with real-world text. However, rendering
# in markdown would break the format. It is because standard markdown
# treat a single "\n" in normal text as a space.
# Our workaround is adding two spaces at the end of each line.
# This is not a perfect solution, as it would
# introduce trailing spaces (only) in code block, but it works well
# especially for console output, because in general the console does not
# care about trailing spaces.
lines = []
for line in accumulated_text.splitlines():
lines.append(line)
if line.startswith("```"):
# Code block marker - do not add trailing spaces, as it would
# break the syntax highlighting
lines.append("\n")
else:
lines.append(" \n")
markdown = Markdown("".join(lines))
# Update the Live console output
live.update(markdown)
self._console.print()
return outputs
class DummyChatIO(ChatIO):
"""
Dummy ChatIO class for testing
"""
def __init__(self):
self.roles = []
self._console = Console()
def prompt_for_input(self, role) -> str:
self.roles.append(role)
if len(self.roles) == 1:
ret = "Hello"
elif len(self.roles) == 2:
ret = "What's the value of 1+1?"
else:
ret = "exit"
self._console.print(f"[bold]{role}:{ret}")
return ret
def prompt_for_output(self, role: str) -> str:
self._console.print(f"[bold]{role}:")
def stream_output(self, output_stream):
"""Stream output from a role."""
# Create a Live context for updating the console output
with Live(console=self._console, refresh_per_second=60) as live:
# Read lines from the stream
for outputs in output_stream:
accumulated_text = outputs
if not accumulated_text:
continue
# Render the accumulated text as Markdown
# NOTE: this is a workaround for the rendering "unstandard markdown"
# in rich. The chatbots output treat "\n" as a new line for
# better compatibility with real-world text. However, rendering
# in markdown would break the format. It is because standard markdown
# treat a single "\n" in normal text as a space.
# Our workaround is adding two spaces at the end of each line.
# This is not a perfect solution, as it would
# introduce trailing spaces (only) in code block, but it works well
# especially for console output, because in general the console does not
# care about trailing spaces.
lines = []
for line in accumulated_text.splitlines():
lines.append(line)
if line.startswith("```"):
# Code block marker - do not add trailing spaces, as it would
# break the syntax highlighting
lines.append("\n")
else:
lines.append(" \n")
markdown = Markdown("".join(lines))
# Update the Live console output
live.update(markdown)
self._console.print()
return outputs
simple_io = SimpleChatIO()
rich_io = RichChatIO()
dummy_io = DummyChatIO()