search

AlibabaResearch / DAMO-ConvAI

DAMO-ConvAI: The official repository which contains the codebase for Alibaba DAMO Conversational AI.
MIT License
1.21k stars 186 forks source link

Hi - Any update on code? Already April... #20

Closed Kshlsingh763 closed 1 year ago

huybery commented 1 year ago

Graphix will be released in next week, thanks for your attention !

Kshlsingh763 commented 1 year ago

Hi @huybery - Any tentative timeline on this?

kanseaveg commented 1 year ago

I've been waiting for two months already, and the official side still hasn't provided the code. The performance of din-sql is even better than this one.

Why haven't they made the code public after completing it? I don't quite understand. @huybery

huybery commented 1 year ago

Graphix has been updated, thanks for your patience.

@kanseaveg There is a lot of work to open source code, especially in companies, and there are issues of copyright and personnel leave involved, which we have been struggling with.

huybery commented 1 year ago

Also, welcome to check out our new benchmark: BIRD-SQL https://bird-bench.github.io/

kanseaveg commented 1 year ago

awesome! thanks for your team's contribution.

kanseaveg commented 1 year ago

by the way, is there any requirements.txt? Although the dockerfile is convenient but it's not convenient for development to the realistic environment. would you please summay the dependencies to an external requirements.txt? thanks.

kanseaveg commented 1 year ago

by the way, is there any requirements.txt? Although the dockerfile is convenient but it's not convenient for development to the realistic environment. would you please summay the dependencies to an external requirements.txt? thanks.

I make an mistake. I had saw thepyproject.tomlnow. No need an requirements.txt now. Thank you for your team's contribution again. image

kanseaveg commented 1 year ago

summay an requirements.txt. anyone need it can copy.

# Necessity dependencies
cython==0.29.28
six==1.16.0
pandas==1.3.5
numpy==1.21.5
# git config --system core.longpaths true | for windows to fix "filename too long" bug
datasets @ git+https://github.com/huggingface/datasets@cf47649eaed608fb7030f692020a0921e16f23c8
nltk==3.7
pyyaml==6.0
sentencepiece==0.1.96
tenacity==8.0.1
rapidfuzz==2.0.5
alive-progress==2.3.1
sqlparse==0.4.2
pillow==9.0.1
transformers==4.17.0
fastapi==0.74.1
uvicorn==0.17.5
torch==2.0.1

# need the cuda tookit version the same as the deepspeed version. If version dismatch, see https://github.com/microsoft/DeepSpeed/issues/2902#issuecomment-1530056532
# deepspeed==0.5.10

# For dev-dependencies
pytest==7.0.1
pytest-check==1.0.4
isort==5.10.1
autopep8==1.6.0
wandb==0.12.11
black==22.1.0
mypy==0.931
flake8==4.0.1