Revision 38fab0bd87371924be7fe6c1efae3121491376b7 authored by Rui Hou on 09 May 2021, 00:44:49 UTC, committed by Facebook GitHub Bot on 09 May 2021, 00:45:41 UTC
Summary: Pull Request resolved: https://github.com/fairinternal/mmf-internal/pull/172 Pull Request resolved: https://github.com/facebookresearch/mmf/pull/874 Make it possible to have the mlm head weights tied to text_encoder embedding weights in case we have a separate text encoder for encoding text modalities. Reviewed By: apsdehal Differential Revision: D27756957 fbshipit-source-id: ce26e79ca9930127f36c3c2adfe280e3c67ac23a
1 parent dbe8dec
requirements.txt
torch>=1.6.0, <=1.8.1
torchvision>=0.7.0, <=0.9.1
numpy>=1.16.6
tqdm>=4.43.0,<4.50.0
demjson==2.2.4
torchtext==0.5.0
GitPython==3.1.0
requests==2.23.0
fasttext==0.9.1
nltk==3.4.5
editdistance==0.5.3
transformers==3.4.0
sklearn==0.0
omegaconf==2.0.6
lmdb==0.98
termcolor==1.1.0
iopath==0.1.7
datasets==1.2.1
matplotlib==3.3.4
pycocotools==2.0.2
ftfy==5.8
pytorch-lightning @ git+https://github.com/PyTorchLightning/pytorch-lightning@7a48db5
torchaudio>=0.6.0, <=0.8.1
Computing file changes ...
