Skip to content

Commit

Permalink
Remove pinning protobuf version since 3.20.3 is old and not supported…
Browse files Browse the repository at this point in the history
… by Tensorflow

Install `tf_nightly[and_cuda]` for Keras GPU builds.

PiperOrigin-RevId: 728362609
  • Loading branch information
tensorflower-gardener committed Feb 18, 2025
1 parent eaa4052 commit c5c04b0
Show file tree
Hide file tree
Showing 2 changed files with 3 additions and 3 deletions.
4 changes: 2 additions & 2 deletions requirements.txt
Original file line number Diff line number Diff line change
Expand Up @@ -8,8 +8,8 @@ scipy
# with both new and old protobuf stubs. This is needed to resolve
# Keras-TensorFlow circular dependency issue, when one of them gets a dependency
# incompatible with another one (protobuf in this specific case).
protobuf==3.20.3
tf-nightly
protobuf==4.21.6
tf-nightly[and_cuda]
portpicker
pyyaml
Pillow
Expand Down
2 changes: 1 addition & 1 deletion tf_keras/kokoro/github/ubuntu/gpu/build.sh
Original file line number Diff line number Diff line change
Expand Up @@ -42,7 +42,7 @@ pip uninstall -y keras-nightly

# LD Library Path needs to be same as TensorFlow Ubuntu Docker build -
# https://github.com/tensorflow/tensorflow/blob/master/tensorflow/tools/tf_sig_build_dockerfiles/
export LD_LIBRARY_PATH="/usr/local/cuda/lib64:/usr/local/cuda/extras/CUPTI/lib64:/usr/local/tensorrt/lib"
# export LD_LIBRARY_PATH="/usr/local/cuda/lib64:/usr/local/cuda/extras/CUPTI/lib64:/usr/local/tensorrt/lib"
CUDA_TOOLKIT_PATH="/usr/local/cuda-11.8"
TF_CUDA_CONFIG_REPO="@ubuntu20.04-gcc9_manylinux2014-cuda11.8-cudnn8.6-tensorrt8.4_config_cuda"
TF_CUDA_COMPUTE_CAPABILITIES="sm_35,sm_50,sm_60,sm_70,sm_75,compute_80"
Expand Down

0 comments on commit c5c04b0

Please sign in to comment.