Skip to content

Releases: Syraxius/TensorflowDockerBuilder

TensorFlow 2.21.0 for RTX 50 / Blackwell sm_120 GPUs

02 May 00:37

Choose a tag to compare

Unofficial TensorFlow 2.21.0 Linux x86_64 wheels built for NVIDIA Blackwell RTX GPUs with CUDA compute capability 12.0.

These wheels include native sm_120 CUDA kernels plus compute_120 PTX fallback, targeting RTX 50-series and RTX PRO Blackwell GPUs such as the GeForce RTX 5090.

Artifacts

  • tensorflow-2.21.0+selfbuild-cp310-cp310-linux_x86_64.whl
  • tensorflow-2.21.0+selfbuild-cp311-cp311-linux_x86_64.whl
  • tensorflow-2.21.0+selfbuild-cp312-cp312-linux_x86_64.whl
  • tensorflow-2.21.0+selfbuild-cp313-cp313-linux_x86_64.whl

Build Details

  • TensorFlow: v2.21.0
  • CUDA/cuDNN: CUDA 12.8.1, cuDNN 9.8.0
  • CUDA architecture flags: sm_120,compute_120
  • Bazel CUDA config: cuda_nvcc
  • Platform: Linux x86_64
  • Python: 3.10, 3.11, 3.12, 3.13

Install

python -m pip install --upgrade pip
python -m pip install './tensorflow-2.21.0+selfbuild-cp312-cp312-linux_x86_64.whl[and-cuda]'

Use the wheel matching your Python version.

Verify

python - <<'PY'
import json
import tensorflow as tf

print(tf.__version__)
print(json.dumps(tf.sysconfig.get_build_info(), indent=2, sort_keys=True))
print(tf.config.list_physical_devices("GPU"))
PY

Expected highlights:

  • TensorFlow version: 2.21.0+selfbuild
  • CUDA compute capabilities include sm_120 and compute_120
  • Your Blackwell RTX GPU appears in TensorFlow's GPU device list

Notes

This is an unofficial downstream TensorFlow build, not an official TensorFlow release.
A recent NVIDIA 570-series or newer driver is recommended for CUDA 12.8 / RTX 50-series systems.

Based on NVIDIA’s CUDA GPU compute capability table, sm_120 / compute capability 12.0 covers these names:

  • GeForce RTX 5090
  • GeForce RTX 5080
  • GeForce RTX 5070 Ti
  • GeForce RTX 5070
  • GeForce RTX 5060 Ti
  • GeForce RTX 5060
  • GeForce RTX 5050
  • NVIDIA RTX PRO 6000 Blackwell Server Edition
  • NVIDIA RTX PRO 4500 Blackwell Server Edition
  • NVIDIA RTX PRO 6000 Blackwell Workstation Edition
  • NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition
  • NVIDIA RTX PRO 5000 Blackwell
  • NVIDIA RTX PRO 4500 Blackwell
  • NVIDIA RTX PRO 4000 Blackwell
  • NVIDIA RTX PRO 4000 Blackwell SFF Edition
  • NVIDIA RTX PRO 2000 Blackwell