Typo and GPU kernel check bug in MXFP4 quantization (v4.55.3)

[HeenaRajan]

System Info

• transformers version: 4.55.3
• Platform: Linux-5.10.0-35-cloud-amd64-x86_64-with-glibc2.31
• Python version: 3.10.18
• Huggingface_hub version: 0.34.4
• Safetensors version: 0.6.2
• Accelerate version: 1.10.0
• Accelerate config: not found
• DeepSpeed version: not installed
• PyTorch version (accelerator?): 2.8.0+cu128 (CUDA)
• Tensorflow version (GPU?): not installed (NA)
• Flax version (CPU?/GPU?/TPU?): not installed (NA)
• Jax version: not installed
• JaxLib version: not installed
• Using distributed or parallel set-up in script?:
• Using GPU in script?:
• GPU type: NVIDIA A100-SXM4-40GB

Who can help?

@SunMarc
@MekkCyber

Information

• The official example scripts
• My own modified scripts

Tasks

• An officially supported task in the examples folder (such as GLUE/SQuAD, ...)
• My own task or dataset (give details below)

Reproduction

In this script, line 72 has a typo in the function name:

if not is_triton_available("3.4.0") or not is_triton_kernels_availalble():

is_triton_kernels_availalble() is misspelled (extra "l"). The correct function is is_triton_kernels_available().

This causes a NameError when validating the environment for MXFP4 quantization.

Expected behavior

Suggested fix:
Replace line 72 with:

if not is_triton_available("3.4.0") or not is_triton_kernels_available():

[SIGSEGV-0x7]

Having the same issue here.

Replaced line 72 in file "src/transformers/quantizers/quantizer_mxfp4.py" with:

if not is_triton_available("3.4.0") or not is_kernels_available():

seems to resolve the issue temporarily.

[SunMarc]

Indeed there was an issue when releasing the patch today. We will fix this soon ! Please use main for now or 4.55.2 if you want to make it work with mxfp4. cc @ArthurZucker

[ArthurZucker]

Damn we have a small issue with the patch helper 😢

[vanshs1]

Having the same issue here.

Replaced line 72 in file "src/transformers/quantizers/quantizer_mxfp4.py" with:

if not is_triton_available("3.4.0") or not is_kernels_available():

seems to resolve the issue temporarily.

Yes - this fixes it! I dropped in to see if someone else had reported/fixed it. Looking forward to the patch.