Facing "NotImplementedError: config.do_stable_layer_norm is False is currently not supported" in FlaxWav2vec2Models

[premmurugan229]

System Info

While running Flax Wav2Vec2 models like FlaxWav2Vec2Model, FlaxWav2Vec2ForCTC, FlaxWav2Vec2ForPreTraining , I am encountering an error as

NotImplementedError: ``config.do_stable_layer_norm is False`` is currently not supported"

when config.do_stable_layer_norm is False, currently there is no implementation

but, by default config.do_stable_layer_norm is False Reference

We need to implement this, since mostly do_stable_layer_norm value is false. We might get this error frequently

System Info

My environment details are as below :

• transformers version: 4.51.3
• Platform: Linux-5.4.0-212-generic-x86_64-with-glibc2.35
• Python version: 3.10.12
• Huggingface_hub version: 0.30.2
• Safetensors version: 0.5.3
• Accelerate version: not installed
• Accelerate config: not found
• DeepSpeed version: not installed
• PyTorch version (GPU?): 2.6.0+cu124 (False)
• Tensorflow version (GPU?): not installed (NA)
• Flax version (CPU?/GPU?/TPU?): 0.10.4 (cpu)
• Jax version: 0.5.0
• JaxLib version: 0.5.0

I am attaching screenshot for reference

Who can help?

@gante @Rocketknight1

I am facing issue for Flax models like

FlaxWav2Vec2Model
FlaxWav2Vec2ForCTC
FlaxWav2Vec2ForPreTraining

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

Steps to recreate the error:

Run the below code in any python editor

from transformers import AutoProcessor, FlaxWav2Vec2Model
 
processor = AutoProcessor.from_pretrained("facebook/wav2vec2-base-960h")
model = FlaxWav2Vec2Model.from_pretrained("facebook/wav2vec2-base-960h", from_pt= True)

Expected behavior

The model should run and produce an error free output with the basic examples

[ctr-pmuruganTT]

I would like to contribute for this issue