AttributeError: 'Gemma3TextConfig' object has no attribute 'rope_local_base_freq' when running pipeline

[lhphanto]

(ltx2) [../data/users/lxh/LTX-2 (main)]$ python -m ltx_pipelines.ti2vid_two_stages     --checkpoint-path ./ltx2_weights/ltx-2-19b-distilled-fp8.safetensors     --distilled-lora ./ltx2_weights/ltx-2-19b-distilled-lora-384.safetensors 0.8     --spatial-upsampler-path ./ltx2_weights/ltx-2-spatial-upscaler-x2-1.0.safetensors     --gemma-root ./gemma     --prompt "A beautiful sunset over the ocean"     --output-path output.mp4
<frozen runpy>:128: RuntimeWarning: 'ltx_pipelines.ti2vid_two_stages' found in sys.modules after import of package 'ltx_pipelines', but prior to execution of 'ltx_pipelines.ti2vid_two_stages'; this may result in unpredictable behaviour
Traceback (most recent call last):
  File "<frozen runpy>", line 198, in _run_module_as_main
  File "<frozen runpy>", line 88, in _run_code
  File "/data/users/lxh/LTX-2/packages/ltx-pipelines/src/ltx_pipelines/ti2vid_two_stages.py", line 299, in <module>
    main()
  File "/home/lxh/.conda/envs/ltx2/lib/python3.12/site-packages/torch/utils/_contextlib.py", line 124, in decorate_context
    return func(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^
  File "/data/users/lxh/LTX-2/packages/ltx-pipelines/src/ltx_pipelines/ti2vid_two_stages.py", line 259, in main
    video, audio = pipeline(
                   ^^^^^^^^^
  File "/home/lxh/.conda/envs/ltx2/lib/python3.12/site-packages/torch/utils/_contextlib.py", line 124, in decorate_context
    return func(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^
  File "/data/users/lxh/LTX-2/packages/ltx-pipelines/src/ltx_pipelines/ti2vid_two_stages.py", line 105, in __call__
    text_encoder = self.stage_1_model_ledger.text_encoder()
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/data/users/lxh/LTX-2/packages/ltx-pipelines/src/ltx_pipelines/utils/model_ledger.py", line 235, in text_encoder
    return self.text_encoder_builder.build(device=self._target_device(), dtype=self.dtype).to(self.device).eval()
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/data/users/lxh/LTX-2/packages/ltx-core/src/ltx_core/loader/single_gpu_model_builder.py", line 75, in build
    meta_model = self.meta_model(config, self.module_ops)
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/data/users/lxh/LTX-2/packages/ltx-core/src/ltx_core/loader/single_gpu_model_builder.py", line 51, in meta_model
    model = module_op.mutator(model)
            ^^^^^^^^^^^^^^^^^^^^^^^^
  File "/data/users/lxh/LTX-2/packages/ltx-core/src/ltx_core/text_encoders/gemma/encoders/av_encoder.py", line 135, in create_and_populate
    base = config.rope_local_base_freq
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/lxh/.conda/envs/ltx2/lib/python3.12/site-packages/transformers/configuration_utils.py", line 164, in __getattribute__
    return super().__getattribute__(key)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AttributeError: 'Gemma3TextConfig' object has no attribute 'rope_local_base_freq'

Has anyone encountered similar issues? I think the code is not compatible with the gemma I downloaded. But it seems like the gemma3 repo hasn't changed for the last 10 months?

Thanks!

[asafjo23]

Worked for me after installing "transformers==4.57.6"

[bjranson]

Had the same issue. LLM wanted me to do all sorts of code hacks, but in the end this worked:
pip install transformers==4.57.6