Three undefined names found by static analysis (one reachable NameError)

[musicalplatypus]

Running pyflakes across the repository surfaces three undefined names on main (378dd15). They are unrelated to each other and differ in severity, so I have listed them separately. Happy to send a PR for any or all of them if that is useful.

Reproduce with:

pip install pyflakes
pyflakes $(git ls-files '*.py')

---

1. T is undefined in the audio keyword-spotting example — reachable

tinyml-modeloptimization/torchmodelopt/examples/audio_keyword_spotting/main.py:243

def compute_torchMFCC(self, audio_tensor):
    __MFCC = T.MFCC( # type: ignore

T is never imported in the file. The method is called at line 203, so this raises NameError: name 'T' is not defined whenever that path executes. The # type: ignore suggests a type checker flagged it previously.

Looks like a missing import torchaudio.transforms as T, but I did not want to guess the intended module.

---

2. __all__ lists a name that does not exist

tinyml-modelzoo/tinyml_modelzoo/models/classification.py:985

    'TEST_NEW_MODEL_2K',

TEST_NEW_MODEL_2K is not defined anywhere in the file. Python raises AttributeError for a from ... import * when __all__ names a missing symbol, so this breaks star-imports of the module. The name suggests leftover scaffolding.

---

3. value is undefined in image_base.py — latent, currently unreachable

tinyml-modelmaker/tinyml_modelmaker/ai_modules/vision/training/tinyml_tinyverse/image_base.py:242-245

    @classmethod
    def init_params(cls, *args, **kwargs):
        params = dict(training=dict())
        params = utils.ConfigDict(params, *args, **kwargs)
        return params

    # @staticmethod
    # def _argv_value(value):
    #     if value is None:
    #         return "None"

        if isinstance(value, (list, tuple)):
            return repr(list(value))

        return str(value)

A _argv_value static method was partially commented out: the decorator, def line and first branch are commented, but the rest of the body is not. Python therefore parses the remaining lines as trailing statements of init_params, after its return, so they are unreachable and cannot raise at runtime today.

Not currently a bug, but the dead fragment is misleading, and it would start executing if init_params were ever edited above it. _argv_value has no callers, so deleting the fragment outright looks correct.

---

For completeness: a separate missing import numpy as np in three references/*/test_onnx.py scripts produces the same class of error on the nn_for_feature_extraction path. That one is already included in PR #8.