[Bug fix] Fake quantized model save after HF accelerate hooks are added (#906)

## What does this PR do?

**Type of change:** Bug fix

**Overview:** Fix `AttributeError: Can't get local object
'add_hook_to_module.<locals>.new_forward'` when saving a quantized model
a second time after restoring it with `device_map="auto"`.

When a model is loaded with `device_map="auto"`, accelerate's
`add_hook_to_module` patches every submodule (including
`TensorQuantizer` instances) and injects three instance attributes:
`_hf_hook`, `_old_forward`, and `forward` (a `functools.partial`
wrapping a local function). These are not picklable and were leaking
into the modelopt state dict collected by `get_modelopt_state()`,
causing `torch.save` to fail.

This PR adds the three accelerate-injected attributes to
`TensorQuantizer._skip_properties_for_save_restore` so they are excluded
from the serialized state, matching the existing pattern used for
HuggingFace and DeepSpeed attributes.

## Usage

```python
mto.enable_huggingface_checkpointing()

# Quantize and save
model = AutoModelForCausalLM.from_pretrained(name, device_map="auto")
model = mtq.quantize(model, mtq.FP8_DEFAULT_CFG, forward_loop=forward_loop)
model.save_pretrained(save_dir)

# Restore and save again (this previously failed)
model2 = AutoModelForCausalLM.from_pretrained(save_dir, device_map="auto")
model2.save_pretrained(save_dir_round2)  # now works
```

## Testing

- Added unit test
`test_tensor_quantizer_modelopt_state_with_accelerate_hook` in
`tests/unit/torch/quantization/plugins/test_accelerate.py` that verifies
accelerate hook attributes are excluded from modelopt state and the
state dict remains picklable.

## Before your PR is "*Ready for review*"

- **Make sure you read and follow [Contributor
guidelines](https://github.com/NVIDIA/Model-Optimizer/blob/main/CONTRIBUTING.md)**
and your commits are signed.
- **Is this change backward compatible?**: Yes — only adds entries to a
skip set; existing saved checkpoints are unaffected.
- **Did you write any new necessary tests?**: Yes
- **Did you add or update any necessary documentation?**: No (internal
fix, no API change)
- **Did you update
[Changelog](https://github.com/NVIDIA/Model-Optimizer/blob/main/CHANGELOG.rst)?**:
No

## Additional Information

The root cause is in accelerate's `add_hook_to_module`, which defines
`new_forward` as a local function and binds it via `functools.partial`
onto `module.forward`. Since local functions cannot be pickled, any
`TensorQuantizer` that has been hooked by accelerate becomes
unserializable unless these attributes are excluded.


<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->

## Summary by CodeRabbit

* **Bug Fixes**
* Enhanced compatibility with accelerate library by excluding
framework-specific hooks and attributes from model state serialization,
preventing issues during save/restore operations.

* **Tests**
* Added test to validate that accelerate-related attributes are properly
excluded from model state and that the state remains picklable.

* **Public API**
  * TensorQuantizer is now publicly exported.

<!-- end of auto-generated comment: release notes by coderabbit.ai -->

Signed-off-by: realAsma <akuriparambi@nvidia.com>

modelopt/torch/quantization/nn/modules/tensor_quantizer.py

@@ -156,6 +156,10 @@ class TensorQuantizer(nn.Module):
         "_padding",
         # Extra flags added by huggingface
         "_is_hf_initialized",
+        # Extra flags added by accelerate
+        "_hf_hook",
+        "_old_forward",
+        "forward",
         # Extra flags added by deepspeed
         "ds_external_parameters",
         "all_parameters",

tests/unit/torch/quantization/plugins/test_accelerate.py

@@ -13,12 +13,14 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+import pickle
+
 import pytest
 import torch
 import torch.nn as nn
 
 import modelopt.torch.quantization as mtq
-from modelopt.torch.quantization.nn import QuantLinearConvBase
+from modelopt.torch.quantization.nn import QuantLinearConvBase, TensorQuantizer
 
 try:
     from accelerate.hooks import ModelHook, add_hook_to_module
@@ -51,3 +53,30 @@ def test_linear_with_accelerate_monkey_patched_forward():
 
     assert module_test.input_quantizer.amax is not None
     assert module_test.weight_quantizer.amax is not None
+
+
+def test_tensor_quantizer_modelopt_state_with_accelerate_hook():
+    """Verify accelerate hook attributes are excluded from modelopt state.
+
+    When accelerate's add_hook_to_module patches a TensorQuantizer, it adds
+    _hf_hook, _old_forward, and an instance-level forward (a functools.partial
+    wrapping a local function). These must be excluded from the modelopt state
+    dict, otherwise torch.save / pickle will fail with:
+        AttributeError: Can't get local object 'add_hook_to_module.<locals>.new_forward'
+    """
+    tq = TensorQuantizer()
+    add_hook_to_module(tq, ModelHook())
+
+    # The hook should have injected these instance attributes
+    assert hasattr(tq, "_hf_hook")
+    assert hasattr(tq, "_old_forward")
+    assert "forward" in tq.__dict__
+
+    # None of the accelerate attributes should appear in the modelopt state
+    state = tq.get_modelopt_state()
+    accelerate_attrs = {"_hf_hook", "_old_forward", "forward"}
+    leaked = accelerate_attrs & state.keys()
+    assert not leaked, f"Accelerate attributes leaked into modelopt state: {leaked}"
+
+    # The state dict must be picklable (torch.save uses pickle internally)
+    pickle.dumps(state)