Delete deprecated QDQLayout and related code

[jerryzh168]

Stack from ghstack (oldest at bottom):

• Delete deprecated TensorCoreTiledLayout and related code #4153
• Move bitpacking.py to prototype and add uintx_utils.py #4152
• Delete deprecated PlainLayout, PlainAQTTensorImpl and related v1 code paths #4151
• Delete deprecated PackedLinearInt8DynamicActivationIntxWeightLayout and related code #4150
• Delete deprecated Int4CPULayout and related code #4149
• Delete deprecated SemiSparseLayout and related code #4148
• Delete deprecated Int4XPULayout and related code #4147
• -> Delete deprecated QDQLayout and related code #4146

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/4146

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[jerryzh168]

@claude review

[claude]

Claude finished @jerryzh168's task in 1m 29s —— View job

---

Review: Delete deprecated QDQLayout and related code

• Gather context and read changed files
• Review the diff for correctness and completeness
• Post review feedback

---

This is a clean deletion of deprecated code. The changes are straightforward and correct.

Summary of changes:

• Deletes torchao/dtypes/uintx/q_dq_layout.py (253 lines) — the entire QDQLayout, QDQTensorImpl, and associated linear/embedding dispatch functions
• Removes QDQLayout from the uintx and dtypes __init__.py exports
• Removes the _linear_q_dq_check/_linear_q_dq_impl dispatch entry from _register_aqt_quantized_linear_dispatches() and the _embedding_q_dq_check/_embedding_q_dq_impl call from the embedding dispatch in affine_quantized_tensor_ops.py
• Updates StretchedIntxWeightConfig.layout default from QDQLayout() to None in config_torchao.py

Observations:

1. No remaining references — I verified with a full codebase grep for QDQLayout, QDQTensorImpl, and q_dq_layout. Zero matches remain, so this is a complete removal.

2. StretchedIntxWeightConfig.layout change is safe — The layout field was changed from QDQLayout() to Optional[Layout] = None. The handler function _int8_dynamic_activation_stretched_intx_transform (config_torchao.py:51-109) doesn't reference config.layout at all (it uses IntxUnpackedToInt8Tensor directly for version 2), so this default value change has no functional impact.

3. Embedding dispatch fallback — Previously, the embedding __torch_function__ handler checked for QDQLayout first and dequantized if matched. With this removed, any model that was using QDQLayout for embeddings would now fall through to the PlainAQTTensorImpl assertion at affine_quantized_tensor_ops.py:212. This is fine since QDQLayout was already deprecated and emitting warnings, and this is a deliberate removal.

4. Linear dispatch removal — The QDQ linear dispatch entry was a simple dequantize-then-call-F.linear implementation. Its removal from the dispatch table is clean — the surrounding entries are unaffected.

Verdict: LGTM. Clean and complete deletion of deprecated code with no dangling references.