[BugFix][310p][Cherry-pick] Handle null quantization config in ShardedStateLoader310&[Feature][310P] Support W8A8 dynamic linear method (#8296)
### What this PR does / why we need it? This PR implements the `AscendW8A8DynamicLinearMethod310` quantization scheme specifically for 310P hardware. It includes the logic for weight retrieval, per-channel parameter generation, and the application of dynamic quantization using NPU-specific kernels. Additionally, it updates `ShardedStateLoader310` to handle quantization configurations more robustly when generating parameter type maps. Feedback from the review identified two critical issues in the implementation: 1. The tensor squeezing logic in the `apply` method incorrectly handles 2D inputs, which may lead to shape mismatches in subsequent layers. 2. The weight tensor in `process_weights_after_loading` is transposed after being converted to the private NZ format; the transpose operation should be performed on the ND tensor before conversion to ensure correct physical layout. cherry-pick from : #7546 #7725 ### Does this PR introduce _any_ user-facing change? No. ### How was this patch tested? New unit tests were added in `tests/ut/_310p/quantization/test_w8a8_dynamic_310.py` to verify the quantization method, and `tests/ut/_310p/test_sharded_state_loader_310p.py` was updated to test the state loader changes. --------- Signed-off-by: csoulnd <daidaicurry@foxmail.com>
This commit is contained in:
@@ -48,7 +48,11 @@ class NPUWorker310(NPUWorker):
|
||||
max_size=max_size,
|
||||
)
|
||||
|
||||
ShardedStateLoader310.generate_quant_description(self.model_runner.model, path)
|
||||
ShardedStateLoader310.generate_quant_description(
|
||||
self.model_runner.model,
|
||||
path,
|
||||
self.vllm_config.quant_config,
|
||||
)
|
||||
|
||||
@torch.inference_mode()
|
||||
def determine_available_memory(self) -> int:
|
||||
|
||||
Reference in New Issue
Block a user