5899438a86
### What this PR does / why we need it?
This pull request introduces significant enhancements for 310P device
support, primarily by enabling W8A8S quantization and facilitating the
saving of models with W8A8SC state outputs. It provides an example
script for saving sharded and compressed model states, implements the
core W8A8S quantization method, and integrates metadata generation
within the 310P worker to accurately describe the quantization types of
saved parameters. These changes aim to improve efficiency and
compatibility for quantized models on 310P hardware.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
W8A8S accuarcy test and W8A8SC states save.
<img width="886" height="184" alt="image"
src="https://github.com/user-attachments/assets/e9bcac54-1f69-4d3a-a5b8-221a147ef99d"
/>
- vLLM version: v0.16.0
- vLLM main:
15d76f74e2
---------
Signed-off-by: pu-zhe <zpuaa@outlook.com>
53 lines
1.8 KiB
Python
53 lines
1.8 KiB
Python
#
|
|
# Copyright (c) 2025 Huawei Technologies Co., Ltd. All Rights Reserved.
|
|
#
|
|
# Licensed under the Apache License, Version 2.0 (the "License");
|
|
# you may not use this file except in compliance with the License.
|
|
# You may obtain a copy of the License at
|
|
#
|
|
# http://www.apache.org/licenses/LICENSE-2.0
|
|
#
|
|
# Unless required by applicable law or agreed to in writing, software
|
|
# distributed under the License is distributed on an "AS IS" BASIS,
|
|
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
# See the License for the specific language governing permissions and
|
|
# limitations under the License.
|
|
# This file is a part of the vllm-ascend project.
|
|
#
|
|
import torch_npu
|
|
from vllm.logger import logger
|
|
|
|
from vllm_ascend._310p.model_runner_310p import NPUModelRunner310
|
|
from vllm_ascend.worker.worker import NPUWorker, init_workspace_manager
|
|
|
|
|
|
class NPUWorker310(NPUWorker):
|
|
def init_device(self):
|
|
self.device = self._init_device()
|
|
torch_npu.npu.set_compile_mode(jit_compile=False)
|
|
|
|
init_workspace_manager(self.device, num_ubatches=1)
|
|
|
|
self.model_runner = NPUModelRunner310(self.vllm_config, self.device)
|
|
|
|
def save_sharded_state(
|
|
self,
|
|
path: str,
|
|
pattern: str | None = None,
|
|
max_size: int | None = None,
|
|
) -> None:
|
|
from vllm_ascend._310p.sharded_state_loader_310p import ShardedStateLoader310
|
|
|
|
ShardedStateLoader310.save_model(
|
|
self.model_runner.model,
|
|
path,
|
|
pattern=pattern,
|
|
max_size=max_size,
|
|
)
|
|
|
|
ShardedStateLoader310.generate_quant_description(self.model_runner.model, path)
|
|
|
|
def _warm_up_atb(self):
|
|
# 310p device do not support torch_npu._npu_matmul_add_fp32 atb ops
|
|
logger.info("Skip warm-up atb ops for 310P device.")
|