mirror of
https://github.com/PaddlePaddle/FastDeploy.git
synced 2026-04-23 00:17:25 +08:00
CSWYF3634076
08c411518f
* [Loader] support dummy load weight * [Loader] support dummy load weight v2 * [Loader] support dummy load weight unittest * [Loader] support dummy load weight unittest v2 * [Loader] support dummy load weight v3 docs and fp8
107 lines
4.1 KiB
Python
107 lines
4.1 KiB
Python
"""
|
|
# Copyright (c) 2026 PaddlePaddle Authors. 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.
|
|
"""
|
|
|
|
import time
|
|
|
|
import paddle
|
|
from paddle import nn
|
|
from paddleformers.utils.log import logger
|
|
from typing_extensions import assert_never
|
|
|
|
from fastdeploy.config import FDConfig, LoadConfig, ModelConfig
|
|
from fastdeploy.model_executor.load_weight_utils import is_weight_cache_enabled
|
|
from fastdeploy.model_executor.model_loader.base_loader import BaseModelLoader
|
|
from fastdeploy.model_executor.models.adapters import as_embedding_model
|
|
from fastdeploy.model_executor.models.model_base import ModelRegistry
|
|
from fastdeploy.model_executor.utils import process_final_after_loading
|
|
|
|
|
|
class DummyModelLoader(BaseModelLoader):
|
|
"""Model loader that initializes model weights with random values."""
|
|
|
|
def __init__(self, load_config: LoadConfig):
|
|
super().__init__(load_config)
|
|
logger.info("Load the model and initialize dummy weights")
|
|
|
|
def download_model(self, model_config: ModelConfig) -> None:
|
|
"""download_model"""
|
|
pass
|
|
|
|
def _initialize_dummy_weights(
|
|
self,
|
|
model: nn.Module,
|
|
low: float = -1e-3,
|
|
high: float = 1e-3,
|
|
) -> None:
|
|
float_dtypes = (
|
|
paddle.float16,
|
|
paddle.float32,
|
|
paddle.float64,
|
|
paddle.bfloat16,
|
|
)
|
|
float8_dtypes = (
|
|
paddle.float8_e4m3fn,
|
|
paddle.float8_e5m2,
|
|
)
|
|
with paddle.no_grad():
|
|
for _, param in model.named_parameters():
|
|
if param is None:
|
|
continue
|
|
if not param.shape or 0 in param.shape:
|
|
continue
|
|
if param.dtype in float8_dtypes:
|
|
tmp = (high - low) * paddle.randn(param.shape, dtype=paddle.float16) + low
|
|
param.copy_(tmp.cast(param.dtype), False)
|
|
elif param.dtype in float_dtypes:
|
|
param.set_value((high - low) * paddle.randn(param.shape, dtype=param.dtype) + low)
|
|
else:
|
|
param.set_value(paddle.zeros(param.shape, dtype=param.dtype))
|
|
|
|
def load_model(self, fd_config: FDConfig) -> nn.Layer:
|
|
start_dummy_weight_time = time.time()
|
|
architectures = fd_config.model_config.architectures[0]
|
|
context = paddle.LazyGuard()
|
|
if fd_config.load_config.dynamic_load_weight:
|
|
import fastdeploy.rl # noqa
|
|
|
|
if fd_config.speculative_config.model_type != "mtp":
|
|
architectures = architectures.replace("Ernie5ForCausalLM", "Ernie5MoeForCausalLM")
|
|
else:
|
|
architectures = architectures.replace("Ernie5ForCausalLM", "Ernie5MTPForCausalLM")
|
|
|
|
architectures = architectures + "RL"
|
|
|
|
enable_cache, _, weight_cache_context = is_weight_cache_enabled(fd_config)
|
|
fd_config.model_config.enable_cache = enable_cache
|
|
with weight_cache_context:
|
|
with context:
|
|
model_cls = ModelRegistry.get_class(architectures)
|
|
convert_type = fd_config.model_config.convert_type
|
|
if convert_type == "none":
|
|
pass
|
|
elif convert_type == "embed":
|
|
model_cls = as_embedding_model(model_cls)
|
|
else:
|
|
assert_never(convert_type)
|
|
|
|
model = model_cls(fd_config)
|
|
|
|
model.eval()
|
|
self._initialize_dummy_weights(model)
|
|
process_final_after_loading(model, fd_config)
|
|
logger.info("dummy weight cost time: {}s".format(time.time() - start_dummy_weight_time))
|
|
return model
|