mirror of
https://git.datalinker.icu/vllm-project/vllm.git
synced 2025-12-12 04:15:02 +08:00
e489ad7a21
- **Add SPDX license headers to python source files**
- **Check for SPDX headers using pre-commit**
commit 9d7ef44c3cfb72ca4c32e1c677d99259d10d4745
Author: Russell Bryant <rbryant@redhat.com>
Date: Fri Jan 31 14:18:24 2025 -0500
Add SPDX license headers to python source files
This commit adds SPDX license headers to python source files as
recommended to
the project by the Linux Foundation. These headers provide a concise way
that is
both human and machine readable for communicating license information
for each
source file. It helps avoid any ambiguity about the license of the code
and can
also be easily used by tools to help manage license compliance.
The Linux Foundation runs license scans against the codebase to help
ensure
we are in compliance with the licenses of the code we use, including
dependencies. Having these headers in place helps that tool do its job.
More information can be found on the SPDX site:
- https://spdx.dev/learn/handling-license-info/
Signed-off-by: Russell Bryant <rbryant@redhat.com>
commit 5a1cf1cb3b80759131c73f6a9dddebccac039dea
Author: Russell Bryant <rbryant@redhat.com>
Date: Fri Jan 31 14:36:32 2025 -0500
Check for SPDX headers using pre-commit
Signed-off-by: Russell Bryant <rbryant@redhat.com>
---------
Signed-off-by: Russell Bryant <rbryant@redhat.com>
124 lines
4.6 KiB
Python
124 lines
4.6 KiB
Python
# SPDX-License-Identifier: Apache-2.0
|
|
|
|
import os
|
|
|
|
import pytest
|
|
|
|
from vllm.model_executor.layers.pooler import CLSPool, PoolingType
|
|
from vllm.model_executor.models.bert import BertEmbeddingModel
|
|
from vllm.model_executor.models.roberta import RobertaEmbeddingModel
|
|
from vllm.platforms import current_platform
|
|
|
|
MAX_MODEL_LEN = 128
|
|
MODEL_NAME = os.environ.get("MODEL_NAME", "BAAI/bge-base-en-v1.5")
|
|
REVISION = os.environ.get("REVISION", "main")
|
|
|
|
MODEL_NAME_ROBERTA = os.environ.get("MODEL_NAME",
|
|
"intfloat/multilingual-e5-large")
|
|
REVISION_ROBERTA = os.environ.get("REVISION", "main")
|
|
|
|
|
|
@pytest.mark.skipif(current_platform.is_rocm(),
|
|
reason="Xformers backend is not supported on ROCm.")
|
|
def test_model_loading_with_params(vllm_runner):
|
|
"""
|
|
Test parameter weight loading with tp>1.
|
|
"""
|
|
with vllm_runner(model_name=MODEL_NAME,
|
|
revision=REVISION,
|
|
dtype="float16",
|
|
max_model_len=MAX_MODEL_LEN) as vllm_model:
|
|
output = vllm_model.encode("Write a short story about a robot that"
|
|
" dreams for the first time.\n")
|
|
|
|
model_config = vllm_model.model.llm_engine.model_config
|
|
model_tokenizer = vllm_model.model.llm_engine.tokenizer
|
|
|
|
# asserts on the bert model config file
|
|
assert model_config.encoder_config["max_seq_length"] == 512
|
|
assert model_config.encoder_config["do_lower_case"]
|
|
|
|
# asserts on the pooling config files
|
|
assert model_config.pooler_config.pooling_type == PoolingType.CLS.name
|
|
assert model_config.pooler_config.pooling_norm
|
|
|
|
# asserts on the tokenizer loaded
|
|
assert model_tokenizer.tokenizer_id == "BAAI/bge-base-en-v1.5"
|
|
assert model_tokenizer.tokenizer_config["do_lower_case"]
|
|
assert model_tokenizer.tokenizer.model_max_length == 512
|
|
|
|
def check_model(model):
|
|
assert isinstance(model, BertEmbeddingModel)
|
|
assert model._pooler.pooling_type == PoolingType.CLS
|
|
assert model._pooler.normalize
|
|
|
|
vllm_model.apply_model(check_model)
|
|
|
|
# assert output
|
|
assert output
|
|
|
|
|
|
@pytest.mark.skipif(current_platform.is_rocm(),
|
|
reason="Xformers backend is not supported on ROCm.")
|
|
def test_roberta_model_loading_with_params(vllm_runner):
|
|
"""
|
|
Test parameter weight loading with tp>1.
|
|
"""
|
|
with vllm_runner(model_name=MODEL_NAME_ROBERTA,
|
|
revision=REVISION_ROBERTA,
|
|
dtype="float16",
|
|
max_model_len=MAX_MODEL_LEN) as vllm_model:
|
|
output = vllm_model.encode("Write a short story about a robot that"
|
|
" dreams for the first time.\n")
|
|
|
|
model_config = vllm_model.model.llm_engine.model_config
|
|
model_tokenizer = vllm_model.model.llm_engine.tokenizer
|
|
|
|
# asserts on the bert model config file
|
|
assert model_config.encoder_config["max_seq_length"] == 512
|
|
assert not model_config.encoder_config["do_lower_case"]
|
|
|
|
# asserts on the pooling config files
|
|
assert model_config.pooler_config.pooling_type == PoolingType.MEAN.name
|
|
assert model_config.pooler_config.pooling_norm
|
|
|
|
# asserts on the tokenizer loaded
|
|
assert model_tokenizer.tokenizer_id == "intfloat/multilingual-e5-large"
|
|
assert not model_tokenizer.tokenizer_config["do_lower_case"]
|
|
|
|
def check_model(model):
|
|
assert isinstance(model, RobertaEmbeddingModel)
|
|
assert model._pooler.pooling_type == PoolingType.MEAN
|
|
assert model._pooler.normalize
|
|
|
|
vllm_model.apply_model(check_model)
|
|
|
|
# assert output
|
|
assert output
|
|
|
|
|
|
@pytest.mark.skipif(current_platform.is_rocm(),
|
|
reason="Xformers backend is not supported on ROCm.")
|
|
def test_facebook_roberta_model_loading_with_params(vllm_runner):
|
|
"""
|
|
Test loading roberta-base model with no lm_head.
|
|
"""
|
|
model_name = "FacebookAI/roberta-base"
|
|
with vllm_runner(model_name=model_name,
|
|
dtype="float16",
|
|
max_model_len=MAX_MODEL_LEN) as vllm_model:
|
|
output = vllm_model.encode("Write a short story about a robot that"
|
|
" dreams for the first time.\n")
|
|
|
|
model_tokenizer = vllm_model.model.llm_engine.tokenizer
|
|
assert model_tokenizer.tokenizer_id == model_name
|
|
|
|
def check_model(model):
|
|
assert isinstance(model, RobertaEmbeddingModel)
|
|
assert not hasattr(model, "lm_head")
|
|
assert isinstance(model._pooler, CLSPool)
|
|
|
|
vllm_model.apply_model(check_model)
|
|
|
|
assert output
|