Support qwen for pytorch engine

[RunningLeon]

Motivation

Support Qwen for pytorch engine

opencompass results:

Modification

Please briefly describe what modification is made in this PR.

BC-breaking (Optional)

Does the modification introduce changes that break the backward-compatibility of the downstream repositories?
If so, please describe how it breaks the compatibility and how the downstream projects should modify their code to keep compatibility with this PR.

Use cases (Optional)

If this PR introduces a new feature, it is better to list some use cases here, and update the documentation.

Checklist

1. Pre-commit or other linting tools are used to fix the potential lint issues.
2. The modification is covered by complete unit tests. If not, please add more unit tests to ensure the correctness.
3. If the modification has a dependency on downstream projects of a newer version, this PR should be tested with all supported versions of downstream projects.
4. The documentation has been modified accordingly, like docstring or example tutorials.

[jjjjohnson]

Looks like not support logn_attn yet

[RunningLeon]

@jjjjohnson hi, thanks for the reminder. Will update later.

[jjjjohnson]

@RunningLeon Look like there is en error in Qwen pytorch backend...

from lmdeploy.messages import PytorchEngineConfig, EngineGenerationConfig
from lmdeploy.pytorch.engine import Engine
from lmdeploy.serve.async_engine import AsyncEngine
import logging
logger = logging.getLogger(__name__)
logger.setLevel(logging.DEBUG)



model_name='qwen-14b'
engine_config = PytorchEngineConfig(model_name=model_name,
                                    tp=1,max_batch_size=8,
                                    adapters=None,num_gpu_blocks=200, num_cpu_blocks=100)


engin= AsyncEngine(model_path=model_path, backend='pytorch', model_name=model_name, backend_config=engine_config)

prompts = 'Hello'
engin(prompts)

[RunningLeon]

@jjjjohnson Hi, thanks for your feedback. I'll try to reproduce the issue. Coming back to you soon.

[RunningLeon]

@jjjjohnson hi, lmdeploy had supported qwen with old version. It seems you are running new version of Qwen and some rewriting fails in this case. Could try PR #1351? thanks for your understanding.

[jjjjohnson]

Cool! THX!