EngineX/xc-llm-ascend
3
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

upgrade vLLM to main (#4608)

Browse Source 
1. fix https://github.com/vllm-project/vllm/pull/28542
The model structure modifications we involved in are:
     - Qwen2.5-VL(still exist some patch)
     - Qwen2-VL
     - Qwen2
     - DeepSeek series
     - Qwen-moe series
2. fix https://github.com/vllm-project/vllm/pull/29121
   the output token now  type changed from np to `list[list[int]]`

3. fix https://github.com/vllm-project/vllm/pull/29262
    `xformers` backend for multimodal now has been deprecated
4. fix https://github.com/vllm-project/vllm/pull/29342

5. fix https://github.com/vllm-project/vllm/pull/28579
6. fix https://github.com/vllm-project/vllm/pull/28718
7. fix https://github.com/vllm-project/vllm/issues/28665
8. fix https://github.com/vllm-project/vllm/pull/26847
vllm introduced the `optimization-level`, some default config has been
changed, and the param `--enforce-eager` has been deprecated
9. fix http://github.com/vllm-project/vllm/pull/29223 it retuns tuple
for sampler.
10. fix https://github.com/vllm-project/vllm/pull/29471 we'll remove the
related patch to avoid this kind of error.

Co-authored-by: hfadzxy <starmoon_zhang@163.com>
Co-authored-by: wangli <wangli858794774@gmail.com>


- vLLM version: v0.11.2

---------

Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
Signed-off-by: wangli <wangli858794774@gmail.com>
Signed-off-by: hfadzxy <starmoon_zhang@163.com>
Co-authored-by: wangli <wangli858794774@gmail.com>
Co-authored-by: hfadzxy <starmoon_zhang@163.com>
This commit is contained in:
wangxiyuan
2025-12-02 22:10:52 +08:00
committed by GitHub
parent 4588cdac02
commit 7f2673ea2d
60 changed files with 383 additions and 374 deletions
Download Patch File Download Diff File Expand all files Collapse all files

75
tests/ut/core/test_scheduler.py
View File

@@ -3,7 +3,7 @@
from typing import Any, Dict, List, Optional, Tuple
from unittest.mock import MagicMock, patch
import numpy as np
import pytest
import torch
from vllm.config import (CacheConfig, KVTransferConfig, ModelConfig,
SchedulerConfig, SpeculativeConfig, VllmConfig)
@@ -81,9 +81,7 @@ def make_output(scheduler):
req.request_id: i
for i, req in enumerate(scheduler.running)
}
sampled_token_ids = [
np.array([1000], dtype=np.int64) for _ in scheduler.running
]
sampled_token_ids = [[1000]] * len(scheduler.running)
logprobs = None
@@ -98,6 +96,7 @@ def make_output(scheduler):
return modelrunner_output
@pytest.mark.skip("Ascend Scheduler has been deprecated")
class TestAscendScheduler(TestBase):
@patch("vllm.config.ModelConfig.__post_init__", MagicMock())
@@ -372,8 +371,7 @@ class TestAscendScheduler(TestBase):
req.request_id: i
for i, req in enumerate(requests)
},
sampled_token_ids=[np.array([EOS_TOKEN_ID]),
np.array([10, 11])
sampled_token_ids=[[EOS_TOKEN_ID], [10, 11]
], # First request hits EOS, second continues
logprobs=None,
prompt_logprobs_dict={},
@@ -424,9 +422,8 @@ class TestAscendScheduler(TestBase):
req.request_id: i
for i, req in enumerate(requests)
},
sampled_token_ids=[np.array([10, 42, 12]),
np.array([13, 14])
], # First request hits stop token
sampled_token_ids=[[10, 42, 12],
[13, 14]], # First request hits stop token
logprobs=None,
prompt_logprobs_dict={},
pooler_output=[])
@@ -475,9 +472,8 @@ class TestAscendScheduler(TestBase):
req.request_id: i
for i, req in enumerate(requests)
},
sampled_token_ids=[np.array([10, 11, 12]),
np.array([13])
], # First request exceeds max_tokens
sampled_token_ids=[[10, 11, 12],
[13]], # First request exceeds max_tokens
logprobs=None,
prompt_logprobs_dict={},
pooler_output=[])
@@ -516,7 +512,7 @@ class TestAscendScheduler(TestBase):
model_output = ModelRunnerOutput(
req_ids=[requests[0].request_id],
req_id_to_index={requests[0].request_id: 0},
sampled_token_ids=[np.array([EOS_TOKEN_ID, 10, 11])],
sampled_token_ids=[[EOS_TOKEN_ID, 10, 11]],
logprobs=None,
prompt_logprobs_dict={},
pooler_output=[])
@@ -573,7 +569,7 @@ class TestAscendScheduler(TestBase):
model_runner_output = ModelRunnerOutput(
req_ids=[requests[0].request_id],
req_id_to_index={requests[0].request_id: 0},
sampled_token_ids=[np.array([0], dtype=np.int64)],
sampled_token_ids=[[0]],
logprobs=None,
prompt_logprobs_dict={},
pooler_output=[])
@@ -589,7 +585,7 @@ class TestAscendScheduler(TestBase):
model_runner_output = ModelRunnerOutput(
req_ids=[requests[1].request_id],
req_id_to_index={requests[1].request_id: 0},
sampled_token_ids=[np.array([0], dtype=np.int64)],
sampled_token_ids=[[0]],
logprobs=None,
prompt_logprobs_dict={},
pooler_output=[])
@@ -607,12 +603,10 @@ class TestAscendScheduler(TestBase):
spec_tokens_list: List[List[List[int]]] = [[[1, 2, 3]], [[1, 2, 3]],
[[1, 2], [3]], [[1]], [[]],
[[1, 2, 3], [4, 5, 6]]]
output_tokens_list: List[List[List[int]]] = [
[np.array([1, 2, 3, 4])], [np.array([1, 5])],
[np.array([1, 2, 5]), np.array([3, 4])], [np.array([1, 2])],
[np.array([5])], [np.array([1, 2, 7]),
np.array([4, 8])]
]
output_tokens_list: List[List[List[int]]] = [[[1, 2, 3, 4]], [[1, 5]],
[[1, 2, 5], [3, 4]],
[[1, 2]], [[5]],
[[1, 2, 7], [4, 8]]]
expected_list: List[Tuple[int, int,
int, List[int]]] = [(1, 3, 3, [1, 1, 1]),
(1, 3, 1, [1, 0, 0]),
@@ -650,9 +644,7 @@ class TestAscendScheduler(TestBase):
model_runner_output = ModelRunnerOutput(
req_ids=req_ids,
req_id_to_index=req_to_index,
sampled_token_ids=[
np.array([0]) for _ in range(len(requests))
],
sampled_token_ids=[[0] for _ in range(len(requests))],
logprobs=None,
prompt_logprobs_dict={},
pooler_output=[])
@@ -892,11 +884,13 @@ class TestSchedulerDynamicBatch(TestBase):
torch.float32, False))
],
)
kv_cache_config.hash_block_size = block_size
cache_config.num_gpu_blocks = 10000
scheduler = SchedulerDynamicBatch(
vllm_config=vllm_config,
kv_cache_config=kv_cache_config,
block_size=block_size,
log_stats=True,
structured_output_manager=MagicMock(spec=StructuredOutputManager),
)
@@ -1064,8 +1058,7 @@ class TestSchedulerDynamicBatch(TestBase):
req.request_id: i
for i, req in enumerate(requests)
},
sampled_token_ids=[np.array([EOS_TOKEN_ID]),
np.array([10, 11])
sampled_token_ids=[[EOS_TOKEN_ID], [10, 11]
], # First request hits EOS, second continues
logprobs=None,
prompt_logprobs_dict={},
@@ -1116,9 +1109,8 @@ class TestSchedulerDynamicBatch(TestBase):
req.request_id: i
for i, req in enumerate(requests)
},
sampled_token_ids=[np.array([10, 42, 12]),
np.array([13, 14])
], # First request hits stop token
sampled_token_ids=[[10, 42, 12],
[13, 14]], # First request hits stop token
logprobs=None,
prompt_logprobs_dict={},
pooler_output=[])
@@ -1167,9 +1159,8 @@ class TestSchedulerDynamicBatch(TestBase):
req.request_id: i
for i, req in enumerate(requests)
},
sampled_token_ids=[np.array([10, 11, 12]),
np.array([13])
], # First request exceeds max_tokens
sampled_token_ids=[[10, 11, 12],
[13]], # First request exceeds max_tokens
logprobs=None,
prompt_logprobs_dict={},
pooler_output=[])
@@ -1208,7 +1199,7 @@ class TestSchedulerDynamicBatch(TestBase):
model_output = ModelRunnerOutput(
req_ids=[requests[0].request_id],
req_id_to_index={requests[0].request_id: 0},
sampled_token_ids=[np.array([EOS_TOKEN_ID, 10, 11])],
sampled_token_ids=[[EOS_TOKEN_ID, 10, 11]],
logprobs=None,
prompt_logprobs_dict={},
pooler_output=[])
@@ -1265,7 +1256,7 @@ class TestSchedulerDynamicBatch(TestBase):
model_runner_output = ModelRunnerOutput(
req_ids=[requests[0].request_id],
req_id_to_index={requests[0].request_id: 0},
sampled_token_ids=[np.array([0])],
sampled_token_ids=[[0]],
logprobs=None,
prompt_logprobs_dict={},
pooler_output=[])
@@ -1281,7 +1272,7 @@ class TestSchedulerDynamicBatch(TestBase):
model_runner_output = ModelRunnerOutput(
req_ids=[requests[1].request_id],
req_id_to_index={requests[1].request_id: 0},
sampled_token_ids=[np.array([0])],
sampled_token_ids=[[0]],
logprobs=None,
prompt_logprobs_dict={},
pooler_output=[])
@@ -1299,12 +1290,10 @@ class TestSchedulerDynamicBatch(TestBase):
spec_tokens_list: List[List[List[int]]] = [[[1, 2, 3]], [[1, 2, 3]],
[[1, 2], [3]], [[1]], [[]],
[[1, 2, 3], [4, 5, 6]]]
output_tokens_list: List[List[List[int]]] = [
[np.array([1, 2, 3, 4])], [np.array([1, 5])],
[np.array([1, 2, 5]), np.array([3, 4])], [np.array([1, 2])],
[np.array([5])], [np.array([1, 2, 7]),
np.array([4, 8])]
]
output_tokens_list: List[List[List[int]]] = [[[1, 2, 3, 4]], [[1, 5]],
[[1, 2, 5], [3, 4]],
[[1, 2]], [[5]],
[[1, 2, 7], [4, 8]]]
expected_list: List[Tuple[int, int,
int, List[int]]] = [(1, 3, 3, [1, 1, 1]),
(1, 3, 1, [1, 0, 0]),
@@ -1342,9 +1331,7 @@ class TestSchedulerDynamicBatch(TestBase):
model_runner_output = ModelRunnerOutput(
req_ids=req_ids,
req_id_to_index=req_to_index,
sampled_token_ids=[
np.array([0]) for _ in range(len(requests))
],
sampled_token_ids=[[0] for _ in range(len(requests))],
logprobs=None,
prompt_logprobs_dict={},
pooler_output=[])