Bump torch version to 2.7.1 #1562

leo-pony · 2025-07-01T09:38:48Z

Signed-off-by: Yikun Jiang yikunkero@gmail.com
Signed-off-by: leo-pony nengjunma@outlook.com

Replace PR 1527

What this PR does / why we need it?

What this PR does / why we need it?
Bump torch version to 2.7.1
Cleanup 857f489 (#837)

Does this PR introduce any user-facing change?

No

How was this patch tested?
CI passed

torch-npu 2.7.1rc1 install guide:
https://gitee.com/ascend/pytorch/tree/v2.7.1/
install depending:

pip3 install pyyaml
pip3 install setuptools

install torch-npu:

vLLM version: v0.9.2
vLLM main: vllm-project/vllm@13e4ee1

github-actions · 2025-07-01T09:41:33Z

This pull request has conflicts, please resolve those before we can evaluate the pull request.

codecov · 2025-07-04T10:02:51Z

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 71.27%. Comparing base (b5ad70e) to head (2e3f6f3).
Report is 6 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1562      +/-   ##
==========================================
+ Coverage   65.78%   71.27%   +5.49%     
==========================================
  Files          78       84       +6     
  Lines        8406     9063     +657     
==========================================
+ Hits         5530     6460     +930     
+ Misses       2876     2603     -273

Flag	Coverage Δ
unittests	`71.27% <100.00%> (+5.49%)`	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

github-actions · 2025-07-06T02:44:40Z

This pull request has conflicts, please resolve those before we can evaluate the pull request.

csrc/kernels/get_masked_input_and_mask_kernel.cpp

github-actions · 2025-07-11T07:35:20Z

This pull request has conflicts, please resolve those before we can evaluate the pull request.

Signed-off-by: Yikun Jiang <yikunkero@gmail.com>

Co-authored-by: leo-pony <nengjunma@outlook.com> Signed-off-by: Yikun Jiang <yikunkero@gmail.com>

Signed-off-by: Yikun Jiang <yikunkero@gmail.com>

…h_common.patch_utils Signed-off-by: leo-pony <nengjunma@outlook.com>

…ions now Signed-off-by: leo-pony <nengjunma@outlook.com>

Signed-off-by: leo-pony <nengjunma@outlook.com>

…nt calculating logical to device Signed-off-by: leo-pony <nengjunma@outlook.com>

…ted liberror_manager.so reference Signed-off-by: leo-pony <nengjunma@outlook.com>

Signed-off-by: leo-pony <nengjunma@outlook.com>

Yikun

Let's wait torch npu 2.7.1 first POC version.

Yikun · 2025-07-24T07:27:29Z

tests/ut/ops/test_expert_load_balancer.py

@@ -12,7 +12,6 @@
 # limitations under the License.
 # This file is a part of the vllm-ascend project.
 #
-


revert unrelated change to avoid cherrypick conflict

Yikun · 2025-07-24T07:28:34Z

csrc/torch_binding.cpp

@@ -193,7 +192,6 @@ std::tuple<at::Tensor, at::Tensor> get_masked_input_and_mask(
            size,
            loop_cnt,
            aiv_num);
-


revert unrelated change to avoid cherrypick conflict

Yikun · 2025-07-24T07:28:36Z

csrc/torch_binding.cpp

@@ -178,7 +178,6 @@ std::tuple<at::Tensor, at::Tensor> get_masked_input_and_mask(
        int64_t aiv_num = 0;
        TORCH_CHECK(aclGetDeviceCapability(device_id, ACL_DEVICE_INFO_VECTOR_CORE_NUM, &aiv_num) == ACL_SUCCESS);
        uint32_t loop_cnt = (size + aiv_num - 1) / aiv_num;
-


revert unrelated change to avoid cherrypick conflict

Yikun · 2025-07-24T15:01:34Z

README.md

@@ -42,7 +42,7 @@ By using vLLM Ascend plugin, popular open-source models, including Transformer-l
 - Software:
  * Python >= 3.9, < 3.12
  * CANN >= 8.1.RC1
-  * PyTorch >= 2.5.1, torch-npu >= 2.5.1.post1.dev20250619
+  * PyTorch >= 2.7.1, torch-npu >= 2.7.1rc1


Suggested change

* PyTorch >= 2.7.1, torch-npu >= 2.7.1rc1

* PyTorch >= 2.7.1, torch-npu >= 2.7.1.dev20250724

Yikun · 2025-07-24T15:02:42Z

README.zh.md

@@ -42,7 +42,7 @@ vLLM 昇腾插件 (`vllm-ascend`) 是一个由社区维护的让vLLM在Ascend NP
 - 软件：
  * Python >= 3.9, < 3.12
  * CANN >= 8.1.RC1
-  * PyTorch >= 2.5.1, torch-npu >= 2.5.1.post1.dev20250619
+  * PyTorch >= 2.5.1, torch-npu >= 2.7.1rc1


Suggested change

* PyTorch >= 2.5.1, torch-npu >= 2.7.1rc1

* PyTorch >= 2.5.1, torch-npu >= 2.7.1.dev20250724

Yikun · 2025-07-24T15:03:55Z

pyproject.toml

-    "torch-npu==2.5.1.post1.dev20250619",
-    "torch>=2.5.1",
-    "torchvision<0.21.0",
+    "torch-npu==2.7.1rc1",


Suggested change

"torch-npu==2.7.1rc1",

"torch-npu==2.7.1.dev20250724",

Yikun · 2025-07-24T15:04:21Z

requirements.txt

@@ -24,4 +24,4 @@ numba
 # Install torch_npu
 --pre
 --extra-index-url https://mirrors.huaweicloud.com/ascend/repos/pypi
-torch-npu==2.5.1.post1.dev20250619
+torch-npu==2.7.1rc1


Suggested change

torch-npu==2.7.1rc1

torch-npu==2.7.1.dev20250724

Yikun · 2025-07-24T15:05:18Z

tests/ut/test_utils.py

@@ -1,4 +1,3 @@
-#


revert unrelated change

Yikun · 2025-07-24T15:06:05Z

vllm_ascend/platform.py

@@ -254,16 +254,10 @@ def stateless_init_device_torch_dist_pg(

        assert is_hccl_available()

-        # TODO(Yizhou): The reason we need to set options while vllm does not


@yiz-liu FYI, we can finally clean up this, thanks for your note.

leo-pony marked this pull request as draft July 1, 2025 09:39

github-actions bot added the merge-conflicts label Jul 1, 2025

leo-pony force-pushed the torch_2_7_adapt branch from 41119fe to 6eb684d Compare July 1, 2025 09:44

github-actions bot added module:tests module:core module:quantization labels Jul 1, 2025

leo-pony closed this Jul 1, 2025

leo-pony force-pushed the torch_2_7_adapt branch from 6eb684d to 6db7dc2 Compare July 1, 2025 13:14

leo-pony reopened this Jul 1, 2025

github-actions bot added documentation Improvements or additions to documentation and removed merge-conflicts labels Jul 1, 2025

leo-pony force-pushed the torch_2_7_adapt branch 2 times, most recently from 04ddefc to 7bab351 Compare July 3, 2025 08:34

zheliuyu mentioned this pull request Jul 3, 2025

[Feature]: Request vllm-ascend to support torch_npu>=2.6 #1390

Open

leo-pony force-pushed the torch_2_7_adapt branch 2 times, most recently from 4f45d58 to b67e468 Compare July 4, 2025 09:45

github-actions bot added the merge-conflicts label Jul 6, 2025

wangxiyuan force-pushed the torch_2_7_adapt branch 2 times, most recently from e113a90 to f10b808 Compare July 9, 2025 03:38

github-actions bot removed the merge-conflicts label Jul 9, 2025

leo-pony mentioned this pull request Jul 9, 2025

[CI] fix wheel build fail: "liberror_manager.so" could not be localed #1690

Closed

github-actions bot added the ci/build label Jul 10, 2025

leo-pony marked this pull request as ready for review July 10, 2025 06:11

leo-pony marked this pull request as draft July 10, 2025 06:13

Yikun added accuracy-test enable all accuracy test for PR ready-for-test start test by label for PR labels Jul 10, 2025

Yikun reviewed Jul 10, 2025

View reviewed changes

csrc/kernels/get_masked_input_and_mask_kernel.cpp Outdated Show resolved Hide resolved

github-actions bot added the merge-conflicts label Jul 11, 2025

This was referenced Jul 14, 2025

[MISC] Move bind_kv_cache to worker module vllm-project/vllm#20900

Merged

[Installation]: Unmatched torch version between vllm=0.9.2 and vllm-ascend=0.9.2rc1 #1796

Closed

Yikun mentioned this pull request Jul 24, 2025

[Optimize]Change AI Vector core number getting function to glibc ABI free funcition #1974

Merged

Yikun and others added 13 commits July 24, 2025 02:23

Bump torch version to 2.7.1

3c72389

Signed-off-by: Yikun Jiang <yikunkero@gmail.com>

revert 857f489 68fb634

2a3af3b

Signed-off-by: Yikun Jiang <yikunkero@gmail.com>

Fix torchvision version

993b73f

Co-authored-by: leo-pony <nengjunma@outlook.com> Signed-off-by: Yikun Jiang <yikunkero@gmail.com>

job test

1f82b4b

Signed-off-by: Yikun Jiang <yikunkero@gmail.com>

Fix platform GetCoreNumByType abi mismach error and remove unuse patc…

d958bd3

…h_common.patch_utils Signed-off-by: leo-pony <nengjunma@outlook.com>

Fix test_platform ut fail: torch2.7 ProcessGroup does not include Opt…

9210897

…ions now Signed-off-by: leo-pony <nengjunma@outlook.com>

Remove unused Options for ProccessGroup in torch2.7

371792c

Signed-off-by: leo-pony <nengjunma@outlook.com>

Fix code formate

abf1ff6

Signed-off-by: leo-pony <nengjunma@outlook.com>

Remove main branch new added patch.utils calls

88c5575

Signed-off-by: leo-pony <nengjunma@outlook.com>

Fix main branch new added ProcessGroup error usage for torch2.7

c63ee8d

Signed-off-by: leo-pony <nengjunma@outlook.com>

Fix torch_binding.cpp abi issue: move aiv core num getting and loop c…

cad0d1e

…nt calculating logical to device Signed-off-by: leo-pony <nengjunma@outlook.com>

Fix build wheel failed test case: torch-npu torch_biinding auto-impor…

a1b302e

…ted liberror_manager.so reference Signed-off-by: leo-pony <nengjunma@outlook.com>

revert custom kernels tiling on device action

2e3f6f3

Signed-off-by: leo-pony <nengjunma@outlook.com>

leo-pony force-pushed the torch_2_7_adapt branch from a7efe04 to 2e3f6f3 Compare July 24, 2025 03:36

github-actions bot removed the merge-conflicts label Jul 24, 2025

leo-pony marked this pull request as ready for review July 24, 2025 06:46

Yikun reviewed Jul 24, 2025

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Bump torch version to 2.7.1 #1562

Bump torch version to 2.7.1 #1562

leo-pony commented Jul 1, 2025 •

edited by github-actions bot

Loading

Uh oh!

github-actions bot commented Jul 1, 2025

Uh oh!

codecov bot commented Jul 4, 2025 •

edited

Loading

Uh oh!

github-actions bot commented Jul 6, 2025

Uh oh!

Uh oh!

github-actions bot commented Jul 11, 2025

Uh oh!

Yikun left a comment

Uh oh!

Yikun Jul 24, 2025

Uh oh!

Yikun Jul 24, 2025

Uh oh!

Yikun Jul 24, 2025

Uh oh!

Yikun Jul 24, 2025

Uh oh!

Yikun Jul 24, 2025

Uh oh!

Yikun Jul 24, 2025

Uh oh!

Yikun Jul 24, 2025

Uh oh!

Yikun Jul 24, 2025

Uh oh!

Yikun Jul 24, 2025

Uh oh!

Uh oh!

	* PyTorch >= 2.7.1, torch-npu >= 2.7.1rc1
	* PyTorch >= 2.7.1, torch-npu >= 2.7.1.dev20250724

	* PyTorch >= 2.5.1, torch-npu >= 2.7.1rc1
	* PyTorch >= 2.5.1, torch-npu >= 2.7.1.dev20250724

		@@ -254,16 +254,10 @@ def stateless_init_device_torch_dist_pg(

		assert is_hccl_available()

		# TODO(Yizhou): The reason we need to set options while vllm does not

Bump torch version to 2.7.1 #1562

Are you sure you want to change the base?

Bump torch version to 2.7.1 #1562

Conversation

leo-pony commented Jul 1, 2025 • edited by github-actions bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What this PR does / why we need it?

Does this PR introduce any user-facing change?

Uh oh!

github-actions bot commented Jul 1, 2025

Uh oh!

codecov bot commented Jul 4, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

github-actions bot commented Jul 6, 2025

Uh oh!

Uh oh!

github-actions bot commented Jul 11, 2025

Uh oh!

Yikun left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

leo-pony commented Jul 1, 2025 •

edited by github-actions bot

Loading

codecov bot commented Jul 4, 2025 •

edited

Loading