Add new dependency library for vulkan tests #10136
Conversation
**Background**: We removed the inline flag of the symbol `torch::jit::GetBackendMetaSerialization()` from and moved it to .cpp in order to make the dynamic registration related to PyTorch serialization work properly. However, after testing, it was found that Vulkan's tests depend on the symbol named `torch::jit::GetBackendMetaSerialization()`, so after the above changes, it is necessary to explicitly link libtorch_cpu (which contains the torch::jit::GetBackendMetaSerialization() symbol) For more information, see [PR](pytorch/pytorch#147095), especially this [comment](pytorch/pytorch#147095 (comment))
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/10136
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit e779c65 with merge base e261ed1 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
pytorch-bot
bot
added
the
module: vulkan
facebook-github-bot
added
the
CLA Signed
|
@pytorchbot label "topic: not user facing" |
|
Hey! So sorry to bother @kirklandsign @shoumikhin @SS-JIA, could you please help to trigger the workflow and merge this pr into the trunk? |
|
@FFFrog apologies for the delayed response. Merging now. |
Thank you a lot. |
Hey @kirklandsign , so sorry for breaking the internal tests, could you mind to show me the detailed information about the failure? |
|
@FFFrog no worries. To address it properly I think we will need to test against the Meta internal repo. I will try to implement your changes in the internal repo and figure out a way to do it that doesn't break the internal tests. |
|
@SS-JIA Thank you very much,please let me know if any progress. |
**Background**: We removed the inline flag of the symbol `torch::jit::GetBackendMetaSerialization()` from and moved it to .cpp in order to make the dynamic registration related to PyTorch serialization work properly. However, after testing, it was found that Vulkan's tests depend on the symbol named `torch::jit::GetBackendMetaSerialization()`, so after the above changes, it is necessary to explicitly link libtorch_cpu (which contains the torch::jit::GetBackendMetaSerialization() symbol) For more information, see [PR](pytorch/pytorch#147095), especially this [comment](pytorch/pytorch#147095 (comment)) cc @SS-JIA @manuelcandales @cbilgin
Background:
We removed the inline flag of the symbol
torch::jit::GetBackendMetaSerialization()from and moved it to .cpp in order to make the dynamic registration related to PyTorch serialization work properly.However, after testing, it was found that Vulkan's tests depend on the symbol named
torch::jit::GetBackendMetaSerialization(), so after the above changes, it is necessary to explicitly link libtorch_cpu (which contains the torch::jit::GetBackendMetaSerialization() symbol)For more information, see PR, especially this comment
cc @SS-JIA @manuelcandales @cbilgin