Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

[ROCm] Fix build problem resulted from previous commit related to FP8 kv-cache support #2790

Merged
merged 2 commits into from
Feb 7, 2024
Merged

[ROCm] Fix build problem resulted from previous commit related to FP8 kv-cache support #2790

merged 2 commits into from
Feb 7, 2024

Conversation

hongxiayang
Copy link
Collaborator

@hongxiayang hongxiayang commented Feb 6, 2024 •
edited
Loading

Fixes: #2725
Current head failed to build on ROCm, and I got errors like:

g++ -pthread -B /opt/conda/envs/py_3.8/compiler_compat -Wl,--sysroot=/ -pthread -shared -B /opt/conda/envs/py_3.8/compiler_compat -L/opt/conda/envs/py_3.8/lib -Wl,-rpath=/opt/conda/envs/py_3.8/lib -WlEN,--no-as-needed -Wl,--sysroot=/ /app/vllm/build/temp.linux-x86_64-cpython-38/csrc/activation_kernels.o /app/vllm/build/temp.linux-x86_64-cpython-38/csrc/attention/attention_kernels.o /app/vllm/build/temp.linux-x86_64-cpython-38/csrc/cache_kernels.o /app/vllm/build/temp.linux-x86_64-cpython-38/csrc/hip_utils_kernels.o /app/vllm/build/temp.linux-x86_64-cpython-38/csrc/layernorm_kernels.o /app/vllm/build/temp.linux-x86_64-cpython-38/csrc/moe_align_block_size_kernels.o /app/vllm/build/temp.linux-x86_64-cpython-38/csrc/pos_encoding_kernels.o /app/vllm/build/temp.linux-x86_64-cpython-38/csrc/pybind.o /app/vllm/build/temp.linux-x86_64-cpython-38/csrc/quantization/gptq/q_gemm.o /app/vllm/build/temp.linux-x86_64-cpython-38/csrc/quantization/squeezellm/quant_hip_kernel.o -L/opt/conda/envs/py_3.8/lib/python3.8/site-packages/torch/lib -L/opt/rocm/lib -L/opt/rocm/hip/lib -lc10 -ltorch -ltorch_cpu -ltorch_python -lamdhip64 -lc10_hip -ltorch_hip -o build/lib.linux-x86_64-cpython-38/vllm/_C.cpython-38-x86_64-linux-gnu.so
/opt/conda/envs/py_3.8/compiler_compat/ld: /app/vllm/build/temp.linux-x86_64-cpython-38/csrc/cache_kernels.o: in function `__float2bfloat16(float)':
cache_kernels.hip:(.text+0x0): multiple definition of `__float2bfloat16(float)'; /app/vllm/build/temp.linux-x86_64-cpython-38/csrc/attention/attention_kernels.o:attention_kernels.hip:(.text+0x0): first defined here
/opt/conda/envs/py_3.8/compiler_compat/ld: /app/vllm/build/temp.linux-x86_64-cpython-38/csrc/cache_kernels.o: in function `__bfloat1622float2(__hip_bfloat162)':
cache_kernels.hip:(.text+0x40): multiple definition of `__bfloat1622float2(__hip_bfloat162)'; /app/vllm/build/temp.linux-x86_64-cpython-38/csrc/attention/attention_kernels.o:attention_kernels.hip:(.text+0x40): first defined here
/opt/conda/envs/py_3.8/compiler_compat/ld: /app/vllm/build/temp.linux-x86_64-cpython-38/csrc/cache_kernels.o: in function `__double2bfloat16(double)':
cache_kernels.hip:(.text+0x60): multiple definition of `__double2bfloat16(double)'; /app/vllm/build/temp.linux-x86_64-cpython-38/csrc/attention/attention_kernels.o:attention_kernels.hip:(.text+0x60): first defined here
/opt/conda/envs/py_3.8/compiler_compat/ld: /app/vllm/build/temp.linux-x86_64-cpython-38/csrc/cache_kernels.o: in function `__float22bfloat162_rn(HIP_vector_type<float, 2u>)':
cache_kernels.hip:(.text+0xa0): multiple definition of `__float22bfloat162_rn(HIP_vector_type<float, 2u>)'; /app/vllm/build/temp.linux-x86_64-cpython-38/csrc/attention/attention_kernels.o:attention_kernels.hip:(.text+0xa0): first defined here
/opt/conda/envs/py_3.8/compiler_compat/ld: /app/vllm/build/temp.linux-x86_64-cpython-

We need a patch to fix the compilation issue before the next ROCm release is available.
This pull request fixed the build issue for ROCm build.

@hongxiayang hongxiayang marked this pull request as ready for review February 6, 2024 19:38
@hongxiayang hongxiayang mentioned this pull request Feb 6, 2024
Closed
@jamestwhedbee
Copy link
Contributor

I am using ROCm 5.7 and before this PR, v0.3.0 of vLLM could not be built. Applying this patch fixes it for me! Just thought it was worth mentioning that this issue was not limited to ROCm 6.0

@hongxiayang
Copy link
Collaborator Author

I am using ROCm 5.7 and before this PR, v0.3.0 of vLLM could not be built. Applying this patch fixes it for me! Just thought it was worth mentioning that this issue was not limited to ROCm 6.0

Thanks for your comment. I will update to apply the patch regardless the version.

@hongxiayang hongxiayang mentioned this pull request Feb 6, 2024
Merged
zhuohan123
zhuohan123 approved these changes Feb 7, 2024
View reviewed changes
Copy link
Member

@zhuohan123 zhuohan123 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM! Thanks for the fix!

@zhuohan123 zhuohan123 merged commit c81dddb into vllm-project:main Feb 7, 2024
17 checks passed
hongxiayang added a commit to hongxiayang/vllm that referenced this pull request Feb 13, 2024
@hongxiayang
[ROCm] Fix build problem resulted from previous commit related to FP8…
dbdb467 
… kv-cache support (vllm-project#2790)
yhu422 added a commit to yhu422/vllm that referenced this pull request Feb 13, 2024
@yhu422
Write info to local json
8a2f18a 
[ROCm] Fix build problem resulted from previous commit related to FP8 kv-cache support  (vllm-project#2790)

Add documentation on how to do incremental builds (vllm-project#2796)

[Ray] Integration compiled DAG off by default (vllm-project#2471)

Disable custom all reduce by default (vllm-project#2808)

add usage context

removed usage_context from Engine_args

Move IO to another process

added http request

[ROCm] support Radeon™ 7900 series (gfx1100) without using flash-attention (vllm-project#2768)

Add documentation section about LoRA (vllm-project#2834)

Refactor 2 awq gemm kernels into m16nXk32 (vllm-project#2723)

Co-authored-by: Chunan Zeng <chunanzeng@Chunans-Air.attlocal.net>

Added additional arg for from_engine_args

comments
alexm-neuralmagic pushed a commit to neuralmagic/nm-vllm that referenced this pull request Feb 13, 2024
@hongxiayang @alexm-neuralmagic
[ROCm] Fix build problem resulted from previous commit related to FP8…
ea19b11 
… kv-cache support (vllm-project#2790)
jvmncs pushed a commit to jvmncs/vllm that referenced this pull request Feb 14, 2024
@hongxiayang @jvmncs
[ROCm] Fix build problem resulted from previous commit related to FP8…
5cb2c3a 
… kv-cache support (vllm-project#2790)
xjpang pushed a commit to xjpang/vllm that referenced this pull request Feb 20, 2024
@hongxiayang
[ROCm] Fix build problem resulted from previous commit related to FP8…
38deef5 
… kv-cache support (vllm-project#2790)
xjpang pushed a commit to xjpang/vllm that referenced this pull request Feb 22, 2024
@hongxiayang
[ROCm] Fix build problem resulted from previous commit related to FP8…
88483a6 
… kv-cache support (vllm-project#2790)
@andy-neuma andy-neuma mentioned this pull request Feb 23, 2024
Closed
xjpang pushed a commit to xjpang/vllm that referenced this pull request Mar 4, 2024
@hongxiayang
[ROCm] Fix build problem resulted from previous commit related to FP8…
8f5d8d1 
… kv-cache support (vllm-project#2790)
@fxmarty
Copy link

fxmarty commented Apr 18, 2024

Thank you @hongxiayang

@fxmarty fxmarty mentioned this pull request Apr 18, 2024
Closed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@zhuohan123 zhuohan123 zhuohan123 approved these changes

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

[BUG] Compile source code error for ROCM platform when using #include <hip/hip_bf16.h>
4 participants
@hongxiayang @jamestwhedbee @fxmarty @zhuohan123