[Bug] AssertionError: sm90 not supported yet - running a model on Nvidia H100

daniel-kukiela commented 1 year ago

🐛 Bug

I'm not sure if this is a bug or should be a feature request. If you have "too new" GPU, you cannot use it. With the model compilation it errors:

[...]
  File "/usr/local/lib/python3.8/dist-packages/tvm/contrib/cutlass/gen_gemm.py", line 197, in __init__
    assert sm in GENERATOR_FUNC_TABLE and sm in DEFAULT_KERNELS, f"sm{sm} not supported yet."
AssertionError: sm90 not supported yet.

Is there a way to use the highest supported sm flag set or to specify one to use for the model compilation?

To Reproduce

Steps to reproduce the behavior:

Use unsupported GPU

Expected behavior

The model can be compiled and run even if it does not utilize all the capabilities of the hardware.

Environment

Platform (e.g. WebGPU/Vulkan/IOS/Android/CUDA): CUDA
Operating system (e.g. Ubuntu/Windows/MacOS/...): Ubuntu 22.04
Device (e.g. iPhone 12 Pro, PC+RTX 3090, ...): Nvidia H100
How you installed MLC-LLM (conda, source): source
How you installed TVM-Unity (pip, source): pip
Python version (e.g. 3.10): 3.8
GPU driver version (if applicable): 525.105.17
CUDA/cuDNN version (if applicable): 11.8

TVM Unity Hash Tag (python -c "import tvm; print('\n'.join(f'{k}: {v}' for k, v in tvm.support.libinfo().items()))", applicable if you compile models):

USE_GTEST: AUTO
SUMMARIZE: OFF
USE_IOS_RPC: OFF
USE_ETHOSU:
CUDA_VERSION: 11.8
USE_LIBBACKTRACE: AUTO
DLPACK_PATH: 3rdparty/dlpack/include
USE_TENSORRT_CODEGEN: OFF
USE_THRUST: OFF
USE_TARGET_ONNX: OFF
USE_AOT_EXECUTOR: ON
BUILD_DUMMY_LIBTVM: OFF
USE_CUDNN: OFF
USE_TENSORRT_RUNTIME: OFF
USE_ARM_COMPUTE_LIB_GRAPH_EXECUTOR: OFF
USE_CCACHE: AUTO
USE_ARM_COMPUTE_LIB: OFF
USE_CPP_RTVM:
USE_OPENCL_GTEST: /path/to/opencl/gtest
USE_MKL: OFF
USE_PT_TVMDSOOP: OFF
USE_CLML: OFF
USE_STACKVM_RUNTIME: OFF
USE_GRAPH_EXECUTOR_CUDA_GRAPH: OFF
ROCM_PATH: /opt/rocm
USE_DNNL: OFF
USE_VITIS_AI: OFF
USE_LLVM: llvm-config --ignore-libllvm --link-static
USE_VERILATOR: OFF
USE_TF_TVMDSOOP: OFF
USE_THREADS: ON
USE_MSVC_MT: OFF
BACKTRACE_ON_SEGFAULT: OFF
USE_GRAPH_EXECUTOR: ON
USE_ROCBLAS: OFF
GIT_COMMIT_HASH: be21b378b284eeab5b6a7721bf417ad7445fddf0
USE_VULKAN: ON
USE_RUST_EXT: OFF
USE_CUTLASS: ON
USE_CPP_RPC: OFF
USE_HEXAGON: OFF
USE_CUSTOM_LOGGING: OFF
USE_UMA: OFF
USE_FALLBACK_STL_MAP: OFF
USE_SORT: ON
USE_RTTI: ON
GIT_COMMIT_TIME: 2023-08-16 07:41:34 -0400
USE_HEXAGON_SDK: /path/to/sdk
USE_BLAS: none
USE_ETHOSN: OFF
USE_LIBTORCH: OFF
USE_RANDOM: ON
USE_CUDA: ON
USE_COREML: OFF
USE_AMX: OFF
BUILD_STATIC_RUNTIME: OFF
USE_CMSISNN: OFF
USE_KHRONOS_SPIRV: OFF
USE_CLML_GRAPH_EXECUTOR: OFF
USE_TFLITE: OFF
USE_HEXAGON_GTEST: /path/to/hexagon/gtest
PICOJSON_PATH: 3rdparty/picojson
USE_OPENCL_ENABLE_HOST_PTR: OFF
INSTALL_DEV: OFF
USE_PROFILER: ON
USE_NNPACK: OFF
LLVM_VERSION: 15.0.7
USE_OPENCL: OFF
COMPILER_RT_PATH: 3rdparty/compiler-rt
RANG_PATH: 3rdparty/rang/include
USE_SPIRV_KHR_INTEGER_DOT_PRODUCT: OFF
USE_OPENMP: OFF
USE_BNNS: OFF
USE_CUBLAS: OFF
USE_METAL: OFF
USE_MICRO_STANDALONE_RUNTIME: OFF
USE_HEXAGON_EXTERNAL_LIBS: OFF
USE_ALTERNATIVE_LINKER: AUTO
USE_BYODT_POSIT: OFF
USE_HEXAGON_RPC: OFF
USE_MICRO: OFF
DMLC_PATH: 3rdparty/dmlc-core/include
INDEX_DEFAULT_I64: ON
USE_RELAY_DEBUG: OFF
USE_RPC: ON
USE_TENSORFLOW_PATH: none
TVM_CLML_VERSION:
USE_MIOPEN: OFF
USE_ROCM: OFF
USE_PAPI: OFF
USE_CURAND: OFF
TVM_CXX_COMPILER_PATH: /opt/rh/gcc-toolset-11/root/usr/bin/c++
HIDE_PRIVATE_SYMBOLS: ON

junrushao commented 11 months ago

H100 is indeed supported, but you will have to disable cutlass for now using --no-cutlass-attn and --no-cutlass-norm

technillogue commented 11 months ago

is there any performance penalty from not using cutlass?

mlc-ai / mlc-llm