tvm
tvm copied to clipboard
apache
Open deep learning compiler stack for cpu, gpu and specialized accelerators
As discussed in issue #17375, the current rule for assigning touched buffers is not appropriate. Consider the following example: ```python code_block_0 for k in range(0, 10): # (the gen point...
In our current lowering pipeline, ThreadSync is placed before the MergeSharedMemoryAllocations Pass, which may lead to unknown behaviors because MergeSharedMemoryAllocations will modify the buffer access region. https://github.com/apache/tvm/blob/main/src/driver/driver_api.cc#L585-L613 ```cpp bool detect_global_barrier...
### Expected behavior A successful compilation, or messages about what part of the program being compiled fails. As far as I can tell this _should_ at least surface a more...
Thanks for participating in the TVM community! We use https://discuss.tvm.ai for any general usage questions and discussions. The issue tracker is used for actionable items such as feature proposals discussion,...
The current implementation does not support bit-serial operators for AArch64 processor architecture. I added the functions in the folder python/tvm/topi/arm_cpu and adjusted the functions bitserial_conv2d_strategy_arm_cpu and schedule_bitserial_dense_arm_cpu in the folder...
I'm encountering a segmentation fault when running an AOT-compiled TVM model for inference. The crash occurs within the TVM-generated code, specifically in the `tvmgen_default___tvm_main___compute_` function. - Build PyTorch Model ```python...
I use TVM to refine my onnx model that contains max tensor dimension is < 4, but when i check the new onnx model output from TVM, i found there...
### Expected behavior build work no matter build folder name ### Actual behavior build fail when build directory contains space shows: ``` [main] 正在生成文件夹: /home/user1/doc/a/apache-tvm-src-v0.17.0.rc0/build-Release-GCC 9.3.1 [build] 正在启动生成 [proc] 执行命令:...
### Actual behavior ``` Traceback (most recent call last): File "/share_container/optfuzz/res/bugs/llm.py", line 35, in ex = relax.build(mod, target='llvm') ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/software/tvm-latest/python/tvm/relax/vm_build.py", line 335, in build mod = pipeline(mod) ^^^^^^^^^^^^^ File...
I encountered an issue when trying to build a TIR-based module. The following error occurs when using `T.match_buffer` for a scalar element: ``` File "/software/tvm/src/tir/transforms/lower_match_buffer.cc", line 222 TVMError: Check failed:...
