tvm
tvm copied to clipboard
apache
Open deep learning compiler stack for cpu, gpu and specialized accelerators
### Actual behavior ``` Traceback (most recent call last): File "test_simple.py", line 46, in ex = relax.build(mod, target='llvm') # crash here! ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/software/tvm/python/tvm/relax/vm_build.py", line 335, in build mod =...
Hi all, I set `check_well_formed=True` in the below Relax IR construction and can run `mod.show()` to show the IR successfully. It seems the Relax IR passed the legitimacy checking. However,...
### Expected behavior I wanted to port MicroTVM as easily as possible on a cortex-m4 (float)-based stm32g474 to test the inference speed of the MCU that was not optimized. Simplest...
Prior to this commit, the `relax.transform.AttachGlobalSymbol` pass could produce a PrimFunc whose `"global_symbol"` attribute does not match the name of the `GlobalVar`. As a result, the PackedFunc that is provided...
Prior to this commit, the `relax.transform.MergeCompositeFunctions` required the module to contain a single Relax function named `"main"`. Lack of this function would result in an error when `mod->GetGlobalVar("main")` was called....
Prior to this commit, the `relax.transform.RealizeVDevice` pass performed in-place update on expressions appearing in its input `IRModule`, overwriting their struct info. In-place mutation of TVM's IR types is only legal...
### Actual behavior ``` Traceback (most recent call last): File "/share_container/optfuzz/res/bugs/3_simple.py", line 36, in mod = relax.transform.MergeCompositeFunctions()(mod) # crash here ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/software/tvm-lunder/python/tvm/ir/transform.py", line 238, in __call__ return _ffi_transform_api.RunPass(self, mod)...
Prior to this commit, the well-formed checker verified that the `"global_symbol"` attribute, if present, matches the name of the `GlobalVar`. However, this check was only applied for Relax functions. As...
### Expected behavior support conv1d QLinearConv,MatMulIntegerToFloat in onnx frontend ### Actual behavior 1. The following operators are not supported for frontend ONNX: MatMulIntegerToFloat 2. Only 2D kernels are supported for...
When building hexagon_launcher as the guide: https://github.com/apache/tvm/tree/main/apps/hexagon_launcher，there are some errors: ``` tvm/apps/hexagon_launcher/cmake/hexagon/../../../../include/tvm/runtime/container/array.h:661:76: error: no member named 'invoke_result_t' in namespace 'std' template ~~~~~^ tvm/apps/hexagon_launcher/cmake/hexagon/../../../../include/tvm/runtime/container/array.h:661:92: error: 'F' does not refer to a...
