tvm
tvm copied to clipboard
apache
Open deep learning compiler stack for cpu, gpu and specialized accelerators
### Expected behaviour: No output mismatch. ### Actual behaviour: When compiled on an `arm_cpu` target, the model results in an output mismatch: ``` E AssertionError: E Not equal to tolerance...
### Expected behaviour: When an `arm_cpu` target is used, the grouped convolution should compile successfully without an error. ### Actual behaviour: When an `arm_cpu` target is used, the model fails...
### Expected behaviour: When an `arm_cpu` target is used, the model should compile successfully without an error. ### Actual behaviour: When an `arm_cpu` target is used, the model fails to...
Add dcnV2 support according to this [repository](https://github.com/xi11xi19/CenterNet2TorchScript/tree/master/DCNv2)
- Add converter for torch2.0 ExportedProgram
- If a Graph executor model have too many node, `tvmc run` with `--profile` option will hang, and if you trace its OpenCL call, it shows CL_OUT_OF_HOST_MEMORY error on the...
Add support of `sum` operator in Keras lambdas. cc: @Hzfengsy
For the LSTM below, when `batch_size != 1` (i.e., the size of first dimension input), compiling will lead to an unexpected crash and throw `Check failed: *axis_ptr == 1 (2...
The commit contains new memory manager which limits amount of cached buffers. The problem appeared on scenario with 4 sequential transformer based networks which generate output with different sizes on...
### This issue is to track progress for SVE upstreaming - [x] P1. `vscale` builtin and its support in LLVM codegen - https://github.com/apache/tvm/pull/16484 - [x] P2. Support for scalable lanes...
