-
### What happened?
Hitting this assert for a `tensor.insert_slice` op
```iree-compile: iree/third_party/llvm-project/llvm/include/llvm/ADT/ArrayRef.h:195: llvm::ArrayRef llvm::ArrayRef::slice(size…
-
### What happened?
In `iree-run-module`, we can check accuracy by providing `expected_output`. Currently it only provides thresholds for F32, F16 and F64 (via `--expected_f32_threshold, --expected_…
-
### What happened?
Several of our tests are failing to compile using the WebGPU compiler target, during compilation from SPIR-V to WGSL using Tint.
From [`tests/e2e/xla_ops/`](https://github.com…
-
Today we are generating native object `.o` files (x86/arm/etc) that we then pass out to the linker (lld/ld/link) - that means we need to be able to generate those native object files with the IREE bin…
-
### Request description
I'm exporting several JAX functions through IREE so I can run them via the IREE runtime, using this function
```
def aot(function, static_argnums, *args, **options):
""…
-
### What happened?
LLaMA/Vicuna 7B fp16 model is producing all zero results on CPU for both local-sync and local-task.
### Steps to reproduce your issue
1. Download the .mlir from https://sto…
-
IREE still lowers `tensor.pad` into a `linalg.fill` followed by `tensor.insert_slice`. This is really the poor-mans handling of `tensor.pad`. There are three ways in which the `tensor.pad` handling ca…
-
### What happened?
I am using this docker image to compile iree: `build_tools/docker/dockerfiles/base.Dockerfile`.
When I try to build the project from the root of the image, I get the following…
-
I have Win10 and Radeon RX 570, I did everything the guide says, tried setting up virtual environment as well as launching from shark .exe and still get this error, where did I miss?
-