-
The MacOS release-mode gcc-14 aarch64 CI started failing after this commit: https://github.com/oneapi-src/oneDNN/commit/55b52c67967b61d439c84df8b34cfb006bf528eb
The log can be seen [here](https://g…
-
### Issue type
Build/Install
### Have you reproduced the bug with TensorFlow Nightly?
No
### Source
source
### TensorFlow version
tf v2.17.0
### Custom code
No
### OS pla…
-
1、双击安装ipex-llm-ollama-installer-20240918.exe
默认安装在C:\Users\OPS17\ipex-llm-ollama
2、拷贝ipex-llm-ollama文件夹到C:\aipc\
3、这个时候运行ipex-llm-ollama里面的start.bat，可以正常运行，并且访问什么都正常
4、这个时候用Python编写一个程序，并打包为…
-
### Describe the issue
First i use commond '--use dnnl' to build a onnxruntime.dll enable oneDNN support.
When i use this code to register a dnnl EP to ORT：
`int threadNum = 4;`
`bool enable_cpu_m…
-
- [x] gtest for integration tests - these are built by enabling `-DGC_TEST_ENABLE=on`. Currently, there's only one built now, run as `cd test/dnnl/ && ./test_dnnl_c_interface`. I think it's missing a …
-
```
Caused by: java.lang.UnsatisfiedLinkError: Can't load library: /Users/booster/.javacpp/cache/dnnl-3.3.4-1.5.10-macosx-x86_64.jar/org/bytedeco/dnnl/macosx-x86_64/libjnidnnl.dylib
```
Caused by…
-
Hello,
I found that dnnl::matmul is slow when they are not using SGEMM.
For example, if I assigned src/weights/bias with 50x50 matrix, it takes 760 microseconds and 100x100 is going to be 5ms.
Ho…
-
### Describe the bug
Original issue: https://github.com/oneapi-src/oneDNN/issues/1611
I am compiling pytorch from source according to the instructions from
https://intel.github.io/intel-extens…
-
# Summary
During our pytorch development, we found Windows system memory alloctor is worse performance, and slow down the whole pytorch performance. After add third party memory alloctor, pytorch imp…
-
**Idea:**
Cast FP32/FP16 to BF16.
Casting will be different based on type:
- FP32 to BF16: truncate last 16 bits from mantissa, exponent stays the same
- FP16 to BF16: more involved process --…