-
This issue happens on the latest test build release of [koboldcpp-rocm](https://github.com/YellowRoseCx/koboldcpp-rocm)
Version: [KoboldCPP-v1.56.yr1-ROCm](https://github.com/YellowRoseCx/koboldcpp…
-
### Component
Dasharo firmware
### Device
NovaCustom V54 14th Gen
### Dasharo version
v0.9.1-rc5
### Dasharo Tools Suite version
--
### Test case ID
CPU003.001
### Brief summary
lspcu shows…
-
This issue is a placeholder for future discussion about supporting 4-dimensional-reducing dot-product instructions taking 8bit inputs and accumulating into 32bit, i.e.
```
int32_accumulator += int…
-
We need to present more useful hardware environments about the benchmark runner, like `cat /proc/cpuinfo`, in [Equinix bare metal](https://github.com/open-telemetry/community/blob/main/assets.md#equin…
-
### Describe the bug
To finetune model on Xeon CPU, we are following the [ai-reference-models/models_v2/pytorch/llama/training/cpu at main · intel/ai-reference-models (github.com)](https://github.com…
-
This is mainly because we are not using VNNI in codegen. The test case is extracted from MobileBert_int8 model.
To repro,
1. Compile the mmt4d kernel with codegen, run `iree-compile --output-for…
-
**Describe the Issue**
Mistral/Nvidia recently released [Nemo 12B](https://mistral.ai/news/mistral-nemo/) and llama.cpp have [added support](https://github.com/ggerganov/llama.cpp/pull/8579) for its …
-
Hi, as the Ryzen AI 9 processors released, can you add support for them?
Mine is an ASUS Vivobook S 16 with HX 365,
`CPU0: AMD Ryzen AI 9 365 w/ Radeon 880M (family: 0x1a, model: 0x24, stepping: 0…
-
I optimized the initial AVX 512 GEMM kernel based on what works best on my 2020 Intel MacBook Pro i5. This is an Ice Lake client architecture system which has a single 512-bit FMA unit. When testing o…
-
Migrate the Caffe2/MKL-DNN int8 operation to support Aten/JIT backend and align with Qint8 direction in Pytorch/Aten
Motivation
With Cascadelake/VNNI, MKL-DNN int8 functions can speedup DL m…