Speech recognition models running on the Ryzen AI NPU with Windows ML.
🔄 In a Training Loop
Von
Harsha001
AI & ML interests
ROCmFPX GGUF quants for AMD GPUs and Strix Halo. Every file is scored by KLD against BF16. Speech models ported to the AMD XDNA2 NPU.
Recent Activity
updated a model 10 minutes ago
Harsha001/DeepSeek-V4.1-Flash-1017-StrixHalo-GGUF published a model 10 minutes ago
Harsha001/DeepSeek-V4.1-Flash-1017-StrixHalo-GGUF new activity about 23 hours ago
Harsha001/Nanbeige4.3-33B-ROCmFPX:DeepSeek V4.1 Flash 1017 (GSQ-RCO)Organizations
None yet
Low-bit GGUF files for AMD GPUs (Vulkan and HIP) with the ROCmFPX llama.cpp fork. Every file is measured against BF16.
-
Harsha001/Qwen3.8-27B-Uncensored-ROCmFPX
Image-Text-to-Text • 27B • Updated • 243 • 1 -
Harsha001/NVIDIA-Nemotron-3.5-Bolt-E4B-ROCmFPX
Text Generation • 8B • Updated • 224 -
Harsha001/flint-ROCmFPX
Text Generation • 40B • Updated • 396 • 1 -
Harsha001/Carbon-X-ROCmFPX
Text Generation • 15B • Updated • 184 • 1
Speech recognition models running on the Ryzen AI NPU with Windows ML.
Image and video GGUF files for stable-diffusion.cpp with ROCmFPX weight types, for AMD GPUs.
Low-bit GGUF files for AMD GPUs (Vulkan and HIP) with the ROCmFPX llama.cpp fork. Every file is measured against BF16.
-
Harsha001/Qwen3.8-27B-Uncensored-ROCmFPX
Image-Text-to-Text • 27B • Updated • 243 • 1 -
Harsha001/NVIDIA-Nemotron-3.5-Bolt-E4B-ROCmFPX
Text Generation • 8B • Updated • 224 -
Harsha001/flint-ROCmFPX
Text Generation • 40B • Updated • 396 • 1 -
Harsha001/Carbon-X-ROCmFPX
Text Generation • 15B • Updated • 184 • 1