Skip to content

Releases: adromir/llama-cpp-turboquant

Release list

llama.cpp ROCm v0.0.4-vanilla [Vanilla TurboQuant]

llama.cpp ROCm v0.0.4-vanilla [Vanilla TurboQuant] Latest
Latest

Choose a tag to compare

View all tags
@github-actions github-actions released this 10 Oct 23:21
v0.0.4-vanilla
1ebeee7

Release Information

  • Flavor: Vanilla TurboQuant
  • Source Branch: custom-workflow
  • Commit: 1ebeee73c5bf6976bda288d8a4c1b41dc2a98d8d
  • ROCm Core SDK: 10.1.0 (TheRock)
  • Target GPU Architectures: gfx900;gfx906;gfx908;gfx90a;gfx1030;gfx1100;gfx1101;gfx1102;gfx1200;gfx1201
  • CPU Target: Standard AVX2 (Universal x86-64 compatibility)

Description

Upstream llama.cpp base with TurboQuant KV cache compression.

Unified ROCm 10.1.0 (TheRock) release built for Windows and Linux.
Packages include required AMD ROCm runtime libraries (rocblas, hipblaslt, hipblas, amdhip64) for portable out-of-the-box execution.

Assets 4
Loading

llama.cpp ROCm v0.0.4-experimental-avx512 [Experimental RDNA Boosts (Specialized AVX-512)]

llama.cpp ROCm v0.0.4-experimental-avx512 [Experimental RDNA Boosts (Specialized AVX-512)] Pre-release
Pre-release

Choose a tag to compare

View all tags
@github-actions github-actions released this 10 Oct 23:21
v0.0.4-experimental-avx512
1ebeee7

Release Information

  • Flavor: Experimental RDNA Boosts (Specialized AVX-512)
  • Source Branch: experiment/rdna-boosts
  • Commit: ce7417cf78cb0e198c38f3acdeaf4dadd19ce9cc
  • ROCm Core SDK: 10.1.0 (TheRock)
  • Target GPU Architectures: gfx900;gfx906;gfx908;gfx90a;gfx1030;gfx1100;gfx1101;gfx1102;gfx1200;gfx1201
  • CPU Target: Specialized AVX-512 (Zen 4/5, modern Intel - high-speed hybrid offloading)

Description

TurboQuant KV cache + Stew's RDNA Boosts (native-BF16 FA, WMMA, fused GDN, fused MoE) + ROCmFPX quantization family + DFlash2 speculative decoding, compiled with specialized native AVX-512 vector extensions (AVX512F, VBMI, VNNI, BF16) for maximum CPU and hybrid offload performance (e.g. Qwen 3.8 Flash-Next 51B, 70B+ models) on AMD Zen 4 / Zen 5 and modern Intel CPUs.

Unified ROCm 10.1.0 (TheRock) release built for Windows and Linux.
Packages include required AMD ROCm runtime libraries (rocblas, hipblaslt, hipblas, amdhip64) for portable out-of-the-box execution.

  • Windows Scripts: llama-rocm-experimental-scripts-windows.zip (PowerShell quantization scripts for Windows)
  • Linux Scripts: llama-rocm-experimental-scripts-linux.zip (Bash quantization scripts for Linux)
Loading

llama.cpp ROCm v0.0.4-experimental [Experimental RDNA Boosts (Universal AVX2)]

llama.cpp ROCm v0.0.4-experimental [Experimental RDNA Boosts (Universal AVX2)] Pre-release
Pre-release

Choose a tag to compare

View all tags
@github-actions github-actions released this 10 Oct 23:20
v0.0.4-experimental
1ebeee7

Release Information

  • Flavor: Experimental RDNA Boosts (Universal AVX2)
  • Source Branch: experiment/rdna-boosts
  • Commit: ce7417cf78cb0e198c38f3acdeaf4dadd19ce9cc
  • ROCm Core SDK: 10.1.0 (TheRock)
  • Target GPU Architectures: gfx900;gfx906;gfx908;gfx90a;gfx1030;gfx1100;gfx1101;gfx1102;gfx1200;gfx1201
  • CPU Target: Standard AVX2 (Universal x86-64 compatibility)

Description

TurboQuant KV cache + Stew's RDNA Boosts (native-BF16 FA, WMMA, fused GDN, fused MoE) + ROCmFPX quantization family + DFlash2 speculative decoding. Universal AVX2 build compatible with all modern x86-64 CPUs.

Unified ROCm 10.1.0 (TheRock) release built for Windows and Linux.
Packages include required AMD ROCm runtime libraries (rocblas, hipblaslt, hipblas, amdhip64) for portable out-of-the-box execution.

  • Windows Scripts: llama-rocm-experimental-scripts-windows.zip (PowerShell quantization scripts for Windows)
  • Linux Scripts: llama-rocm-experimental-scripts-linux.zip (Bash quantization scripts for Linux)
Loading

llama.cpp ROCm v0.0.3-vanilla [Vanilla TurboQuant]

llama.cpp ROCm v0.0.3-vanilla [Vanilla TurboQuant]

Choose a tag to compare

View all tags
@github-actions github-actions released this 26 Sep 18:50
v0.0.3-vanilla
57f0c9f

Release Information

  • Flavor: Vanilla TurboQuant
  • Source Branch: custom-workflow
  • Commit: 57f0c9f7a3e67119939f77ffc310f354c4febf6a
  • ROCm Core SDK: 10.0.0 (TheRock)
  • Target GPU Architectures: gfx900;gfx906;gfx908;gfx90a;gfx1030;gfx1100;gfx1101;gfx1102;gfx1200;gfx1201

Description

Upstream llama.cpp base with TurboQuant KV cache compression.

Unified ROCm 10.0.0 (TheRock) release built for Windows and Linux.
Packages include required AMD ROCm runtime libraries (rocblas, hipblaslt, hipblas, amdhip64) for portable out-of-the-box execution.

Loading

llama.cpp ROCm v0.0.3-experimental [Experimental RDNA Boosts]

llama.cpp ROCm v0.0.3-experimental [Experimental RDNA Boosts] Pre-release
Pre-release

Choose a tag to compare

View all tags
@github-actions github-actions released this 26 Sep 18:49
v0.0.3-experimental
57f0c9f

Release Information

  • Flavor: Experimental RDNA Boosts
  • Source Branch: experiment/rdna-boosts
  • Commit: f699ca98a524df07f853f732f6940e67fe3b13e3
  • ROCm Core SDK: 10.0.0 (TheRock)
  • Target GPU Architectures: gfx900;gfx906;gfx908;gfx90a;gfx1030;gfx1100;gfx1101;gfx1102;gfx1200;gfx1201

Description

TurboQuant KV cache + Stew's RDNA Boosts (native-BF16 FA, WMMA, fused GDN, fused MoE) + ROCmFPX quantization family + DFlash2 speculative decoding.

Unified ROCm 10.0.0 (TheRock) release built for Windows and Linux.
Packages include required AMD ROCm runtime libraries (rocblas, hipblaslt, hipblas, amdhip64) for portable out-of-the-box execution.

  • Windows Scripts: llama-rocm-experimental-scripts-windows.zip (PowerShell quantization scripts for Windows)
  • Linux Scripts: llama-rocm-experimental-scripts-linux.zip (Bash quantization scripts for Linux)
Loading

llama.cpp ROCm v0.0.2-vanilla [Vanilla TurboQuant]

llama.cpp ROCm v0.0.2-vanilla [Vanilla TurboQuant]

Choose a tag to compare

View all tags
@github-actions github-actions released this 08 Sep 06:16
v0.0.2-vanilla
55776c9

Release Information

  • Flavor: Vanilla TurboQuant
  • Source Branch: custom-workflow
  • Commit: 55776c968d3b87fabd2cfc7769b4575f87be8a1f
  • ROCm Core SDK: 10.0.0 (TheRock)
  • Target GPU Architectures: gfx900;gfx906;gfx908;gfx90a;gfx1030;gfx1100;gfx1101;gfx1102;gfx1200;gfx1201

Description

Upstream llama.cpp base with TurboQuant KV cache compression.

Unified ROCm 10.0.0 (TheRock) release built for Windows and Linux.
Packages include required AMD ROCm runtime libraries (rocblas, hipblaslt, hipblas, amdhip64) for portable out-of-the-box execution.

Loading

llama.cpp ROCm v0.0.2-experimental [Experimental RDNA Boosts]

llama.cpp ROCm v0.0.2-experimental [Experimental RDNA Boosts] Pre-release
Pre-release

Choose a tag to compare

View all tags
@github-actions github-actions released this 08 Sep 06:16
v0.0.2-experimental
55776c9

Release Information

  • Flavor: Experimental RDNA Boosts
  • Source Branch: experiment/rdna-boosts
  • Commit: 6b05c7e38228600dd467f2bcd9dbb69d21594c24
  • ROCm Core SDK: 10.0.0 (TheRock)
  • Target GPU Architectures: gfx900;gfx906;gfx908;gfx90a;gfx1030;gfx1100;gfx1101;gfx1102;gfx1200;gfx1201

Description

TurboQuant KV cache + Stew's RDNA Boosts (native-BF16 FA, WMMA, fused GDN, fused MoE) + ROCmFPX quantization family + DFlash2 speculative decoding.

Unified ROCm 10.0.0 (TheRock) release built for Windows and Linux.
Packages include required AMD ROCm runtime libraries (rocblas, hipblaslt, hipblas, amdhip64) for portable out-of-the-box execution.

  • Windows Scripts: llama-rocm-experimental-scripts-windows.zip (PowerShell quantization scripts for Windows)
  • Linux Scripts: llama-rocm-experimental-scripts-linux.zip (Bash quantization scripts for Linux)
Loading

llama.cpp ROCm v0.0.1-vanilla [Vanilla TurboQuant]

llama.cpp ROCm v0.0.1-vanilla [Vanilla TurboQuant]

Choose a tag to compare

View all tags
@github-actions github-actions released this 04 Sep 17:44
v0.0.1-vanilla
04dc4df

Release Information

  • Flavor: Vanilla TurboQuant
  • Source Branch: custom-workflow
  • Commit: 04dc4df136352144dfed23de72574d398fc3a1c9
  • ROCm Core SDK: 10.0.0 (TheRock)
  • Target GPU Architectures: gfx900;gfx906;gfx908;gfx90a;gfx1030;gfx1100;gfx1101;gfx1102;gfx1200;gfx1201

Description

Upstream llama.cpp base with TurboQuant KV cache compression.

Unified ROCm 10.0.0 (TheRock) release built for Windows and Linux.
Packages include required AMD ROCm runtime libraries (rocblas, hipblaslt, hipblas, amdhip64) for portable out-of-the-box execution.

Loading

llama.cpp ROCm v0.0.1-experimental [Experimental RDNA Boosts]

llama.cpp ROCm v0.0.1-experimental [Experimental RDNA Boosts] Pre-release
Pre-release

Choose a tag to compare

View all tags
@github-actions github-actions released this 04 Sep 17:44
v0.0.1-experimental
04dc4df

Release Information

  • Flavor: Experimental RDNA Boosts
  • Source Branch: experiment/rdna-boosts
  • Commit: 277ea9bd0fef76c6f366689f5c723e9711da7a3b
  • ROCm Core SDK: 10.0.0 (TheRock)
  • Target GPU Architectures: gfx900;gfx906;gfx908;gfx90a;gfx1030;gfx1100;gfx1101;gfx1102;gfx1200;gfx1201

Description

TurboQuant KV cache + Stew's RDNA Boosts (native-BF16 FA, WMMA, fused GDN, fused MoE) + ROCmFPX quantization family + DFlash2 speculative decoding.

Unified ROCm 10.0.0 (TheRock) release built for Windows and Linux.
Packages include required AMD ROCm runtime libraries (rocblas, hipblaslt, hipblas, amdhip64) for portable out-of-the-box execution.

  • Windows Scripts: llama-rocm-experimental-scripts-windows.zip (PowerShell quantization scripts for Windows)
  • Linux Scripts: llama-rocm-experimental-scripts-linux.zip (Bash quantization scripts for Linux)
Loading