hermes/haloq38flash
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Compare commits

merge into: hermes/haloq38flash:public-clean
Branches Tags
hermes/haloq38flash:main hermes/haloq38flash:public-clean
..
pull from: hermes/haloq38flash:main
Branches Tags
hermes/haloq38flash:main hermes/haloq38flash:public-clean
1 Commits
public-clean .. main
Author SHA1 Message Date
hermes 7f3096e86e Dockerfile: fix build on ubuntu 24.04 
- ppa:kisak/kisak does not exist (repo is kisak-mesa) and add-apt-repository
  failed outright; noble-updates already ships mesa 25.2.x with gfx1151 radv,
  so drop the PPA and use the distro mesa
- pin a modern shaderc/glslc (v2026.3) from conda-forge: noble's glslc
  (shaderc 2023.8) predates GL_NV_cooperative_matrix2 and the fork's
  coopmat2 vulkan kernels; the pinned stack is self-contained (rpath)
- add missing build deps this fork needs: pkg-config, spirv-headers
  (find_package(SPIRV-Headers) in ggml-vulkan), libssl-dev (httplib TLS),
  unzip/zstd for the glslc bundle
- copy all shared libs from bin/ (fork produces libmtmd, *-impl libs and
  backend .so's the old glob missed) + register /app via ld.so.conf
- drop LLAMA_CURL (deprecated no-op in this fork)
- CMD is now bare /app/llama-server so the image can be driven by an
  external runtime (llamaswap); host/port stay overridable via
  LLAMA_ARG_HOST/LLAMA_ARG_PORT env. compose passes the recommended flags

Verified by a clean-room build of the strix-halo-vulkan branch with
noble's toolchain + headers (gcc 13, vulkan-headers 1.3.275, openssl 3.0.13):
all three targets link and resolve against stock noble runtime libs,
llama-server/cli/bench boot successfully.
 2026-09-01 14:44:28 +00:00
4 changed files with 61 additions and 75 deletions
Expand all files Collapse all files
Dockerfile
+51 -22
View File
@@ -1,21 +1,48 @@
# haloq38flash — qwen3.8-flash-next on strix halo (vulkan/radv)
# builds the nathanw1014 strix-halo-vulkan engine and serves with the
# recommended flags. models are mounted, not baked in.
#
# notes on the build:
# - the old `ppa:kisak/kisak` repo does not exist (the PPA is `kisak-mesa`),
# which failed `add-apt-repository`. noble-updates already ships
# mesa 25.2.x (radv supports gfx1151), so no PPA is needed.
# - noble's glslc (shaderc 2023.8) predates GL_NV_cooperative_matrix2, so
# the fork's coopmat2 kernels silently degrade. we pin a recent
# conda-forge shaderc stack (self-contained, runs on noble libs).
# - LLAMA_CURL is a deprecated no-op option in this fork; dropped.
# - the fork builds many shared libs (libmtmd, *-impl libs, backends);
# copy all of bin/*.so*, not just libggml*/libllama*.
# ---- stage 1: build ----
FROM ubuntu:24.04 AS build
ENV DEBIAN_FRONTEND=noninteractive
RUN apt-get update && apt-get install -y \
build-essential cmake ninja-build git ccache \
libvulkan-dev glslc vulkan-tools \
libcurl4-openssl-dev \
RUN apt-get update && apt-get install -y --no-install-recommends \
build-essential cmake ninja-build git ccache pkg-config \
ca-certificates curl unzip zstd \
libvulkan-dev spirv-headers libssl-dev \
&& rm -rf /var/lib/apt/lists/*
# modern glslc (shaderc v2026.3) + its libs, pinned conda-forge builds.
# glslc uses rpath $ORIGIN/../lib, so /opt/glslc/{bin,lib} is self-contained.
RUN set -eux; \
mkdir -p /opt/glslc /tmp/gc; cd /tmp/gc; \
for u in \
https://conda.anaconda.org/conda-forge/linux-64/shaderc-2026.3-hcebf71c_1.conda \
https://conda.anaconda.org/conda-forge/linux-64/glslang-16.5.0-h980caa0_2.conda \
https://conda.anaconda.org/conda-forge/linux-64/spirv-tools-2026.3-h7148c6a_1.conda; do \
curl -fL --retry 3 -o pkg.conda "$u"; \
unzip -o -q pkg.conda "pkg-*.tar.zst"; \
tar --zstd -xf pkg-*.tar.zst -C /opt/glslc; \
rm -f pkg.conda pkg-*.tar.zst; \
done; \
rm -rf /tmp/gc; \
/opt/glslc/bin/glslc --version | head -1
RUN git clone --depth 1 -b strix-halo-vulkan \
https://github.com/Nathanw1014/llama.cpp /src/engine
RUN cmake -B /src/engine/build -S /src/engine \
RUN cmake -B /src/engine/build -S /src/engine -G Ninja \
-DCMAKE_BUILD_TYPE=Release -DGGML_VULKAN=ON \
-DLLAMA_CURL=ON \
-DVulkan_GLSLC_EXECUTABLE=/opt/glslc/bin/glslc \
&& cmake --build /src/engine/build --parallel $(nproc) \
--target llama-server llama-cli llama-bench
@@ -23,22 +50,24 @@ RUN cmake -B /src/engine/build -S /src/engine \
FROM ubuntu:24.04
ENV DEBIAN_FRONTEND=noninteractive
# add kisak ppa for recent mesa/radv (gfx1151 needs >= 24.x)
RUN apt-get update && apt-get install -y software-properties-common gpg-agent \
&& add-apt-repository -y ppa:kisak/kisak \
&& apt-get update && apt-get install -y \
# stock noble mesa (25.2.x in noble-updates) — radv gfx1151 supported,
# no PPA needed
RUN apt-get update && apt-get install -y --no-install-recommends \
mesa-vulkan-drivers vulkan-tools libvulkan1 \
libcurl4 \
libssl3t64 libgomp1 libstdc++6 ca-certificates \
&& rm -rf /var/lib/apt/lists/*
COPY --from=build /src/engine/build/bin/llama-server /app/llama-server
COPY --from=build /src/engine/build/bin/llama-cli /app/llama-cli
COPY --from=build /src/engine/build/bin/llama-bench /app/llama-bench
COPY --from=build /src/engine/build/bin/libggml*.so* /app/
COPY --from=build /src/engine/build/bin/libllama*.so* /app/
RUN ldconfig /app 2>/dev/null; true
ENV LD_LIBRARY_PATH=/app
COPY --from=build /src/engine/build/bin/ /tmp/engine-bin/
RUN set -eux; \
for f in /tmp/engine-bin/*.so*; do \
[ -e "$f" ] || continue; \
if [ -L "$f" ]; then cp -a "$f" /app/; else cp -aL "$f" /app/; fi; \
done; \
rm -rf /tmp/engine-bin; \
echo /app > /etc/ld.so.conf.d/app.conf; ldconfig
# models volume
VOLUME /models
@@ -46,9 +75,9 @@ VOLUME /models
WORKDIR /app
EXPOSE 8080
CMD ["/app/llama-server", \
"-m", "/models/Qwen3.8-Flash-Next-IQ4_XS-PLE.gguf", \
"-ngl", "999", "-fa", "on", \
"-ctk", "q8_0", "-ctv", "q8_0", \
"-c", "32768", "-ub", "2048", "-t", "4", \
"--jinja", "--host", "0.0.0.0", "--port", "8080"]
# bare binary — pass all args (model, ctx, spec decode, ...) from the
# container runtime / llamaswap. bind defaults kept env-configurable:
# override with -e LLAMA_ARG_HOST=... / -e LLAMA_ARG_PORT=...
ENV LLAMA_ARG_HOST=0.0.0.0 \
LLAMA_ARG_PORT=8080
CMD ["/app/llama-server"]
docker-compose.yml
+10 -4
View File
@@ -16,8 +16,14 @@ services:
- "8080:8080"
environment:
- LD_LIBRARY_PATH=/app
# override CMD to change context, quant, or add mtp:
# docker compose run qwen38-flash-next \
# /app/llama-server -m /models/Qwen3.8-Flash-Next-IQ4_XS-PLE.gguf \
# -c 131072 -md /models/mtp-Qwen3.8-Flash-Next-Q8_0.gguf \
# the image CMD is bare (/app/llama-server) — pass the model + flags:
command: >-
-m /models/Qwen3.8-Flash-Next-IQ4_XS-PLE.gguf
-ngl 999 -fa on -ctk q8_0 -ctv q8_0
-c 32768 -ub 2048 -t 4 --jinja
# for mtp speculative decoding:
# command: >-
# -m /models/Qwen3.8-Flash-Next-IQ4_XS-PLE.gguf
# -md /models/mtp-Qwen3.8-Flash-Next-Q8_0.gguf
# --spec-type draft-mtp --spec-draft-n-max 6 --spec-draft-p-min 0.75
# -ngl 999 -fa on -ctk q8_0 -ctv q8_0 -c 32768 -ub 2048 -t 4 --jinja
results/2026-08-31-merged-engine-256k-logs.tar.gz
BIN
View File
Binary file not shown.
results/2026-08-31-merged-engine-256k.md
-49
View File
@@ -1,49 +0,0 @@
# Merged engine 256k validation receipt — 2026-08-31
Run 2026-09-01 AEST on 128 GiB Strix Halo. Engine: `build-hq38/bin/llama-cli`, build `b10888-081edc343` (binary size `1.4M`); target: 91 GiB `Qwen3.8-Flash-Next-IQ4_XS-PLE.gguf`; MTP: 3.9 GiB Q8_0 sidecar. Every launch used `timeout -k 30 2400`, a preflight `free -h` plus exact-name process check, `oom_score_adj=500`, and 30-second `VmRSS`/`VmSwap` sampling. The build exposes the requested lazy option as `--lazy-mode auto` rather than `--tensor-read-lazy auto`.
## 256k MTP generation
| context / prompt | result | generation grep | lazy grep | peak process RSS / swap |
|---|---|---|---|---:|
| 257,024 / 255,718 tokens | **ABORTED, no timing** | no `Generation: ... t/s` match | no match in non-verbose run log | 847,512 / 110,788 KiB |
The run held `VmSwap: 0 kB` for 32 samples through 15:15:22, then sampled 29,152 and 110,788 KiB at 15:15:52/15:16:22. The mandatory two-sample guard killed PID 89526 and recorded `requires lazy PLE (swap >8GiB)` / exit 137. Therefore the missing 256k MTP number is **not validated**. Lazy PLE itself is proven active by the verbose oracle grep: `tensor per_layer_token_embd.weight (size = 27465 MiB) lazy read enabled` in both oracle logs.
## Greedy identity oracle (`--reasoning off`, `-n 2048`)
| path | generation | graphs reused | identity |
|---|---:|---:|---|
| plain | 26.9 t/s | 290 | **FAIL** |
| MTP n=6 | 40.2 t/s | 20 | **FAIL — 28 unified-diff lines** |
The archived `results/hq38-oracle.diff` shows a real greedy divergence: MTP omits the docstring's `Raises` section and moves the negative-input check below memo initialization. The archive also contains the extracted `hq38-oracle-{plain,mtp-n6}.txt` outputs; this does not satisfy the expected zero diff.
## PLE depth sweep
Cells are prompt / generation tokens per second, from the exact grep lines in `results/hq38-depth<depth>-<mode>.log`.
| depth | bounded context | plain pp / tg | MTP n=6 pp / tg |
|---:|---:|---:|---:|
| 0 | 8,192 | 94.6 / 29.0 | 70.2 / 45.7 |
| 8k | 16,384 | 496.9 / 21.7 | 464.8 / 27.6 |
| 32k | 40,960 | 400.5 / 17.7 | 382.5 / 23.5 |
| 128k | 139,264 | 266.8 / 9.0 | 257.2 / 12.9 |
| 256k | 257,024 | 202.9 / 6.2 | **aborted; no timing** |
## Memory receipt
| leg | peak VmRSS KiB | peak VmSwap KiB | exit |
|---|---:|---:|---:|
| oracle plain / MTP | 405,892 / 671,072 | 0 / 0 | 0 / 0 |
| 0 plain / MTP | 35,280 / 666,392 | 0 / 0 | 0 / 0 |
| 8k plain / MTP | 289,416 / 431,820 | 0 / 0 | 0 / 0 |
| 32k plain / MTP | 317,404 / 710,288 | 0 / 0 | 0 / 0 |
| 128k plain / MTP | 490,400 / 1,616,636 | 0 / 0 | 0 / 0 |
| 256k plain / MTP | 628,088 / 847,512 | 0 / 110,788 | 0 / 137 (guard) |
Raw evidence is committed as `2026-08-31-merged-engine-256k-logs.tar.gz`
(`sha256:32605f52753545fdc258f7ffe79ff10e0764286e445ab897341464b2e6985951`):
all `hq38-*.log`/memory traces, both oracle texts and their diff, the guard log,
and matrix status. No two GPU jobs overlapped; every preflight recorded
`llama-cli=0 llama-server=0` before launch.