From 852481ea739a7d9dfddcbf8a34f6869c6d6bf8fb Mon Sep 17 00:00:00 2001 From: Luca Sacchi Ricciardi Date: Fri, 27 Mar 2026 14:48:52 +0000 Subject: [PATCH] fix docker-compose and Dockerfile --- Dockerfile | 5 ++--- docker-compose.yml | 2 +- 2 files changed, 3 insertions(+), 4 deletions(-) diff --git a/Dockerfile b/Dockerfile index 3235d91..6f9b834 100644 --- a/Dockerfile +++ b/Dockerfile @@ -5,20 +5,19 @@ ENV HSA_OVERRIDE_GFX_VERSION="11.0.0" ENV AMDGPU_TARGETS="gfx1100" ENV HIP_VISIBLE_DEVICES="0" +# Aggiunti hipblas-dev e rocblas-dev necessari per la compilazione GGML_HIP RUN apt-get update && apt-get install -y \ build-essential cmake git libcurl4-openssl-dev \ + hipblas-dev rocblas-dev \ && rm -rf /var/lib/apt/lists/* WORKDIR /app -# Little often: cloniamo e compiliamo RUN git clone https://github.com/ggerganov/llama.cpp.git . -# Compilazione con supporto HIP nativo per AMD RUN cmake -B build -DGGML_HIP=ON -DAMDGPU_TARGETS=gfx1100 -DCMAKE_BUILD_TYPE=Release RUN cmake --build build --config Release -j $(nproc) EXPOSE 8000 -# Punto d'ingresso nudo e crudo ENTRYPOINT ["/app/build/bin/llama-server"] diff --git a/docker-compose.yml b/docker-compose.yml index 05a50df..98d16b9 100644 --- a/docker-compose.yml +++ b/docker-compose.yml @@ -12,7 +12,7 @@ services: # -ngl 99 offloada tutti i layer sulla GPU AMD # --host 0.0.0.0 lo rende accessibile fuori dal container command: > - -m /models/qwen1_5-0_8b-chat-q8_0.gguf + -m /models/Qwen3.5-0.8B-UD-Q8_K_XL.gguf --host 0.0.0.0 --port 8000 -ngl 99