Spaces:
Paused
Paused
fix: compilar llama-cpp-python sin AVX2/FMA para evitar SIGILL en runtime
Browse files- Dockerfile +2 -1
Dockerfile
CHANGED
|
@@ -16,7 +16,8 @@ COPY requirements.txt .
|
|
| 16 |
RUN pip install --no-cache-dir --upgrade pip \
|
| 17 |
&& pip install --no-cache-dir numpy==1.26.0 \
|
| 18 |
&& pip install --no-cache-dir torch torchvision --index-url https://download.pytorch.org/whl/cpu \
|
| 19 |
-
&&
|
|
|
|
| 20 |
&& pip install --no-cache-dir -r requirements.txt
|
| 21 |
|
| 22 |
# Pre-descarga del modelo Qwen GGUF en build (arranque rápido en runtime)
|
|
|
|
| 16 |
RUN pip install --no-cache-dir --upgrade pip \
|
| 17 |
&& pip install --no-cache-dir numpy==1.26.0 \
|
| 18 |
&& pip install --no-cache-dir torch torchvision --index-url https://download.pytorch.org/whl/cpu \
|
| 19 |
+
&& CMAKE_ARGS="-DLLAMA_NATIVE=OFF -DLLAMA_AVX2=OFF -DLLAMA_AVX512=OFF -DLLAMA_FMA=OFF -DLLAMA_F16C=OFF" \
|
| 20 |
+
pip install --no-cache-dir llama-cpp-python==0.3.2 \
|
| 21 |
&& pip install --no-cache-dir -r requirements.txt
|
| 22 |
|
| 23 |
# Pre-descarga del modelo Qwen GGUF en build (arranque rápido en runtime)
|