Spaces:

KoboldAI
/

Koboldcpp-Tiefighter

Running on T4

Henk717 commited on Nov 6, 2023

Commit

ec210a9

•

1 Parent(s): 396d27a

Disable MMQ

Files changed (1) hide show

Dockerfile CHANGED Viewed

@@ -8,5 +8,5 @@ RUN git clone https://github.com/lostruins/koboldcpp /opt/koboldcpp
 WORKDIR /opt/koboldcpp
 RUN make LLAMA_OPENBLAS=1 LLAMA_CUBLAS=1 LLAMA_PORTABLE=1
 RUN wget -O model.ggml $MODEL
-CMD /bin/python3 ./koboldcpp.py --model model.ggml --usecublas mmq --gpulayers 99 --multiuser --contextsize $CONTEXT_SIZE --port 7860 --hordeconfig $MODEL_NAME 1 1

 WORKDIR /opt/koboldcpp
 RUN make LLAMA_OPENBLAS=1 LLAMA_CUBLAS=1 LLAMA_PORTABLE=1
 RUN wget -O model.ggml $MODEL
+CMD /bin/python3 ./koboldcpp.py --model model.ggml --usecublas --gpulayers 99 --multiuser --contextsize $CONTEXT_SIZE --port 7860 --hordeconfig $MODEL_NAME 1 1