remote worker

runpod/runpod-worker-oobabooga-api/Dockerfile

@@ -82,7 +82,7 @@ RUN --mount=type=cache,target=/root/.cache,sharing=locked \
     cd /workspace/text-generation-webui/ && mkdir repositories && cd repositories && \
     # Triton
     git clone https://github.com/qwopqwop200/GPTQ-for-LLaMa.git && \
-    git reset --hard 508de42ff45ec560a4504e12b0d42114d599cf38 && \
+    (cd GPTQ-for-LLaMa && git reset --hard 508de42ff45ec560a4504e12b0d42114d599cf38) && \
     (cd GPTQ-for-LLaMa && pip3 install -r requirements.txt && rm -rf .git)
     # Cuda
 #    git clone --branch cuda --single-branch https://github.com/qwopqwop200/GPTQ-for-LLaMa.git && \
@@ -107,6 +107,7 @@ RUN cd /workspace/text-generation-webui && python3 download-model.py ${MODEL_NAM
     #https://huggingface.co/BlinkDL/rwkv-4-pile-7b/resolve/main/RWKV-4-Pile-7B-20230406-ctx8192-test949.pth
 
 #RUN cd /workspace/text-generation-webui && python3 download-model.py MetaIX/GPT4-X-Alpaca-30B-Int4 --text-only && \
+#    rm /workspace/text-generation-webui/models/MetaIX_GPT4-X-Alpaca-30B-Int4/pytorch_model.bin.index.json
 #    wget -P /workspace/text-generation-webui/models/MetaIX_GPT4-X-Alpaca-30B-Int4/ https://huggingface.co/MetaIX/GPT4-X-Alpaca-30B-Int4/resolve/main/gpt4-x-alpaca-30b-4bit.safetensors
 
 FROM ${BASE_IMAGE}
@@ -123,13 +124,14 @@ RUN --mount=type=cache,target=/var/cache/apt,sharing=locked apt-get update --yes
 
 RUN --mount=type=cache,target=/root/.cache,sharing=locked pip3 install --upgrade pip setuptools && \
     pip3 install cuda-python==11.8.0 && \
-    pip3 install torch torchvision torchaudio --extra-index-url=https://download.pytorch.org/whl/cu118 && \
+    pip3 install torch torchvision torchaudio --extra-index-url=https://download.pytorch.org/whl/cu118
+RUN --mount=type=cache,target=/root/.cache,sharing=locked \
     pip3 install bitsandbytes && \
     cp /usr/local/lib/python3.10/dist-packages/bitsandbytes/libbitsandbytes_cuda118.so /usr/local/lib/python3.10/dist-packages/bitsandbytes/libbitsandbytes_cpu.so && \
     pip3 install safetensors && \
     pip3 install sentencepiece && \
     pip3 install diffusers && \
-    pip3 install deepspeed && \
+#    pip3 install deepspeed && \
     pip3 install accelerate xformers triton && \
     pip3 install git+https://github.com/huggingface/transformers.git && \
     pip3 install rwkv && \