Merge remote-tracking branch 'origin/dev' into jp6

Author: dusty-nv

packages/llm/exllama/Dockerfile

@@ -1,5 +1,6 @@
 #---
-# name: exllama
+# name: exllama:v1
+# alias: exllama
 # group: llm
 # config: config.py
 # depends: [pytorch, huggingface_hub]

packages/llm/llava/Dockerfile

@@ -18,10 +18,10 @@ ADD https://api.github.com/repos/${LLAVA_REPO}/git/refs/heads/${LLAVA_BRANCH} /t
 
 RUN git clone --branch=${LLAVA_BRANCH} --depth=1 https://github.com/${LLAVA_REPO} llava && \
     cd llava && \
-    sed 's|"deepspeed==0.9.5",.*||' -i pyproject.toml && \
-    sed 's|"transformers.*"|"transformers"|' -i pyproject.toml && \
-    sed 's|"accelerate.*"|"accelerate"|' -i pyproject.toml && \
-    sed 's|"bitsandbytes.*"|"bitsandbytes"|' -i pyproject.toml && \
+#    sed 's|"deepspeed==0.9.5",.*||' -i pyproject.toml && \
+#    sed 's|"transformers.*"|"transformers"|' -i pyproject.toml && \
+#    sed 's|"accelerate.*"|"accelerate"|' -i pyproject.toml && \
+    sed 's|"bitsandbytes.*",||' -i pyproject.toml && \
     cat pyproject.toml
 
 COPY benchmark.py llava/llava/serve/
@@ -31,5 +31,8 @@ RUN cd llava && \
     cp dist/llava*.whl /opt
 
 RUN pip3 install --no-cache-dir --verbose llava*.whl
- 
+
+RUN pip3 install --verbose /opt/torch*.whl
+RUN pip3 install --verbose /opt/torchvision*.whl
+
 WORKDIR /

packages/llm/text-generation-webui/Dockerfile

@@ -2,7 +2,7 @@
 # name: text-generation-webui
 # group: llm
 # config: config.py
-# depends: [transformers, auto_gptq, gptq-for-llama, exllama, exllama:v2, llama_cpp:gguf]
+# depends: [transformers, auto_gptq, gptq-for-llama, exllama:v1, exllama:v2, llama_cpp:gguf]
 # requires: '>=34.1.0'
 # docs: docs.md
 #---