fix: all docker filepaths weights folders, use local weights

kaito-project · Oct 16, 2023 · 413d1ed · 413d1ed
1 parent 4017ead
commit 413d1ed
Show file tree

Hide file tree

Showing 5 changed files with 29 additions and 22 deletions.
diff --git a/.github/workflows/preset-image-build.yml b/.github/workflows/preset-image-build.yml
@@ -68,13 +68,13 @@ jobs:
           IFS=';' read -ra ADDR <<< "$FILES_MODIFIED"
           for file in "${ADDR[@]}"; do
             echo "Checking file: $file"
-            if [[ "$file" == pkg/presets/falcon/* ]] && [[ "$FALCON_MODIFIED" == false ]]; then
+            if [[ "$file" == presets/falcon/* ]] && [[ "$FALCON_MODIFIED" == false ]]; then
               echo "File matches falcon path: $file"
               FALCON_MODIFIED=true
-            elif [[ "$file" == pkg/presets/llama-2/* ]] && [[ "$LLAMA2_MODIFIED" == false ]]; then
+            elif [[ "$file" == presets/llama-2/* ]] && [[ "$LLAMA2_MODIFIED" == false ]]; then
               echo "File matches llama-2 path: $file"
               LLAMA2_MODIFIED=true
-            elif [[ "$file" == pkg/presets/llama-2-chat/* ]] && [[ "$LLAMA2_CHAT_MODIFIED" == false ]]; then
+            elif [[ "$file" == presets/llama-2-chat/* ]] && [[ "$LLAMA2_CHAT_MODIFIED" == false ]]; then
               echo "File matches llama-2-chat path: $file"
               LLAMA2_CHAT_MODIFIED=true
             else
@@ -139,19 +139,19 @@ jobs:
 
           - name: llama-2-7b
             dockerfile: docker/presets/llama-2/Dockerfile
-            build_args: "--build-arg EXTERNAL_IP=__EXTERNAL_IP__ --build-arg EXTERNAL_PORT=__EXTERNAL_PORT__ --build-arg LLAMA_VERSION=llama-2-7b --build-arg SRC_DIR=pkg/presets/llama-2 --build-arg WEB_SERVER_AUTH_TOKEN=__WEB_SERVER_AUTH_TOKEN__"
+            build_args: "--build-arg LLAMA_VERSION=llama-2-7b --build-arg SRC_DIR=/home/presets/llama-2"
 
           - name: llama-2-7b-chat
             dockerfile: docker/presets/llama-2/Dockerfile
-            build_args: "--build-arg EXTERNAL_IP=__EXTERNAL_IP__ --build-arg EXTERNAL_PORT=__EXTERNAL_PORT__ --build-arg LLAMA_VERSION=llama-2-7b-chat --build-arg SRC_DIR=pkg/presets/llama-2-chat --build-arg WEB_SERVER_AUTH_TOKEN=__WEB_SERVER_AUTH_TOKEN__"
+            build_args: "--build-arg LLAMA_VERSION=llama-2-7b-chat --build-arg SRC_DIR=/home/presets/llama-2-chat"
 
           - name: llama-2-13b
             dockerfile: docker/presets/llama-2/Dockerfile
-            build_args: "--build-arg EXTERNAL_IP=__EXTERNAL_IP__ --build-arg EXTERNAL_PORT=__EXTERNAL_PORT__ --build-arg LLAMA_VERSION=llama-2-13b --build-arg SRC_DIR=pkg/presets/llama-2 --build-arg WEB_SERVER_AUTH_TOKEN=__WEB_SERVER_AUTH_TOKEN__"
+            build_args: "--build-arg LLAMA_VERSION=llama-2-13b --build-arg SRC_DIR=/home/presets/llama-2"
 
           - name: llama-2-13b-chat
             dockerfile: docker/presets/llama-2/Dockerfile
-            build_args: "--build-arg EXTERNAL_IP=__EXTERNAL_IP__ --build-arg EXTERNAL_PORT=__EXTERNAL_PORT__ --build-arg LLAMA_VERSION=llama-2-13b-chat --build-arg SRC_DIR=pkg/presets/llama-2-chat --build-arg WEB_SERVER_AUTH_TOKEN=__WEB_SERVER_AUTH_TOKEN__"
+            build_args: "--build-arg LLAMA_VERSION=llama-2-13b-chat --build-arg SRC_DIR=/home/presets/llama-2-chat"
 
         include:
           - name: falcon-7b
@@ -235,11 +235,11 @@ jobs:
 
           - name: llama-2-70b
             dockerfile: docker/presets/llama-2/Dockerfile
-            build_args: "--build-arg EXTERNAL_IP=__EXTERNAL_IP__ --build-arg EXTERNAL_PORT=__EXTERNAL_PORT__ --build-arg LLAMA_VERSION=llama-2-70b --build-arg SRC_DIR=/home/pkg/presets/llama-2 --build-arg WEB_SERVER_AUTH_TOKEN=__WEB_SERVER_AUTH_TOKEN__"
+            build_args: "--build-arg LLAMA_VERSION=llama-2-70b --build-arg SRC_DIR=/home/presets/llama-2"
 
           - name: llama-2-70b-chat
             dockerfile: docker/presets/llama-2/Dockerfile
-            build_args: "--build-arg EXTERNAL_IP=__EXTERNAL_IP__ --build-arg EXTERNAL_PORT=__EXTERNAL_PORT__ --build-arg LLAMA_VERSION=llama-2-70b-chat --build-arg SRC_DIR=/home/pkg/presets/llama-2-chat --build-arg WEB_SERVER_AUTH_TOKEN=__WEB_SERVER_AUTH_TOKEN__"
+            build_args: "--build-arg LLAMA_VERSION=llama-2-70b-chat --build-arg SRC_DIR=/home/presets/llama-2-chat"
 
         include:
           # TODO: Support large models
@@ -324,9 +324,6 @@ jobs:
           retries=3
           while [ $retries -gt 0 ]; do
             BUILD_ARGS="${{ matrix.model.build_args }}"
-            BUILD_ARGS=${BUILD_ARGS/__EXTERNAL_IP__/${{ secrets.AZURE_WEB_SERVER_EXTERNAL_IP }}}
-            BUILD_ARGS=${BUILD_ARGS/__EXTERNAL_PORT__/${{ secrets.AZURE_WEB_SERVER_EXTERNAL_PORT }}}
-            BUILD_ARGS=${BUILD_ARGS/__WEB_SERVER_AUTH_TOKEN__/${{ secrets.WEB_SERVER_AUTH_TOKEN }}}
 
             echo "Docker BUILD_ARGS: $BUILD_ARGS"
 

diff --git a/docker/presets/download_script.go b/docker/presets/download_script.go
@@ -1,3 +1,8 @@
+/*
+ * This script allows you to download llama and falcon weights using parallel wget.
+ * Llama is to be downloaded via hosted web server.
+ * Falcon from public url.
+ */
 package main
 
 import (

diff --git a/docker/presets/falcon/Dockerfile b/docker/presets/falcon/Dockerfile
@@ -15,8 +15,7 @@ RUN pip install --no-cache-dir -r requirements.txt
 
 ARG FALCON_MODEL_NAME
 
-COPY /home/docker/presets/download_script.go /workspace/download_script.go
-RUN go run /workspace/download_script.go "public" ${FALCON_MODEL_NAME} "/workspace/falcon/weights"
-
+# Copy the entire model to the weights directory
+COPY /falcon/${FALCON_MODEL_NAME} /workspace/falcon/weights
 # Copy the entire 'presets/falcon' folder to the working directory
 COPY /home/presets/falcon /workspace/falcon
diff --git a/docker/presets/llama-2/Dockerfile b/docker/presets/llama-2/Dockerfile
@@ -26,16 +26,16 @@ RUN pip install 'uvicorn[standard]'
 
 ARG LLAMA_VERSION
 ARG SRC_DIR
-ARG EXTERNAL_IP
-ARG EXTERNAL_PORT
-ARG WEB_SERVER_AUTH_TOKEN
+# ARG EXTERNAL_IP
+# ARG EXTERNAL_PORT
+# ARG WEB_SERVER_AUTH_TOKEN
 
-ENV AUTH_TOKEN_ENV_VAR=${WEB_SERVER_AUTH_TOKEN}
+# ENV AUTH_TOKEN_ENV_VAR=${WEB_SERVER_AUTH_TOKEN}
 
 # Copy Go download script into the Docker image
-COPY /home/docker/presets/download_script.go /workspace/download_script.go
+# COPY /home/docker/presets/download_script.go /workspace/download_script.go
 
 # Use Go download script to fetch model weights
-RUN go run /workspace/download_script.go "private" ${LLAMA_VERSION} "/workspace/llama/llama-2/weights" ${EXTERNAL_IP} ${EXTERNAL_PORT}
-
+# RUN go run /workspace/download_script.go "private" ${LLAMA_VERSION} "/workspace/llama/llama-2/weights" ${EXTERNAL_IP} ${EXTERNAL_PORT}
+ADD /llama/${LLAMA_VERSION} /workspace/llama/llama-2/weights
 ADD ${SRC_DIR} /workspace/llama/llama-2
diff --git a/presets/k8s/docker.yaml b/presets/k8s/docker.yaml
@@ -22,6 +22,8 @@ spec:
           mountPath: /home
         - name: llama-volume
           mountPath: /home/llama
+        - name: falcon-volume
+          mountPath: /home/falcon
       volumes:
       - name: host-volume
         hostPath:
@@ -30,4 +32,8 @@ spec:
       - name: llama-volume
         hostPath: 
           path: /llama
+          type: Directory
+      - name: falcon-volume
+        hostPath:
+          path: /falcon
           type: Directory