fix docker cache handling
Browse files- Dockerfile +6 -3
- README.md +13 -0
- start.sh +21 -2
Dockerfile
CHANGED
|
@@ -2,7 +2,7 @@ FROM ubuntu:22.04 AS build
|
|
| 2 |
WORKDIR /src
|
| 3 |
|
| 4 |
RUN apt-get update && \
|
| 5 |
-
apt-get install -y build-essential cmake git curl && \
|
| 6 |
rm -rf /var/lib/apt/lists/* /var/cache/apt/archives/*
|
| 7 |
|
| 8 |
ARG CRISPASR_REPO=https://github.com/CrispStrobe/CrispASR
|
|
@@ -12,8 +12,10 @@ RUN git clone --depth 1 "${CRISPASR_REPO}" /src/CrispASR && \
|
|
| 12 |
git -C /src/CrispASR fetch --depth 1 origin "${CRISPASR_REF}" && \
|
| 13 |
git -C /src/CrispASR checkout --detach FETCH_HEAD
|
| 14 |
WORKDIR /src/CrispASR
|
| 15 |
-
|
| 16 |
-
|
|
|
|
|
|
|
| 17 |
|
| 18 |
FROM ubuntu:22.04 AS runtime
|
| 19 |
WORKDIR /space
|
|
@@ -35,6 +37,7 @@ RUN chmod +x /space/start.sh && \
|
|
| 35 |
|
| 36 |
ENV PATH=/opt/crispasr/build/bin:$PATH
|
| 37 |
ENV CRISPASR_SERVER_URL=http://127.0.0.1:8080
|
|
|
|
| 38 |
ENV GRADIO_SERVER_NAME=0.0.0.0
|
| 39 |
ENV GRADIO_SERVER_PORT=7860
|
| 40 |
|
|
|
|
| 2 |
WORKDIR /src
|
| 3 |
|
| 4 |
RUN apt-get update && \
|
| 5 |
+
apt-get install -y build-essential cmake git curl ninja-build && \
|
| 6 |
rm -rf /var/lib/apt/lists/* /var/cache/apt/archives/*
|
| 7 |
|
| 8 |
ARG CRISPASR_REPO=https://github.com/CrispStrobe/CrispASR
|
|
|
|
| 12 |
git -C /src/CrispASR fetch --depth 1 origin "${CRISPASR_REF}" && \
|
| 13 |
git -C /src/CrispASR checkout --detach FETCH_HEAD
|
| 14 |
WORKDIR /src/CrispASR
|
| 15 |
+
ARG CRISPASR_BUILD_JOBS
|
| 16 |
+
RUN jobs="${CRISPASR_BUILD_JOBS:-$(nproc)}" && \
|
| 17 |
+
cmake -S . -B build -G Ninja -DWHISPER_BUILD_TESTS=OFF && \
|
| 18 |
+
cmake --build build -j"${jobs}" --target whisper-cli
|
| 19 |
|
| 20 |
FROM ubuntu:22.04 AS runtime
|
| 21 |
WORKDIR /space
|
|
|
|
| 37 |
|
| 38 |
ENV PATH=/opt/crispasr/build/bin:$PATH
|
| 39 |
ENV CRISPASR_SERVER_URL=http://127.0.0.1:8080
|
| 40 |
+
ENV CRISPASR_CACHE_DIR=/cache
|
| 41 |
ENV GRADIO_SERVER_NAME=0.0.0.0
|
| 42 |
ENV GRADIO_SERVER_PORT=7860
|
| 43 |
|
README.md
CHANGED
|
@@ -19,6 +19,7 @@ This folder contains a Hugging Face Docker Space wrapper for CrispASR:
|
|
| 19 |
- `CRISPASR_BACKEND=whisper` or another backend name
|
| 20 |
- `CRISPASR_LANGUAGE=en`
|
| 21 |
- `CRISPASR_AUTO_DOWNLOAD=1`
|
|
|
|
| 22 |
- `CRISPASR_EXTRA_ARGS=`
|
| 23 |
|
| 24 |
## Local build
|
|
@@ -30,3 +31,15 @@ docker run --rm -p 7860:7860 -p 8080:8080 \
|
|
| 30 |
-v "$PWD/models:/models" \
|
| 31 |
crispasr-hf-space
|
| 32 |
```
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 19 |
- `CRISPASR_BACKEND=whisper` or another backend name
|
| 20 |
- `CRISPASR_LANGUAGE=en`
|
| 21 |
- `CRISPASR_AUTO_DOWNLOAD=1`
|
| 22 |
+
- `CRISPASR_CACHE_DIR=/cache`
|
| 23 |
- `CRISPASR_EXTRA_ARGS=`
|
| 24 |
|
| 25 |
## Local build
|
|
|
|
| 31 |
-v "$PWD/models:/models" \
|
| 32 |
crispasr-hf-space
|
| 33 |
```
|
| 34 |
+
|
| 35 |
+
For auto-downloads, mount a writable cache volume if you want models to survive container restarts:
|
| 36 |
+
|
| 37 |
+
```bash
|
| 38 |
+
docker volume create crispasr-cache
|
| 39 |
+
docker run --rm -p 7860:7860 -p 8080:8080 \
|
| 40 |
+
-e CRISPASR_AUTO_DOWNLOAD=1 \
|
| 41 |
+
-v crispasr-cache:/cache \
|
| 42 |
+
crispasr-hf-space
|
| 43 |
+
```
|
| 44 |
+
|
| 45 |
+
Build parallelism can be tuned with `--build-arg CRISPASR_BUILD_JOBS=8`.
|
start.sh
CHANGED
|
@@ -15,10 +15,29 @@ MODEL_PATH="${CRISPASR_MODEL:-/models/model.gguf}"
|
|
| 15 |
LANGUAGE="${CRISPASR_LANGUAGE:-en}"
|
| 16 |
BACKEND="${CRISPASR_BACKEND:-whisper}"
|
| 17 |
AUTO_DOWNLOAD="${CRISPASR_AUTO_DOWNLOAD:-1}"
|
|
|
|
| 18 |
EXTRA_ARGS="${CRISPASR_EXTRA_ARGS:-}"
|
| 19 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 20 |
declare -a cmd
|
| 21 |
-
cmd=(crispasr --server --host "$SERVER_HOST" --port "$SERVER_PORT" -l "$LANGUAGE")
|
| 22 |
|
| 23 |
if [[ "$AUTO_DOWNLOAD" == "1" ]]; then
|
| 24 |
cmd+=(-m auto --auto-download)
|
|
@@ -35,7 +54,7 @@ if [[ -n "$EXTRA_ARGS" ]]; then
|
|
| 35 |
fi
|
| 36 |
|
| 37 |
log "startup"
|
| 38 |
-
log "server_host=$SERVER_HOST server_port=$SERVER_PORT backend=$BACKEND language=$LANGUAGE auto_download=$AUTO_DOWNLOAD"
|
| 39 |
if [[ "$AUTO_DOWNLOAD" == "1" ]]; then
|
| 40 |
log "model=auto"
|
| 41 |
else
|
|
|
|
| 15 |
LANGUAGE="${CRISPASR_LANGUAGE:-en}"
|
| 16 |
BACKEND="${CRISPASR_BACKEND:-whisper}"
|
| 17 |
AUTO_DOWNLOAD="${CRISPASR_AUTO_DOWNLOAD:-1}"
|
| 18 |
+
CACHE_DIR="${CRISPASR_CACHE_DIR:-/cache}"
|
| 19 |
EXTRA_ARGS="${CRISPASR_EXTRA_ARGS:-}"
|
| 20 |
|
| 21 |
+
ensure_writable_dir() {
|
| 22 |
+
local dir="$1"
|
| 23 |
+
if [[ ! -d "$dir" ]]; then
|
| 24 |
+
if ! mkdir -p "$dir" 2>/dev/null; then
|
| 25 |
+
log "ERROR: could not create cache directory '$dir'"
|
| 26 |
+
exit 70
|
| 27 |
+
fi
|
| 28 |
+
fi
|
| 29 |
+
if [[ ! -w "$dir" ]]; then
|
| 30 |
+
log "ERROR: cache directory '$dir' is not writable by uid=$(id -u) gid=$(id -g)"
|
| 31 |
+
ls -ld "$dir" >&2 || true
|
| 32 |
+
log "If this is a bind mount, chown the host directory or set CRISPASR_CACHE_DIR to a writable path."
|
| 33 |
+
exit 70
|
| 34 |
+
fi
|
| 35 |
+
}
|
| 36 |
+
|
| 37 |
+
ensure_writable_dir "$CACHE_DIR"
|
| 38 |
+
|
| 39 |
declare -a cmd
|
| 40 |
+
cmd=(crispasr --server --host "$SERVER_HOST" --port "$SERVER_PORT" -l "$LANGUAGE" --cache-dir "$CACHE_DIR")
|
| 41 |
|
| 42 |
if [[ "$AUTO_DOWNLOAD" == "1" ]]; then
|
| 43 |
cmd+=(-m auto --auto-download)
|
|
|
|
| 54 |
fi
|
| 55 |
|
| 56 |
log "startup"
|
| 57 |
+
log "server_host=$SERVER_HOST server_port=$SERVER_PORT backend=$BACKEND language=$LANGUAGE auto_download=$AUTO_DOWNLOAD cache_dir=$CACHE_DIR"
|
| 58 |
if [[ "$AUTO_DOWNLOAD" == "1" ]]; then
|
| 59 |
log "model=auto"
|
| 60 |
else
|