Fix challenge-cli not setting storage when using --volume and friends. Improve error messages. Fix multiple issues with Dockerfile.gpu. Change make docker-run and make docker-run-gpu to use docker flags that match the Dyff cloud environment by default.

Dockerfile.gpu

@@ -1,25 +1,23 @@
-FROM nvidia/cuda:12.6.3-cudnn-runtime-ubuntu24.04
+# This Dockerfile is based on this post in the uv GitHub:
+# https://github.com/astral-sh/uv/issues/7758#issuecomment-3263282018
+FROM python:3.12-slim AS build
 
 # Install 'uv', which we will use to install Python dependencies
 COPY --from=ghcr.io/astral-sh/uv:latest /uv /uvx /bin/
+WORKDIR /app
 
-ENV PYTHONDONTWRITEBYTECODE="1" \
-    PYTHONUNBUFFERED="1" \
-    DEBIAN_FRONTEND="noninteractive"
+# uv configuration
+# Ref: https://docs.astral.sh/uv/guides/integration/docker/#compiling-bytecode
+ENV UV_COMPILE_BYTECODE=1 \
+    # Ref: https://docs.astral.sh/uv/guides/integration/docker/#caching
+    UV_LINK_MODE=copy \
+    # Ref: https://docs.astral.sh/uv/guides/integration/docker/#managing-python-interpreters
+    UV_PYTHON_INSTALL_DIR=/opt/python
 
-# Install Python and create venv
-WORKDIR /app/
 RUN uv python install 3.12
 RUN uv venv
 RUN uv pip install --no-cache-dir --upgrade pip setuptools wheel
 
-# Download models during build instead of copying from local
-COPY scripts/model_download.bash /tmp/model_download.bash
-RUN . .venv/bin/activate && \
-    uv pip install --no-cache-dir huggingface-hub && \
-    bash /tmp/model_download.bash && \
-    rm /tmp/model_download.bash
-
 # Install CPU requirements
 COPY requirements.cpu.txt ./
 RUN uv pip install --no-cache-dir -r ./requirements.cpu.txt
@@ -28,9 +26,51 @@ RUN uv pip install --no-cache-dir -r ./requirements.cpu.txt
 COPY requirements.torch.gpu.txt ./
 RUN uv pip install --no-cache-dir -r ./requirements.torch.gpu.txt
 
+# Download models during build instead of copying from local
+COPY scripts/model_download.bash /tmp/model_download.bash
+RUN . .venv/bin/activate && \
+    uv pip install --no-cache-dir huggingface-hub && \
+    bash /tmp/model_download.bash && \
+    rm /tmp/model_download.bash
+
 COPY app ./app
 COPY main.py ./
 
+# ----------------------------------------------------------------------------
+
+FROM nvidia/cuda:12.6.3-cudnn-runtime-ubuntu24.04 AS run
+WORKDIR /app
+
+ENV PYTHONDONTWRITEBYTECODE="1" \
+    PYTHONUNBUFFERED="1" \
+    DEBIAN_FRONTEND="noninteractive"
+
+# The container will run as a non-root user with unknown UID and GID, so we
+# need to give everyone permission to all of the files
+COPY --from=build --chmod=777 /opt/python /opt/python
+COPY --from=build --chmod=777 /app /app
+
+# Add python env to PATH and app folder to PYTHONPATH
+ENV PATH="/app/.venv/bin:${PATH}" \
+    PYTHONPATH="/app"
+
 EXPOSE 8000
 
-ENTRYPOINT ["python3", "-m", "uvicorn", "main:app", "--host", "0.0.0.0", "--port", "8000"]
+# torch.compile tries to cache things in /tmp/torchinductor_{username},
+# which doesn't work because we don't have a user account
+ENV TORCHINDUCTOR_CACHE_DIR="/tmp/torchinductor"
+
+# COPY entrypoint.sh ./entrypoint.sh
+# RUN chmod a+x /app/entrypoint.sh
+# RUN chmod -R a+x /app/.venv/bin
+# ENTRYPOINT ["./entrypoint.sh", ".venv/bin/python", "-m", "uvicorn", "main:app", "--host", "0.0.0.0", "--port", "8000"]
+
+# ENV PATH="/app/.venv/bin:$PATH"
+# ENTRYPOINT [".", "/app/.venv/bin/activate;", "python", "-m", "uvicorn", "main:app", "--host", "0.0.0.0", "--port", "8000"]
+
+# ENTRYPOINT ["uv", "run", "--no-cache", "--no-sync", "python", "-m", "uvicorn", "main:app", "--host", "0.0.0.0", "--port", "8000"]
+
+# RUN chmod a+x .venv/bin/python
+# ENTRYPOINT ["/app/.venv/bin/python", "-m", "uvicorn", "main:app", "--host", "0.0.0.0", "--port", "8000"]
+
+ENTRYPOINT ["python", "-m", "uvicorn", "main:app", "--host", "0.0.0.0", "--port", "8000"]

challenge-cli.py

@@ -41,6 +41,25 @@ def _wait_for_status(
     raise AssertionError("timeout")
 
 
+def _directory_size_bytes(directory: Path | str) -> int:
+    directory = Path(directory)
+    size: int = 0
+    for f in directory.rglob("*"):
+        if f.exists() and not f.is_symlink():
+            size += f.stat().st_size
+    return size
+
+
+def _storage_for_size(size_bytes: int) -> str:
+    """Convert bytes to k8s Quantity and add some overhead in case the size
+    is different on different filesystems.
+    """
+    Gi = int(1024**3)
+    size_B_with_overhead = size_bytes + 2*Gi
+    size_Gi_with_overhead = size_B_with_overhead // Gi
+    return f"{size_Gi_with_overhead}Gi"
+
+
 def _common_options(f):
     @click.option(
         "--account",
@@ -49,6 +68,13 @@ def _common_options(f):
         help="Your account ID",
         metavar="ID",
     )
+    @click.option(
+        "--timeout-seconds",
+        type=int,
+        default=120,
+        help="Timeout for Dyff client operations",
+        metavar="SECONDS",
+    )
     @functools.wraps(f)
     def wrapper(*args, **kwargs):
         return f(*args, **kwargs)
@@ -130,8 +156,9 @@ def upload_submission(
     artifact_id: str | None,
     model_id: str | None,
     gpu: bool,
+    timeout_seconds: int,
 ) -> None:
-    dyffapi = Client()
+    dyffapi = Client(timeout=httpx.Timeout(timeout_seconds))
 
     # Upload the image
     if artifact_id is None:
@@ -164,15 +191,20 @@ def upload_submission(
         assert artifact is not None
 
     model: Model | None = None
+    model_storage_quantity: str | None = None
     if model_id is None:
         if volume is not None:
             if volume_mount is None:
                 raise click.UsageError("--volume-mount is required when --volume is used")
             
-            click.echo("creating Model from local directory ...")
+            model_storage_quantity = _storage_for_size(_directory_size_bytes(volume))
+            click.echo(f"creating Model from local directory with storage={model_storage_quantity} ...")
 
             model = dyffapi.models.create_from_volume(
-                volume, name="model_volume", account=account, resources=ModelResources()
+                volume,
+                name="model_volume",
+                account=account,
+                resources=ModelResources(storage=model_storage_quantity),
             )
             assert model is not None
             click.echo(f"model.id: \"{model.id}\"")
@@ -197,11 +229,22 @@ def upload_submission(
         model = dyffapi.models.get(model_id)
         if model is None:
             raise click.UsageError(f"--model={model_id}: model not found")
+        model_storage_quantity = model.resources.storage
+        if model_storage_quantity is None:
+            raise click.UsageError(
+                f"--model={model_id}: model.resources.storage not set;"
+                " was this model created with the challenge-cli tool?"
+            )
 
     # Create a runnable InferenceService
     if volume_mount is not None:
         if model is None:
             raise click.UsageError("--volume-mount requires --volume or --model")
+        if model_storage_quantity is None:
+            raise click.UsageError(
+                f"model {model.id}: model.resources.storage not set;"
+                " was this model created with the challenge-cli tool?"
+            )
         if not volume_mount.is_absolute():
             raise click.UsageError("--volume-mount must be an absolute path")
         volumeMounts=[
@@ -238,7 +281,7 @@ def upload_submission(
         runner=InferenceServiceRunner(
             kind=InferenceServiceRunnerKind.CONTAINER,
             imageRef=EntityIdentifier.of(artifact),
-            resources=ModelResources(),
+            resources=ModelResources(storage=model_storage_quantity),
             volumeMounts=volumeMounts,
             accelerator=accelerator,
         ),
@@ -287,8 +330,8 @@ def upload_submission(
     help="The Challenge ID to submit to.",
     metavar="ID",
 )
-def submit(account: str, task_id: str, team_id: str, service_id: str, challenge_id: str) -> None:
-    dyffapi = Client()
+def submit(account: str, task_id: str, team_id: str, service_id: str, challenge_id: str, timeout_seconds: int) -> None:
+    dyffapi = Client(timeout=httpx.Timeout(timeout_seconds))
 
     challenge_tasks = {
         "dc509a8c771b492b90c43012fde9a04f": {

makefile

@@ -23,15 +23,11 @@ docker-build:
 
 .PHONY: docker-run
 docker-run:
-	docker run --rm -it -p 8000:8000 $(IMAGE)
+	docker run --rm -it --read-only --tmpfs "/tmp" -u 1001:1001 -p 8000:8000 $(IMAGE)
 
 .PHONY: docker-test-no-internet
 docker-test-no-internet:
-	docker run --rm -it --network none $(IMAGE)
-
-.PHONY: docker-test-readonly
-docker-test-readonly:
-	docker run --rm -it --read-only --tmpfs "/tmp" $(IMAGE)
+	docker run --rm -it --read-only --tmpfs "/tmp" -u 1001:1001 -p 8000:8000 --network none $(IMAGE)
 
 .PHONY: compile
 compile:
@@ -48,6 +44,10 @@ compile-gpu:
 docker-build-gpu:
 	docker build -t $(IMAGE)-gpu -f Dockerfile.gpu .
 
+.PHONY: docker-run
+docker-run-gpu:
+	docker run --rm -it --read-only --tmpfs "/tmp" -u 1001:1001 -p 8000:8000 $(IMAGE)-gpu
+
 requirements.cpu.txt: requirements.in requirements.torch.cpu.txt | $(VENV)
 	uv pip compile --python-version 3.12 --upgrade -o requirements.cpu.txt requirements.cpu.in
 

requirements.cpu.txt

@@ -6,7 +6,7 @@ annotated-doc==0.0.4
     # via fastapi
 annotated-types==0.7.0
     # via pydantic
-anyio==4.11.0
+anyio==4.12.1
     # via
     #   httpx
     #   jupyter-server
@@ -20,17 +20,17 @@ arrow==1.4.0
     # via isoduration
 asttokens==3.0.1
     # via stack-data
-async-lru==2.0.5
+async-lru==2.1.0
     # via jupyterlab
 attrs==25.4.0
     # via
     #   jsonschema
     #   referencing
-azure-core==1.36.0
+azure-core==1.38.0
     # via dyff-client
 babel==2.17.0
     # via jupyterlab-server
-beautifulsoup4==4.14.2
+beautifulsoup4==4.14.3
     # via
     #   dyff-audit
     #   nbconvert
@@ -38,7 +38,7 @@ bleach==6.3.0
     # via nbconvert
 canonicaljson==2.0.0
     # via dyff-schema
-certifi==2025.11.12
+certifi==2026.1.4
     # via
     #   httpcore
     #   httpx
@@ -53,7 +53,7 @@ click==8.3.1
     #   uvicorn
 comm==0.2.3
     # via ipykernel
-debugpy==1.8.17
+debugpy==1.8.19
     # via ipykernel
 decorator==5.2.1
     # via ipython
@@ -61,13 +61,13 @@ defusedxml==0.7.1
     # via nbconvert
 dnspython==2.8.0
     # via email-validator
-dyff-audit==0.16.1
+dyff-audit==0.16.8
     # via -r requirements.in
-dyff-client==0.23.5
+dyff-client==0.23.11
     # via
     #   -r requirements.in
     #   dyff-audit
-dyff-schema==0.39.1
+dyff-schema==0.43.4
     # via
     #   -r requirements.in
     #   dyff-audit
@@ -76,17 +76,17 @@ email-validator==2.3.0
     # via dyff-schema
 executing==2.2.1
     # via stack-data
-fastapi==0.121.2
+fastapi==0.128.0
     # via -r requirements.in
 fastjsonschema==2.21.2
     # via nbformat
-filelock==3.20.0
+filelock==3.20.3
     # via
     #   huggingface-hub
     #   transformers
 fqdn==1.5.1
     # via jsonschema
-fsspec==2025.10.0
+fsspec==2026.1.0
     # via huggingface-hub
 google-i18n-address==3.1.1
     # via dyff-schema
@@ -108,7 +108,7 @@ huggingface-hub==0.36.0
     # via
     #   tokenizers
     #   transformers
-hypothesis==6.148.1
+hypothesis==6.150.2
     # via
     #   dyff-schema
     #   hypothesis-jsonschema
@@ -123,7 +123,7 @@ idna==3.11
     #   requests
 ipykernel==7.1.0
     # via jupyterlab
-ipython==9.7.0
+ipython==9.9.0
     # via ipykernel
 ipython-pygments-lexers==1.1.1
     # via ipython
@@ -139,7 +139,7 @@ jinja2==3.1.6
     #   jupyterlab
     #   jupyterlab-server
     #   nbconvert
-json5==0.12.1
+json5==0.13.0
     # via jupyterlab-server
 jsonpath-ng==1.7.0
     # via
@@ -147,7 +147,7 @@ jsonpath-ng==1.7.0
     #   dyff-schema
 jsonpointer==3.0.0
     # via jsonschema
-jsonschema==4.25.1
+jsonschema==4.26.0
     # via
     #   hypothesis-jsonschema
     #   jupyter-events
@@ -155,7 +155,7 @@ jsonschema==4.25.1
     #   nbformat
 jsonschema-specifications==2025.9.1
     # via jsonschema
-jupyter-client==8.6.3
+jupyter-client==8.8.0
     # via
     #   ipykernel
     #   jupyter-server
@@ -180,9 +180,9 @@ jupyter-server==2.17.0
     #   jupyterlab-server
     #   notebook
     #   notebook-shim
-jupyter-server-terminals==0.5.3
+jupyter-server-terminals==0.5.4
     # via jupyter-server
-jupyterlab==4.4.10
+jupyterlab==4.5.2
     # via notebook
 jupyterlab-pygments==0.3.0
     # via nbconvert
@@ -202,9 +202,9 @@ matplotlib-inline==0.2.1
     # via
     #   ipykernel
     #   ipython
-mistune==3.1.4
+mistune==3.2.0
     # via nbconvert
-nbclient==0.10.2
+nbclient==0.10.4
     # via nbconvert
 nbconvert==7.16.6
     # via
@@ -218,7 +218,7 @@ nbformat==5.10.4
     #   nbconvert
 nest-asyncio==1.6.0
     # via ipykernel
-notebook==7.4.7
+notebook==7.5.2
     # via dyff-audit
 notebook-shim==0.2.4
     # via
@@ -232,7 +232,7 @@ numpy==1.26.4
     #   dyff-schema
     #   pandas
     #   transformers
-packaging==25.0
+packaging==26.0
     # via
     #   huggingface-hub
     #   ipykernel
@@ -242,7 +242,7 @@ packaging==25.0
     #   jupyterlab-server
     #   nbconvert
     #   transformers
-pandas==2.3.3
+pandas==3.0.0
     # via
     #   -r requirements.in
     #   dyff-audit
@@ -253,17 +253,17 @@ parso==0.8.5
     # via jedi
 pexpect==4.9.0
     # via ipython
-pillow==12.0.0
+pillow==12.1.0
     # via -r requirements.in
-platformdirs==4.5.0
+platformdirs==4.5.1
     # via jupyter-core
 ply==3.11
     # via jsonpath-ng
-prometheus-client==0.23.1
+prometheus-client==0.24.1
     # via jupyter-server
 prompt-toolkit==3.0.52
     # via ipython
-psutil==7.1.3
+psutil==7.2.1
     # via ipykernel
 ptyprocess==0.7.0
     # via
@@ -271,15 +271,15 @@ ptyprocess==0.7.0
     #   terminado
 pure-eval==0.2.3
     # via stack-data
-pyarrow==22.0.0
+pyarrow==23.0.0
     # via
     #   -r requirements.in
     #   dyff-audit
     #   dyff-client
     #   dyff-schema
-pycparser==2.23
+pycparser==3.0
     # via cffi
-pydantic==2.12.4
+pydantic==2.12.5
     # via
     #   -r requirements.in
     #   dyff-audit
@@ -308,10 +308,8 @@ python-dotenv==1.2.1
     #   uvicorn
 python-json-logger==4.0.0
     # via jupyter-events
-python-multipart==0.0.20
+python-multipart==0.0.21
     # via -r requirements.in
-pytz==2025.2
-    # via pandas
 pyyaml==6.0.3
     # via
     #   huggingface-hub
@@ -328,7 +326,7 @@ referencing==0.37.0
     #   jsonschema
     #   jsonschema-specifications
     #   jupyter-events
-regex==2025.11.3
+regex==2026.1.15
     # via transformers
 requests==2.32.5
     # via
@@ -348,33 +346,29 @@ rfc3986-validator==0.1.1
     #   jupyter-events
 rfc3987-syntax==1.1.0
     # via jsonschema
-rpds-py==0.29.0
+rpds-py==0.30.0
     # via
     #   jsonschema
     #   referencing
-ruamel-yaml==0.18.16
+ruamel-yaml==0.19.1
     # via dyff-audit
-ruamel-yaml-clib==0.2.15
-    # via ruamel-yaml
-safetensors==0.6.2
+safetensors==0.7.0
     # via transformers
-send2trash==1.8.3
+send2trash==2.1.0
     # via jupyter-server
-setuptools==80.9.0
+setuptools==80.10.1
     # via jupyterlab
 six==1.17.0
     # via
     #   python-dateutil
     #   rfc3339-validator
-sniffio==1.3.1
-    # via anyio
 sortedcontainers==2.4.0
     # via hypothesis
-soupsieve==2.8
+soupsieve==2.8.3
     # via beautifulsoup4
 stack-data==0.6.3
     # via ipython
-starlette==0.49.3
+starlette==0.50.0
     # via fastapi
 terminado==0.18.1
     # via
@@ -382,9 +376,9 @@ terminado==0.18.1
     #   jupyter-server-terminals
 tinycss2==1.4.0
     # via bleach
-tokenizers==0.22.1
+tokenizers==0.22.2
     # via transformers
-tornado==6.5.2
+tornado==6.5.4
     # via
     #   ipykernel
     #   jupyter-client
@@ -410,7 +404,7 @@ traitlets==5.14.3
     #   nbclient
     #   nbconvert
     #   nbformat
-transformers==4.57.1
+transformers==4.57.6
     # via -r requirements.in
 typing-extensions==4.15.0
     # via
@@ -428,21 +422,19 @@ typing-inspection==0.4.2
     # via
     #   pydantic
     #   pydantic-settings
-tzdata==2025.2
-    # via
-    #   arrow
-    #   pandas
+tzdata==2025.3
+    # via arrow
 uri-template==1.3.0
     # via jsonschema
-urllib3==2.5.0
+urllib3==2.6.3
     # via requests
-uvicorn==0.38.0
+uvicorn==0.40.0
     # via -r requirements.in
 uvloop==0.22.1
     # via uvicorn
 watchfiles==1.1.1
     # via uvicorn
-wcwidth==0.2.14
+wcwidth==0.3.0
     # via prompt-toolkit
 webcolors==25.10.0
     # via jsonschema
@@ -452,7 +444,7 @@ webencodings==0.5.1
     #   tinycss2
 websocket-client==1.9.0
     # via jupyter-server
-websockets==15.0.1
+websockets==16.0
     # via
     #   dyff-client
     #   uvicorn

requirements.in

@@ -1,6 +1,6 @@
 dyff-audit
 dyff-client
-dyff-schema>=0.39.1
+dyff-schema>=0.43.4
 
 click
 

requirements.torch.cpu.txt

@@ -1,3 +1,3 @@
 --index-url https://download.pytorch.org/whl/cpu
-torch==2.9.1+cpu
-torchvision==0.24.1+cpu
+torch==2.10.0+cpu
+torchvision==0.25.0+cpu

requirements.torch.gpu.txt

@@ -1,8 +1,12 @@
 # This file was autogenerated by uv via the following command:
 #    uv pip compile --python-version 3.12 -o requirements.torch.gpu.txt requirements.torch.gpu.in
-filelock==3.20.0
+cuda-bindings==12.9.4
     # via torch
-fsspec==2025.10.0
+cuda-pathfinder==1.3.3
+    # via cuda-bindings
+filelock==3.20.3
+    # via torch
+fsspec==2026.1.0
     # via torch
 jinja2==3.1.6
     # via torch
@@ -10,9 +14,9 @@ markupsafe==3.0.3
     # via jinja2
 mpmath==1.3.0
     # via sympy
-networkx==3.6
+networkx==3.6.1
     # via torch
-numpy==2.3.5
+numpy==2.4.1
     # via torchvision
 nvidia-cublas-cu12==12.8.4.1
     # via
@@ -49,23 +53,23 @@ nvidia-nvjitlink-cu12==12.8.93
     #   nvidia-cusolver-cu12
     #   nvidia-cusparse-cu12
     #   torch
-nvidia-nvshmem-cu12==3.3.20
+nvidia-nvshmem-cu12==3.4.5
     # via torch
 nvidia-nvtx-cu12==12.8.90
     # via torch
-pillow==12.0.0
+pillow==12.1.0
     # via torchvision
-setuptools==80.9.0
+setuptools==80.10.1
     # via torch
 sympy==1.14.0
     # via torch
-torch==2.9.1
+torch==2.10.0
     # via
     #   -r requirements.torch.gpu.in
     #   torchvision
-torchvision==0.24.1
+torchvision==0.25.0
     # via -r requirements.torch.gpu.in
-triton==3.5.1
+triton==3.6.0
     # via torch
 typing-extensions==4.15.0
     # via torch