update alternative Dockerfiles for different CUDA versions (#170)

* update alternative Dockerfiles for different CUDA versions

* update comment

CHANGELOG.md

@@ -7,6 +7,10 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## Unreleased
 
+### Changed
+
+- Updated `Dockerfile.release` and `Dockerfile.commit` to work with different CUDA versions.
+
 
 ## [v1.2.2](https://github.com/allenai/allennlp-models/releases/tag/v1.2.2) - 2020-11-17
 

Dockerfile.commit

@@ -1,10 +1,13 @@
 # This Dockerfile is used to build an image from specific commits of allennlp and
-# allennlp-models. It requires two build args: ALLENNLP_COMMIT and ALLENNLP_MODELS_COMMIT,
-# which should be set to the desired commit SHAs for each repo, respectively.
+# allennlp-models. It requires three build args: ALLENNLP_COMMIT, ALLENNLP_MODELS_COMMIT,
+# and CUDA.
+# ALLENNLP_COMMIT and ALLENNLP_MODELS_COMMIT should be set to the desired commit SHAs for each repo.
+# CUDA should be set to a supported CUDA version such as '10.2' or '11.0'.
 
 ARG ALLENNLP_COMMIT
+ARG CUDA
 
-FROM allennlp/commit:${ALLENNLP_COMMIT}
+FROM allennlp/commit:${ALLENNLP_COMMIT}-cuda${CUDA}
 
 ARG ALLENNLP_MODELS_COMMIT
 

Dockerfile.release

@@ -1,13 +1,15 @@
 # This Dockerfile is used to build an image from a specific PyPI release of allennlp and
-# allennlp-models. It requires a single build arg, ALLENNLP_VERSION, which should
-# be the PyPI release version (allennlp and allennlp-models are released together
-# with the same version). For example, you could set ALLENNLP_VERSION=1.0.0rc3.
+# allennlp-models.
+# It requires two build args: RELEASE and CUDA.
+# RELEASE should be a allennlp/allennlp-models version, such as '1.2.2',
+# CUDA should be a supported CUDA version, such '10.2'.
 
-ARG ALLENNLP_VERSION
+ARG RELEASE
+ARG CUDA
 
-FROM allennlp/allennlp:v${ALLENNLP_VERSION}
+FROM allennlp/allennlp:v${RELEASE}-cuda${CUDA}
 
 # Need to specify this ARG again because the FROM stage consumes all args before it.
-ARG ALLENNLP_VERSION
+ARG RELEASE
 
-RUN pip install --no-cache-dir allennlp-models==${ALLENNLP_VERSION}
+RUN pip install --no-cache-dir allennlp-models==${RELEASE}

README.md

@@ -70,32 +70,36 @@ environment to a compute cluster.
 
 Once you have [installed Docker](https://docs.docker.com/engine/installation/) you can either use a [prebuilt image from a release](https://hub.docker.com/r/allennlp/models) or build an image locally with any version of `allennlp` and `allennlp-models`.
 
+If you have GPUs available, you also need to install the [nvidia-docker](https://github.com/NVIDIA/nvidia-docker) runtime.
+
 To build an image locally from a specific release, run
 
 ```bash
 docker build \
-    --build-arg ALLENNLP_VERSION=1.0.0rc3 \
+    --build-arg RELEASE=1.2.2 \
+    --build-arg CUDA=10.2 \
     -t allennlp/models - < Dockerfile.release
 ```
 
-Just replace "1.0.0rc3" with the desired version.
+Just replace the `RELEASE` and `CUDA` build args what you need. Currently only CUDA 10.2 and 11.0 are officially supported.
 
 Alternatively, you can build against specific commits of `allennlp` and `allennlp-models` with
 
 ```bash
 docker build \
-    --build-arg ALLENNLP_COMMIT=e3d72fcb1664caf9554ef4e611191c33a7a5cbbd \
-    --build-arg ALLENNLP_MODELS_COMMIT=54a5df89da64d8d3869e746bc6dab940552dbfc4 \
+    --build-arg ALLENNLP_COMMIT=d823a2591e94912a6315e429d0fe0ee2efb4b3ee \
+    --build-arg ALLENNLP_MODELS_COMMIT=01bc777e0d89387f03037d398cd967390716daf1 \
+    --build-arg CUDA=10.2 \
     -t allennlp/models - < Dockerfile.commit
 ```
 
-Just change the `ALLENNLP_COMMIT` and `ALLENNLP_MODELS_COMMIT` build args to the desired commit SHAs.
+Just change the `ALLENNLP_COMMIT` / `ALLENNLP_MODELS_COMMIT` and `CUDA` build args to the desired commit SHAs and CUDA versions, respectively.
 
-Now run the following command to get an environment that will run on either the cpu or gpu.
+Once you've built your image, you can run it like this:
 
 ```bash
 mkdir -p $HOME/.allennlp/
-docker run --rm -v $HOME/.allennlp:/root/.allennlp allennlp/models
+docker run --rm --gpus all -v $HOME/.allennlp:/root/.allennlp allennlp/models
 ```
 
-If you have GPUs available, you can install the [nvidia-docker](https://github.com/NVIDIA/nvidia-docker) runtime and then add the flag `--gpus all` right before `--rm`.
+> Note: the `--gpus all` is only valid if you've installed the nvidia-docker runtime.