Update links to public l5kit resources (#421)

* Update links in colab setup

* Update links in train notebooks

* Update requirements

examples/planning/train.ipynb

@@ -290,10 +290,10 @@
  "\n",
  "## Pre-trained models\n",
  "we provide a collection of pre-trained models for the planning task:\n",
- "- [model](https://lyft-l5-datasets-public.s3-us-west-2.amazonaws.com/models/planning_models/planning_model_20201208.pt) trained on train.zarr for 15 epochs;\n",
- "- [model](https://lyft-l5-datasets-public.s3-us-west-2.amazonaws.com/models/planning_models/planning_model_20201208_early.pt) trained on train.zarr for 2 epochs;\n",
- "- [model](https://lyft-l5-datasets-public.s3-us-west-2.amazonaws.com/models/planning_models/planning_model_20201208_nopt.pt) trained on train.zarr with perturbations disabled for 15 epochs;\n",
- "- [model](https://lyft-l5-datasets-public.s3-us-west-2.amazonaws.com/models/planning_models/planning_model_20201208_nopt_early.pt) trained on train.zarr with perturbations disabled for 2 epochs;\n",
+ "- [model](https://d20lyvjneielsk.cloudfront.net/planning_model_20201208.pt) trained on train.zarr for 15 epochs;\n",
+ "- [model](https://d20lyvjneielsk.cloudfront.net/planning_model_20201208_early.pt) trained on train.zarr for 2 epochs;\n",
+ "- [model](https://d20lyvjneielsk.cloudfront.net/planning_model_20201208_nopt.pt) trained on train.zarr with perturbations disabled for 15 epochs;\n",
+ "- [model](https://d20lyvjneielsk.cloudfront.net/planning_model_20201208_nopt_early.pt) trained on train.zarr with perturbations disabled for 2 epochs;\n",
  "\n",
  "We include two partially trained models to emphasise the important role of perturbations during training, especially during the first stage of training.\n",
  "\n",

examples/setup_notebook_colab.sh

@@ -6,23 +6,23 @@ TEMP_DATASET_DIR=$(mktemp -d)
 
 # Download sample zarr
 echo "Downloading sample zarr dataset..."
-wget https://lyft-l5-datasets-public.s3-us-west-2.amazonaws.com/prediction/v1.1/sample.tar \
+wget https://d20lyvjneielsk.cloudfront.net/prediction-sample.tar \
  -q --show-progress -P $TEMP_DOWNLOAD_DIR
 
 mkdir -p $TEMP_DATASET_DIR/scenes
 tar xf $TEMP_DOWNLOAD_DIR/sample.tar -C $TEMP_DATASET_DIR/scenes
 
 # Download semantic map
 echo "Downloading semantic map..."
-wget https://lyft-l5-datasets-public.s3-us-west-2.amazonaws.com/prediction/v1.1/semantic_map.tar \
+wget https://d20lyvjneielsk.cloudfront.net/prediction-semantic_map.tar \
  -q --show-progress -P $TEMP_DOWNLOAD_DIR
 mkdir -p $TEMP_DATASET_DIR/semantic_map
 tar xf $TEMP_DOWNLOAD_DIR/semantic_map.tar -C $TEMP_DATASET_DIR/semantic_map
 cp $TEMP_DATASET_DIR/semantic_map/meta.json $TEMP_DATASET_DIR/meta.json
 
 # Download aerial maps
 echo "Downloading aerial maps (this can take a while)..."
-wget https://lyft-l5-datasets-public.s3-us-west-2.amazonaws.com/prediction/v1.1/aerial_map.tar \
+wget https://d20lyvjneielsk.cloudfront.net/prediction-aerial_map.tar \
  -q --show-progress -P $TEMP_DOWNLOAD_DIR
 tar xf $TEMP_DOWNLOAD_DIR/aerial_map.tar -C $TEMP_DATASET_DIR
 

examples/simulation/train.ipynb

@@ -257,8 +257,8 @@
  "\n",
  "## Pre-trained models\n",
  "we provide a collection of pre-trained models for the simulation task:\n",
- "- [simulation model](https://lyft-l5-datasets-public.s3-us-west-2.amazonaws.com/models/simulation_models/simulation_model_20210416_5steps.pt) trained on agents over the semantic rasteriser with history of 0.5s;\n",
- "- [planning model](https://lyft-l5-datasets-public.s3-us-west-2.amazonaws.com/models/simulation_models/planning_model_20210421_5steps.pt) trained on the AV over the semantic rasteriser with history of 0.5s;\n",
+ "- [simulation model](https://d20lyvjneielsk.cloudfront.net/simulation_model_20210416_5steps.pt) trained on agents over the semantic rasteriser with history of 0.5s;\n",
+ "- [planning model](https://d20lyvjneielsk.cloudfront.net/planning_model_20210421_5steps.pt) trained on the AV over the semantic rasteriser with history of 0.5s;\n",
  "\n",
  "To use one of the models simply download the corresponding `.pt` file and load it in the evaluation notebooks."
  ]

examples/urban_driver/train.ipynb

@@ -351,10 +351,10 @@
  "\n",
  "## Pre-trained models\n",
  "We provide a collection of pre-trained models, including both our proposed method and several ablations from our paper:\n",
- "- [Urban Driver](https://lyft-l5-datasets-public.s3-us-west-2.amazonaws.com/models/urban_driver/BPTT.pt);\n",
- "- [Urban Driver without BPTT](https://lyft-l5-datasets-public.s3-us-west-2.amazonaws.com/models/urban_driver/MS.pt);\n",
- "- [Open Loop](https://lyft-l5-datasets-public.s3-us-west-2.amazonaws.com/models/urban_driver/OL.pt);\n",
- "- [Open Loop with history](https://lyft-l5-datasets-public.s3-us-west-2.amazonaws.com/models/urban_driver/OL_HS.pt);\n",
+ "- [Urban Driver](https://d20lyvjneielsk.cloudfront.net/BPTT.pt);\n",
+ "- [Urban Driver without BPTT](https://d20lyvjneielsk.cloudfront.net/MS.pt);\n",
+ "- [Open Loop](https://d20lyvjneielsk.cloudfront.net/OL.pt);\n",
+ "- [Open Loop with history](https://d20lyvjneielsk.cloudfront.net/OL_HS.pt);\n",
  "\n",
  "To use one of the models simply download the corresponding `.pt` file and load it in the evaluation notebooks - further, please see the comments above regarding necessary config changes."
  ]

l5kit/Pipfile

@@ -20,7 +20,7 @@ types-pyyaml = "*"
 l5kit = {editable = true, path = "."}
 zarr = "*"
 numpy = "*"
-opencv-contrib-python-headless = "*"
+opencv-contrib-python-headless = "<=4.5.3.56"
 pymap3d = "*"
 scipy = "*"
 tqdm = "*"

l5kit/setup.py

@@ -16,7 +16,7 @@
  "imageio",
  "matplotlib",
  "numpy~=1.19.0",
- "opencv-contrib-python-headless",
+ "opencv-contrib-python-headless<4.8.0",
  "protobuf>=3.12.2,<=3.20",
  "pymap3d",
  "scipy",