update readme and dependency

Summary: resolve facebookresearch#3

Reviewed By: rbgirshick

Differential Revision: D17872890

fbshipit-source-id: 4222bafc3d83da85d931b937d45064eae2a1fd4f

Author: ppwwyyxx

INSTALL.md

@@ -27,6 +27,7 @@ python setup.py build develop
 # or, as an alternative to `setup.py`, do
 # pip install .
 ```
+Note: you may need to rebuild detectron2 after reinstalling a different build of PyTorch.
 
 ### Common Installation Issues
 

README.md

@@ -12,7 +12,7 @@ learning framework.
   <img src="https://user-images.githubusercontent.com/1381301/66535560-d3422200-eace-11e9-9123-5535d469db19.png"/>
 </div>
 
-See our [blog post](https://ai.facebook.com/blog/-detectron2-a-pytorch-based-modular-object-detection-library-/)
+Check our [blog post](https://ai.facebook.com/blog/-detectron2-a-pytorch-based-modular-object-detection-library-/)
 to see more demos and learn what's new in detectron2.
 
 ## Installation

configs/quick_schedules/README.md

@@ -1,51 +1 @@
 These are quick configs for performance or accuracy regression tracking purposes.
-
-## Perf testing configs:
-
-### Inference
-
-Reference devgpu configuration:
-
- - 48 core Intel(R) Xeon(R) CPU E5-2680 v3 @ 2.50GHz
- - 2x M40 (12GB)
- - buck build @mode/dev-nosan
-
-configs/quick_schedules/mask_rcnn_R_50_C4_inference_acc_test.yaml
-```
-# Before https://github.com/fairinternal/detectron2/pull/84
-Total inference time: 0:00:30.808294 (0.6161658811569214 s / img per device, on 2 devices)
-# After https://github.com/fairinternal/detectron2/pull/84
-Total inference time: 0:00:36.952044 (0.7390408849716187 s / img per device, on 2 devices)
-```
-
-configs/quick_schedules/mask_rcnn_R_50_FPN_inference_acc_test.yaml
-```
-# Before https://github.com/fairinternal/detectron2/pull/84
-Total inference time: 0:00:21.773355 (0.435467095375061 s / img per device, on 2 devices)
-# After https://github.com/fairinternal/detectron2/pull/84
-Total inference time: 0:00:28.766723 (0.5753344583511353 s / img per device, on 2 devices)
-```
-
-### Training
-
-TODO
-
-They are equivalent to the standard C4 / FPN models, only with extremely short schedules.
-
-Metrics to look at:
-
-```
-INFO: Total training time: 0:3:20.276231
-...
-INFO: Total inference time: 0:01:20.276231
-```
-
-
-## Accuracy testing configs:
-
-They are simplified versions of standard models, trained and tested on the same
-minival dataset, with short schedules.
-
-The schedule is designed to provide a stable enough mAP within minimal amount of training time.
-
-Metrics to look at: mAPs.

configs/semantic_R_101_FPN_1x.yaml

@@ -55,6 +55,7 @@ def launch(main_func, num_gpus_per_machine, num_machines=1, machine_rank=0, dist
 def _distributed_worker(
     local_rank, main_func, world_size, num_gpus_per_machine, machine_rank, dist_url, args
 ):
+    assert torch.cuda.is_available(), "cuda is not available. Please check your installation."
     global_rank = machine_rank * num_gpus_per_machine + local_rank
     try:
         dist.init_process_group(

detectron2/engine/launch.py

@@ -1,6 +1,8 @@
 # -*- coding: utf-8 -*-
 # Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved
 
+# flake8: noqa
+
 # Configuration file for the Sphinx documentation builder.
 #
 # This file does only contain a selection of the most common options. For a
@@ -48,6 +50,8 @@
         "torch.autograd.function",
         "torch.nn.modules",
         "torch.nn.modules.utils",
+        "torchvision",
+        "torchvision.ops",
     ]:
         sys.modules[m] = mock.Mock(name=m)
 

docs/conf.py

@@ -12,3 +12,4 @@ yacs
 tabulate
 git+git://github.com/facebookresearch/fvcore.git
 https://download.pytorch.org/whl/nightly/cpu/torch-1.3.0.dev20191010%2Bcpu-cp37-cp37m-linux_x86_64.whl
+https://download.pytorch.org/whl/nightly/cpu/torchvision-0.5.0.dev20191008%2Bcpu-cp37-cp37m-linux_x86_64.whl

docs/requirements.txt

@@ -60,8 +60,9 @@ def get_extensions():
     name="detectron2",
     version="0.1",
     author="FAIR",
-    url="unknown",
-    description="object detection in pytorch",
+    url="https://github.com/facebookresearch/detectron2",
+    description="Detectron2 is FAIR's next-generation research "
+        "platform for object detection and segmentation.",
     packages=find_packages(exclude=("configs", "tests")),
     install_requires=[
         "termcolor>=1.1",
@@ -72,6 +73,7 @@ def get_extensions():
         "matplotlib",
         "tqdm>4.29.0",
         "shapely",
+        "tensorboard",
     ],
     ext_modules=get_extensions(),
     cmdclass={"build_ext": torch.utils.cpp_extension.BuildExtension},