Add fairseq to PyPI (#495)

Summary:
- fairseq can now be installed via pip: `pip install fairseq`
- command-line tools are globally accessible: `fairseq-preprocess`, `fairseq-train`, `fairseq-generate`, etc.
Pull Request resolved: https://github.com/pytorch/fairseq/pull/495

Differential Revision: D14017761

Pulled By: myleott

fbshipit-source-id: 10c9f6634a3056074eac2f33324b4f1f404d4235

fairseq_cli/generate.py

+../generate.py
\ No newline at end of file

fairseq_cli/interactive.py

+../interactive.py
\ No newline at end of file

fairseq_cli/preprocess.py

+../preprocess.py
\ No newline at end of file

fairseq_cli/score.py

+../score.py
\ No newline at end of file

fairseq_cli/setup.py

+../setup.py
\ No newline at end of file

fairseq_cli/train.py

+../train.py
\ No newline at end of file

requirements.txt

-cffi
-numpy
-torch
-tqdm

scripts/compound_split_bleu.sh

@@ -17,4 +17,4 @@ fi
 
 grep ^H $GEN | cut -f3- | perl -ple 's{(\S)-(\S)}{$1 ##AT##-##AT## $2}g' > $SYS
 grep ^T $GEN | cut -f2- | perl -ple 's{(\S)-(\S)}{$1 ##AT##-##AT## $2}g' > $REF
-python score.py --sys $SYS --ref $REF
+fairseq-score --sys $SYS --ref $REF

setup.py

@@ -16,12 +16,6 @@ if sys.version_info < (3,):
 with open('README.md') as f:
     readme = f.read()
 
-with open('LICENSE') as f:
-    license = f.read()
-
-with open('requirements.txt') as f:
-    reqs = f.read()
-
 
 bleu = Extension(
     'fairseq.libbleu',
@@ -35,22 +29,33 @@ bleu = Extension(
 
 setup(
     name='fairseq',
-    version='0.6.0',
+    version='0.6.1',
     description='Facebook AI Research Sequence-to-Sequence Toolkit',
+    url='https://github.com/pytorch/fairseq',
+    classifiers=[
+        'Intended Audience :: Science/Research',
+        'License :: OSI Approved :: BSD License',
+        'Programming Language :: Python :: 3.6',
+        'Topic :: Scientific/Engineering :: Artificial Intelligence',
+    ],
     long_description=readme,
-    license=license,
-    install_requires=reqs.strip().split('\n'),
-    packages=find_packages(),
+    install_requires=[
+        'cffi',
+        'numpy',
+        'torch',
+        'tqdm',
+    ],
+    packages=find_packages(exclude=['scripts', 'tests']),
     ext_modules=[bleu],
     test_suite='tests',
     entry_points={
         'console_scripts': [
-            'fairseq-eval-lm = eval_lm:cli_main',
-            'fairseq-generate = generate:cli_main',
-            'fairseq-interactive = interactive:cli_main',
-            'fairseq-preprocess = preprocess:cli_main',
-            'fairseq-train = train:cli_main',
-            'fairseq-score = score:main',
+            'fairseq-eval-lm = fairseq_cli.eval_lm:cli_main',
+            'fairseq-generate = fairseq_cli.generate:cli_main',
+            'fairseq-interactive = fairseq_cli.interactive:cli_main',
+            'fairseq-preprocess = fairseq_cli.preprocess:cli_main',
+            'fairseq-train = fairseq_cli.train:cli_main',
+            'fairseq-score = fairseq_cli.score:main',
         ],
     },
 )

train.py

@@ -395,18 +395,8 @@ def cli_main():
         port = random.randint(10000, 20000)
         args.distributed_init_method = 'tcp://localhost:{port}'.format(port=port)
         args.distributed_rank = None  # set based on device id
-        print(
-            '''| NOTE: you may get better performance with:
-
-            python -m torch.distributed.launch --nproc_per_node {ngpu} train.py {no_c10d}(...)
-            '''.format(
-                ngpu=args.distributed_world_size,
-                no_c10d=(
-                    '--ddp-backend=no_c10d ' if max(args.update_freq) > 1 and args.ddp_backend != 'no_c10d'
-                    else ''
-                ),
-            )
-        )
+        if max(args.update_freq) > 1 and args.ddp_backend != 'no_c10d':
+            print('| NOTE: you may get better performance with: --ddp-backend=no_c10d')
         torch.multiprocessing.spawn(
             fn=distributed_main,
             args=(args, ),