formatting fix (#2816)

Summary: # Before submitting - [ ] Was this discussed/approved via a Github issue? (no need for typos, doc improvements) - [ ] Did you read the [contributor guideline](https://github.com/pytorch/fairseq/blob/main/CONTRIBUTING.md)? - [ ] Did you make sure to update the docs? - [ ] Did you write any new necessary tests? ## What does this PR do? fix `black` failures ## PR review Anyone in the community is free to review the PR once the tests have passed. If we didn't discuss your PR in Github issues there's a high chance it will not be merged. ## Did you have fun? Make sure you had fun coding � Pull Request resolved: fairinternal/fairseq-py#2816 Reviewed By: alexeib Differential Revision: D33172615 Pulled By: dianaml0 fbshipit-source-id: 36b141f42941670f1bfa981041d878042feb0428
facebookresearch · Dec 17, 2021 · a540213 · a540213
1 parent 1e4055d
commit a540213
Show file tree

Hide file tree

Showing 9 changed files with 60 additions and 41 deletions.
diff --git a/fairseq/checkpoint_utils.py b/fairseq/checkpoint_utils.py
@@ -7,15 +7,18 @@
 import collections
 import contextlib
 import logging
-import numpy as np
 import os
 import re
 import time
 import traceback
 from collections import OrderedDict
+from pathlib import Path
 from typing import Any, Dict, Optional, Union
 
+import numpy as np
 import torch
+from omegaconf import DictConfig, OmegaConf, open_dict
+
 from fairseq.data import data_utils
 from fairseq.dataclass.configs import CheckpointConfig
 from fairseq.dataclass.utils import (
@@ -25,10 +28,6 @@
 from fairseq.distributed.fully_sharded_data_parallel import FSDP, has_FSDP
 from fairseq.file_io import PathManager
 from fairseq.models import FairseqDecoder, FairseqEncoder
-from omegaconf import DictConfig, open_dict, OmegaConf
-
-from pathlib import Path
-
 
 logger = logging.getLogger(__name__)
 
@@ -509,7 +508,7 @@ def load_model_ensemble_and_task_from_hf_hub(
     _arg_overrides["data"] = cache_dir
     return load_model_ensemble_and_task(
         [p.as_posix() for p in Path(cache_dir).glob("*.pt")],
-        arg_overrides=_arg_overrides
+        arg_overrides=_arg_overrides,
     )
 
 

diff --git a/fairseq/models/hubert/hubert.py b/fairseq/models/hubert/hubert.py
@@ -4,13 +4,14 @@
 # LICENSE file in the root directory of this source tree.
 
 import logging
+from dataclasses import dataclass, field
 from typing import Dict, List, Optional, Tuple
 
 import numpy as np
-
 import torch
 import torch.nn as nn
-from dataclasses import dataclass, field
+from omegaconf import II
+
 from fairseq import utils
 from fairseq.data.data_utils import compute_mask_indices
 from fairseq.data.dictionary import Dictionary
@@ -25,7 +26,6 @@
     HubertPretrainingConfig,
     HubertPretrainingTask,
 )
-from omegaconf import II
 
 logger = logging.getLogger(__name__)
 
@@ -205,9 +205,11 @@ class HubertConfig(FairseqDataclass):
     )
 
     checkpoint_activations: bool = field(
-        default=False, metadata={"help": "recompute activations and save memory for extra compute"}
+        default=False,
+        metadata={"help": "recompute activations and save memory for extra compute"},
     )
 
+
 @register_model("hubert", dataclass=HubertConfig)
 class HubertModel(BaseFairseqModel):
     def __init__(

diff --git a/fairseq/modules/layer_norm.py b/fairseq/modules/layer_norm.py
@@ -21,7 +21,6 @@ def forward(self, x):
                 with torch.cuda.device(x.device):
                     return super().forward(x)
 
-
 except ImportError:
     has_fused_layernorm = False
 

diff --git a/tests/speech/__init__.py b/tests/speech/__init__.py
@@ -4,8 +4,8 @@
 # LICENSE file in the root directory of this source tree.
 
 import os
-from pathlib import Path
 import unittest
+from pathlib import Path
 
 import torch
 
@@ -31,7 +31,7 @@ def set_up_librispeech(self):
             "spm_librispeech_unigram10000.model",
             "spm_librispeech_unigram10000.txt",
             "librispeech_test-other.tsv",
-            "librispeech_test-other.zip"
+            "librispeech_test-other.zip",
         ]
         self.base_url = f"{S3_BASE_URL}/s2t/librispeech"
         for filename in self.data_filenames:
@@ -47,7 +47,7 @@ def set_up_ljspeech(self):
             "ljspeech_g2p_gcmvn_stats.npz",
             "ljspeech_g2p.txt",
             "ljspeech_test.tsv",
-            "ljspeech_test.zip"
+            "ljspeech_test.zip",
         ]
         self.base_url = f"{S3_BASE_URL}/s2/ljspeech"
         for filename in self.data_filenames:

diff --git a/tests/speech/test_fastspeech2.py b/tests/speech/test_fastspeech2.py
@@ -8,8 +8,8 @@
 import torch
 from tqdm import tqdm
 
-from fairseq.checkpoint_utils import load_model_ensemble_and_task
 from fairseq import utils
+from fairseq.checkpoint_utils import load_model_ensemble_and_task
 from fairseq.tasks.text_to_speech import batch_mel_cepstral_distortion
 from tests.speech import TestFairseqSpeech
 
@@ -25,12 +25,13 @@ def test_ljspeech_fastspeech2_checkpoint(self):
         path = self.download(self.base_url, self.root, checkpoint_filename)
 
         models, cfg, task = load_model_ensemble_and_task(
-            [path.as_posix()], arg_overrides={
+            [path.as_posix()],
+            arg_overrides={
                 "data": self.root.as_posix(),
                 "config_yaml": "cfg_ljspeech_g2p.yaml",
                 "vocoder": "griffin_lim",
-                "fp16": False
-            }
+                "fp16": False,
+            },
         )
         if self.use_cuda:
             for model in models:
@@ -40,19 +41,21 @@ def test_ljspeech_fastspeech2_checkpoint(self):
         task.load_dataset(test_split)
         batch_iterator = task.get_batch_iterator(
             dataset=task.dataset(test_split),
-            max_tokens=65_536, max_positions=4_096, num_workers=1
+            max_tokens=65_536,
+            max_positions=4_096,
+            num_workers=1,
         ).next_epoch_itr(shuffle=False)
         progress = tqdm(batch_iterator, total=len(batch_iterator))
         generator = task.build_generator(models, cfg)
 
-        mcd, n_samples = 0., 0
+        mcd, n_samples = 0.0, 0
         for sample in progress:
             sample = utils.move_to_cuda(sample) if self.use_cuda else sample
             hypos = generator.generate(models[0], sample, has_targ=True)
             rets = batch_mel_cepstral_distortion(
                 [hypo["targ_waveform"] for hypo in hypos],
                 [hypo["waveform"] for hypo in hypos],
-                sr=task.sr
+                sr=task.sr,
             )
             mcd += sum(d.item() for d, _ in rets)
             n_samples += len(sample["id"].tolist())

diff --git a/tests/speech/test_s2t_transformer.py b/tests/speech/test_s2t_transformer.py
@@ -3,14 +3,14 @@
 # This source code is licensed under the MIT license found in the
 # LICENSE file in the root directory of this source tree.
 
-from argparse import Namespace
 import unittest
+from argparse import Namespace
 
 import torch
 from tqdm import tqdm
 
-from fairseq.checkpoint_utils import load_model_ensemble_and_task
 from fairseq import utils
+from fairseq.checkpoint_utils import load_model_ensemble_and_task
 from fairseq.scoring.wer import WerScorer
 from tests.speech import TestFairseqSpeech
 
@@ -26,10 +26,11 @@ def test_librispeech_s2t_transformer_s_checkpoint(self):
         path = self.download(self.base_url, self.root, checkpoint_filename)
 
         models, cfg, task = load_model_ensemble_and_task(
-            [path.as_posix()], arg_overrides={
+            [path.as_posix()],
+            arg_overrides={
                 "data": self.root.as_posix(),
-                "config_yaml": "cfg_librispeech.yaml"
-            }
+                "config_yaml": "cfg_librispeech.yaml",
+            },
         )
         if self.use_cuda:
             for model in models:
@@ -38,21 +39,33 @@ def test_librispeech_s2t_transformer_s_checkpoint(self):
         test_split = "librispeech_test-other"
         task.load_dataset(test_split)
         batch_iterator = task.get_batch_iterator(
-                dataset=task.dataset(test_split),
-                max_tokens=65_536, max_positions=(4_096, 1_024), num_workers=1
-            ).next_epoch_itr(shuffle=False)
+            dataset=task.dataset(test_split),
+            max_tokens=65_536,
+            max_positions=(4_096, 1_024),
+            num_workers=1,
+        ).next_epoch_itr(shuffle=False)
 
-        scorer_args = {"wer_tokenizer": "none", "wer_lowercase": False,
-                       "wer_remove_punct": False, "wer_char_level": False}
+        scorer_args = {
+            "wer_tokenizer": "none",
+            "wer_lowercase": False,
+            "wer_remove_punct": False,
+            "wer_char_level": False,
+        }
         scorer = WerScorer(Namespace(**scorer_args))
         progress = tqdm(enumerate(batch_iterator), total=len(batch_iterator))
         for batch_idx, sample in progress:
             sample = utils.move_to_cuda(sample) if self.use_cuda else sample
             hypo = task.inference_step(generator, models, sample)
             for i, sample_id in enumerate(sample["id"].tolist()):
-                tgt_tokens = utils.strip_pad(sample["target"][i, :], task.tgt_dict.pad()).int().cpu()
+                tgt_tokens = (
+                    utils.strip_pad(sample["target"][i, :], task.tgt_dict.pad())
+                    .int()
+                    .cpu()
+                )
                 tgt_str = task.tgt_dict.string(tgt_tokens, "sentencepiece")
-                hypo_str = task.tgt_dict.string(hypo[i][0]["tokens"].int().cpu(), "sentencepiece")
+                hypo_str = task.tgt_dict.string(
+                    hypo[i][0]["tokens"].int().cpu(), "sentencepiece"
+                )
                 if batch_idx == 0 and i < 3:
                     print(f"T-{sample_id} {tgt_str}")
                     print(f"H-{sample_id} {hypo_str}")

diff --git a/tests/speech/test_tts_transformer.py b/tests/speech/test_tts_transformer.py
@@ -8,8 +8,8 @@
 import torch
 from tqdm import tqdm
 
-from fairseq.checkpoint_utils import load_model_ensemble_and_task
 from fairseq import utils
+from fairseq.checkpoint_utils import load_model_ensemble_and_task
 from fairseq.tasks.text_to_speech import batch_mel_cepstral_distortion
 from tests.speech import TestFairseqSpeech
 
@@ -25,12 +25,13 @@ def test_ljspeech_tts_transformer_checkpoint(self):
         path = self.download(self.base_url, self.root, checkpoint_filename)
 
         models, cfg, task = load_model_ensemble_and_task(
-            [path.as_posix()], arg_overrides={
+            [path.as_posix()],
+            arg_overrides={
                 "data": self.root.as_posix(),
                 "config_yaml": "cfg_ljspeech_g2p.yaml",
                 "vocoder": "griffin_lim",
-                "fp16": False
-            }
+                "fp16": False,
+            },
         )
         if self.use_cuda:
             for model in models:
@@ -40,19 +41,21 @@ def test_ljspeech_tts_transformer_checkpoint(self):
         task.load_dataset(test_split)
         batch_iterator = task.get_batch_iterator(
             dataset=task.dataset(test_split),
-            max_tokens=65_536, max_positions=768, num_workers=1
+            max_tokens=65_536,
+            max_positions=768,
+            num_workers=1,
         ).next_epoch_itr(shuffle=False)
         progress = tqdm(batch_iterator, total=len(batch_iterator))
         generator = task.build_generator(models, cfg)
 
-        mcd, n_samples = 0., 0
+        mcd, n_samples = 0.0, 0
         for sample in progress:
             sample = utils.move_to_cuda(sample) if self.use_cuda else sample
             hypos = generator.generate(models[0], sample, has_targ=True)
             rets = batch_mel_cepstral_distortion(
                 [hypo["targ_waveform"] for hypo in hypos],
                 [hypo["waveform"] for hypo in hypos],
-                sr=task.sr
+                sr=task.sr,
             )
             mcd += sum(d.item() for d, _ in rets)
             n_samples += len(sample["id"].tolist())

diff --git a/tests/test_hf_hub.py b/tests/test_hf_hub.py
@@ -7,6 +7,7 @@
 import unittest
 
 import torch
+
 try:
     import huggingface_hub
 except ImportError:

diff --git a/tests/test_sequence_generator.py b/tests/test_sequence_generator.py
@@ -562,7 +562,6 @@ def test_diverse_beam_search(self):
         self.assertHypoScore(hypos[1][1], [0.7, 0.35, 0.9], [0, 2, 1], 0.5)
 
 
-
 class TestTopPSamplingSearch(TestSequenceGeneratorBase):
     def setUp(self):
         # construct dummy dictionary