2025-09-09 nightly release (fa51429)

Author: pytorchbot

.github/workflows/lint.yml

@@ -1,43 +1,43 @@
-# name: Lint
+name: Lint
 
-# on:
-#   pull_request:
-#   push:
-#     branches:
-#       - nightly
-#       - main
-#       - release/*
-#   workflow_dispatch:
+on:
+  push:
+    branches: [ main ]
+  pull_request:
 
-# jobs:
-#   python-source-and-configs:
-#     uses: pytorch/test-infra/.github/workflows/linux_job_v2.yml@main
-#     permissions:
-#       id-token: write
-#       contents: read
-#     with:
-#       repository: pytorch/audio
-#       script: |
-#         set -euo pipefail
+concurrency:
+  group: unit-test${{ github.workflow }}-${{ github.ref == 'refs/heads/main' && github.run_number || github.ref }}
+  cancel-in-progress: true
 
-#         echo '::group::Setup environment'
-#         eval "$("$(which conda)" shell.bash hook)"
-#         # libcst does not have 3.11 pre-built binaries yet. Use python 3.10
-#         conda create -y --name env python=3.10
-#         conda activate env
-#         pip3 install --progress-bar=off pre-commit
-#         echo '::endgroup::'
+defaults:
+  run:
+    shell: bash -l -eo pipefail {0}
 
-#         set +e
-#         pre-commit run --all-files --show-diff-on-failure
-#         status=$?
-
-#         echo '::group::Add Summry'
-#         if [ $status -ne 0 ]; then
-#           echo '### Lint failure'  >> $GITHUB_STEP_SUMMARY
-#           echo '```diff'           >> $GITHUB_STEP_SUMMARY
-#           git --no-pager diff      >> $GITHUB_STEP_SUMMARY
-#           echo '```'               >> $GITHUB_STEP_SUMMARY
-#         fi
-#         echo '::endgroup::'
-#         exit $status
+jobs:
+  pre-commit-checks:
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        python-version: ['3.11']
+    steps:
+      - name: Check out repo
+        uses: actions/checkout@v3
+      - name: Setup conda env
+        uses: conda-incubator/setup-miniconda@v2
+        with:
+          auto-update-conda: true
+          miniconda-version: "latest"
+          activate-environment: test
+          python-version: ${{ matrix.python-version }}
+      - name: Update pip
+        run: python -m pip install --upgrade pip
+      - name: Install pre-commit
+        run: |
+          python -m pip install pre-commit
+      - name: Run pre-commit checks
+        run: |
+          pre-commit run --all-files
+      - name: Check to see what files pre-commit modified
+        run: |
+          git diff

.github/workflows/pr-labels.yml

@@ -1,4 +1,6 @@
-# name: pr-labels
+name: pr-labels
+
+# This workflow is disabled and kept only for reference.
 
 # on:
 #   push:

cmake/LoadHIP.cmake

@@ -267,4 +267,3 @@ if(HIP_FOUND)
     set(roctracer_INCLUDE_DIRS ${ROCTRACER_PATH}/include)
   endif() 
 endif()
-

docs/source/conf.py

@@ -59,7 +59,7 @@
 nbsphinx_requirejs_path = ""
 
 autodoc_member_order = "bysource"
-autodoc_mock_imports = ['torchaudio.models.decoder']
+autodoc_mock_imports = ["torchaudio.models.decoder"]
 
 # katex options
 #

examples/tutorials/audio_feature_augmentation_tutorial.py

@@ -36,6 +36,7 @@
 def _get_sample(path):
     return torchaudio.load(path)
 
+
 def get_speech_sample():
     return _get_sample(SAMPLE_WAV_SPEECH_PATH)
 

examples/tutorials/audio_resampling_tutorial.py

@@ -26,7 +26,6 @@
 import math
 import timeit
 
-import matplotlib.colors as mcolors
 import matplotlib.pyplot as plt
 from IPython.display import Audio
 import numpy as np
@@ -427,7 +426,7 @@ def plot(data, cols, rows):
     x_data = np.arange(len(rows))
     bar_width = 0.8 / len(cols)
     for (i, (c, d)) in enumerate(zip(cols, data.T)):
-        x_pos = x_data + (i - len(cols)/2 + 0.5) * bar_width
+        x_pos = x_data + (i - len(cols) / 2 + 0.5) * bar_width
         ax.bar(x_pos, d, bar_width, label=c)
     ax.legend()
     ax.set_xticks(x_data)

examples/tutorials/mvdr_tutorial.py

@@ -128,6 +128,7 @@ def generate_mixture(waveform_clean, waveform_noise, target_snr):
     waveform_noise *= 10 ** (-(target_snr - current_snr) / 20)
     return waveform_clean + waveform_noise
 
+
 # If you have mir_eval installed, you can use it to evaluate the separation quality of the estimated sources.
 # You can also evaluate the intelligibility of the speech with the Short-Time Objective Intelligibility (STOI) metric
 # available in the `pystoi` package, or the Perceptual Evaluation of Speech Quality (PESQ) metric available in the `pesq` package.

examples/tutorials/squim_tutorial.py

@@ -284,8 +284,8 @@ def plot(waveform, title, sample_rate=16000):
 # ```
 # These values are precomputed and hard-coded below.
 print(f"Reference metrics for distorted speech at {snr_dbs[0]}dB are\n")
-print(f"STOI: 0.9670831113894452")
-print(f"PESQ: 2.7961528301239014")
+print("STOI: 0.9670831113894452")
+print("PESQ: 2.7961528301239014")
 si_sdr_ref = si_snr(WAVEFORM_DISTORTED[0:1], WAVEFORM_SPEECH)
 print(f"SI-SDR: {si_sdr_ref}")
 
@@ -304,8 +304,8 @@ def plot(waveform, title, sample_rate=16000):
 si_sdr_ref = si_snr(WAVEFORM_DISTORTED[1:2], WAVEFORM_SPEECH)
 # STOI and PESQ metrics are precomputed and hardcoded below.
 print(f"Reference metrics for distorted speech at {snr_dbs[1]}dB are\n")
-print(f"STOI: 0.5743247866630554")
-print(f"PESQ: 1.1112866401672363")
+print("STOI: 0.5743247866630554")
+print("PESQ: 1.1112866401672363")
 print(f"SI-SDR: {si_sdr_ref}")
 
 

src/libtorchaudio/forced_align/cpu/compute.cpp

@@ -1,9 +1,9 @@
-#include <torch/script.h>
-#include <torch/torch.h>
+#include <torch/csrc/inductor/aoti_torch/c/shim.h>
 #include <torch/csrc/stable/library.h>
-#include <torch/csrc/stable/tensor.h>
 #include <torch/csrc/stable/ops.h>
-#include <torch/csrc/inductor/aoti_torch/c/shim.h>
+#include <torch/csrc/stable/tensor.h>
+#include <torch/script.h>
+#include <torch/torch.h>
 
 using namespace std;
 
@@ -81,18 +81,21 @@ void forced_align_impl(
     auto curIdxOffset = t % 2;
     auto prevIdxOffset = (t - 1) % 2;
     for (auto j = 0; j < S; ++j) {
-      alphas_a[curIdxOffset * S + j] = -std::numeric_limits<scalar_t>::infinity(); // alphas_a[curIdxOffset][j]
+      alphas_a[curIdxOffset * S + j] = -std::numeric_limits<
+          scalar_t>::infinity(); // alphas_a[curIdxOffset][j]
     }
     if (start == 0) {
-      alphas_a[curIdxOffset * S] =
-          alphas_a[prevIdxOffset * S] + logProbs_a[batchIndex][t][blank]; // alphas_a[curIdxOffset][0]
+      alphas_a[curIdxOffset * S] = alphas_a[prevIdxOffset * S] +
+          logProbs_a[batchIndex][t][blank]; // alphas_a[curIdxOffset][0]
       backPtr_a[S * t] = 0; // backPtr_a[t][0] = 0
       startloop += 1;
     }
 
     for (auto i = startloop; i < end; i++) {
       auto x0 = alphas_a[prevIdxOffset * S + i]; // alphas_a[prevIdxOffset][i];
-      auto x1 = alphas_a[prevIdxOffset * S + i - 1]; // alphas_a[prevIdxOffset][i - 1];
+      auto x1 =
+          alphas_a[prevIdxOffset * S + i - 1]; // alphas_a[prevIdxOffset][i
+                                               // - 1];
       auto x2 = -std::numeric_limits<scalar_t>::infinity();
 
       auto labelIdx = (i % 2 == 0) ? blank : targets_a[batchIndex][i / 2];
@@ -103,7 +106,8 @@ void forced_align_impl(
       // (i != 1) just ensures we don't access targets[i - 2] if its i < 2
       if (i % 2 != 0 && i != 1 &&
           targets_a[batchIndex][i / 2] != targets_a[batchIndex][i / 2 - 1]) {
-        x2 = alphas_a[prevIdxOffset * S + i - 2]; // alphas_a[prevIdxOffset][i - 2];
+        x2 = alphas_a[prevIdxOffset * S + i - 2]; // alphas_a[prevIdxOffset][i -
+                                                  // 2];
       }
       scalar_t result = 0.0;
       if (x2 > x1 && x2 > x0) {
@@ -116,12 +120,14 @@ void forced_align_impl(
         result = x0;
         backPtr_a[t * S + i] = 0; // backPtr_a[t][i] = 0
       }
-      alphas_a[curIdxOffset * S + i] = result + logProbs_a[batchIndex][t][labelIdx]; // alphas_a[curIdxOffset][i]
+      alphas_a[curIdxOffset * S + i] = result +
+          logProbs_a[batchIndex][t][labelIdx]; // alphas_a[curIdxOffset][i]
     }
   }
   auto idx1 = (T - 1) % 2;
-  auto ltrIdx = alphas_a[S * idx1 + S - 1] >
-    alphas_a[S * idx1 + S - 2] ? S - 1 : S - 2; // alphas_a[idx1][S - 1], alphas_a[idx1][S - 2]
+  auto ltrIdx = alphas_a[S * idx1 + S - 1] > alphas_a[S * idx1 + S - 2]
+      ? S - 1
+      : S - 2; // alphas_a[idx1][S - 1], alphas_a[idx1][S - 2]
   delete[] alphas_a;
   // path stores the token index for each time step after force alignment.
   for (auto t = T - 1; t > -1; t--) {
@@ -194,15 +200,9 @@ std::tuple<torch::Tensor, torch::Tensor> compute(
               logProbs, targets, blank, paths);
         }
       });
-  return std::make_tuple(
-      paths,
-      logProbs
-      );
+  return std::make_tuple(paths, logProbs);
 }
 
-
-
-
 TORCH_LIBRARY_IMPL(torchaudio, CPU, m) {
   m.impl("forced_align", &compute);
 }

src/libtorchaudio/lfilter.cpp

@@ -82,9 +82,9 @@ void lfilter_core_generic_loop(
   auto coeff = a_coeff_flipped.unsqueeze(2);
   for (int64_t i_sample = 0; i_sample < n_samples_input; i_sample++) {
     auto windowed_output_signal =
-      torch::narrow(padded_output_waveform, 2, i_sample, i_sample + n_order).transpose(0, 1);
-    auto o0 =
-        torch::select(input_signal_windows, 2, i_sample) -
+        torch::narrow(padded_output_waveform, 2, i_sample, i_sample + n_order)
+            .transpose(0, 1);
+    auto o0 = torch::select(input_signal_windows, 2, i_sample) -
         at::matmul(windowed_output_signal, coeff).squeeze(2).transpose(0, 1);
     padded_output_waveform.index_put_(
         {torch::indexing::Slice(),