optimize: some training optimizations (#95)

* optimzie(train&uvr5): rm sf & simp. AudioPre * fix(audio): too many mallocs * feat(audio): load_audio support stereo * fix(audio): float32 wav saving * fix(train): missing ckpt var
2026-06-07 19:40:44 +08:00 · 2024-11-28 03:20:14 +09:00
parent f4644ec1ec
commit a8783c6639
19 changed files with 163 additions and 433 deletions
--- a/requirements/amd.txt
+++ b/requirements/amd.txt
@@ -10,7 +10,6 @@ faiss-cpu==1.7.3
 gradio
 Cython
 pydub>=0.25.1
-soundfile>=0.12.1
 tensorboardX
 Jinja2>=3.1.2
 json5
--- a/requirements/dml.txt
+++ b/requirements/dml.txt
@@ -9,7 +9,6 @@ faiss-cpu==1.7.3
 gradio
 Cython
 pydub>=0.25.1
-soundfile>=0.12.1
 tensorboardX
 Jinja2>=3.1.2
 json5
--- a/requirements/gui-dml.txt
+++ b/requirements/gui-dml.txt
@@ -18,7 +18,6 @@ PyYAML
 resampy
 scikit_learn
 scipy
-SoundFile
 tensorboard
 tqdm
 wave
--- a/requirements/gui.txt
+++ b/requirements/gui.txt
@@ -18,7 +18,6 @@ PyYAML
 resampy
 scikit_learn
 scipy
-SoundFile
 tensorboard
 tqdm
 wave
--- a/requirements/ipex.txt
+++ b/requirements/ipex.txt
@@ -14,7 +14,6 @@ faiss-cpu==1.7.3
 gradio
 Cython
 pydub>=0.25.1
-soundfile>=0.12.1
 tensorboardX
 Jinja2>=3.1.2
 json5
--- a/requirements/main.txt
+++ b/requirements/main.txt
@@ -9,7 +9,6 @@ faiss-cpu
 gradio
 Cython
 pydub>=0.25.1
-soundfile>=0.12.1
 tensorboardX
 Jinja2>=3.1.2
 json5
--- a/requirements/py311.txt
+++ b/requirements/py311.txt
@@ -9,7 +9,6 @@ faiss-cpu
 gradio
 Cython
 pydub>=0.25.1
-soundfile>=0.12.1
 tensorboardX
 Jinja2>=3.1.2
 json5