From 976869e5fb35f4ca1d24845708c9001e2b76bf68 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E6=BA=90=E6=96=87=E9=9B=A8?=
 <41315874+fumiama@users.noreply.github.com>
Date: Sun, 2 Jun 2024 01:46:18 +0900
Subject: [PATCH] fix(vc): replace cache_harvest_f0 with md5 key

---
 infer/modules/vc/modules.py  |  1 -
 infer/modules/vc/pipeline.py | 13 ++++++-------
 2 files changed, 6 insertions(+), 8 deletions(-)

diff --git a/infer/modules/vc/modules.py b/infer/modules/vc/modules.py
index e739c9f..b67ae64 100644
--- a/infer/modules/vc/modules.py
+++ b/infer/modules/vc/modules.py
@@ -188,7 +188,6 @@ class VC:
                 self.net_g,
                 sid,
                 audio,
-                input_audio_path,
                 times,
                 f0_up_key,
                 f0_method,
diff --git a/infer/modules/vc/pipeline.py b/infer/modules/vc/pipeline.py
index fdfcc8c..78a542f 100644
--- a/infer/modules/vc/pipeline.py
+++ b/infer/modules/vc/pipeline.py
@@ -27,8 +27,8 @@ input_audio_path2wav = {}
 
 
 @lru_cache
-def cache_harvest_f0(input_audio_path, fs, f0max, f0min, frame_period):
-    audio = input_audio_path2wav[input_audio_path]
+def cache_harvest_f0(f0_cache_key, fs, f0max, f0min, frame_period):
+    audio = input_audio_path2wav[f0_cache_key]
     f0, t = pyworld.harvest(
         audio,
         fs=fs,
@@ -83,7 +83,6 @@ class Pipeline(object):
 
     def get_f0(
         self,
-        input_audio_path,
         x,
         p_len,
         f0_up_key,
@@ -114,8 +113,10 @@ class Pipeline(object):
                     f0, [[pad_size, p_len - len(f0) - pad_size]], mode="constant"
                 )
         elif f0_method == "harvest":
-            input_audio_path2wav[input_audio_path] = x.astype(np.double)
-            f0 = cache_harvest_f0(input_audio_path, self.sr, f0_max, f0_min, 10)
+            from hashlib import md5
+            f0_cache_key = md5(x.tobytes()).digest()
+            input_audio_path2wav[f0_cache_key] = x.astype(np.double)
+            f0 = cache_harvest_f0(f0_cache_key, self.sr, f0_max, f0_min, 10)
             if filter_radius > 2:
                 f0 = signal.medfilt(f0, 3)
         elif f0_method == "crepe":
@@ -304,7 +305,6 @@ class Pipeline(object):
         net_g,
         sid,
         audio,
-        input_audio_path,
         times,
         f0_up_key,
         f0_method,
@@ -372,7 +372,6 @@ class Pipeline(object):
         pitch, pitchf = None, None
         if if_f0 == 1:
             pitch, pitchf = self.get_f0(
-                input_audio_path,
                 audio_pad,
                 p_len,
                 f0_up_key,