From 7cae84872c088166b544d789de29632f2a7911dc Mon Sep 17 00:00:00 2001
From: Mehdi Boutayeb <mehdi.boutayeb@asc-it.fr>
Date: Tue, 15 Sep 2026 09:50:01 +0200
Subject: [PATCH] Fix invalid escape sequences in text normalization and models

Three modules use invalid escapes in non-raw string literals:
- fun_text_processing/text_normalization/normalize.py - \., \w
- funasr/models/fsmn_kws/encoder.py - \[
- funasr/models/fun_asr_nano/tools/format5res.py - \.

On Python 3.12+ each raises SyntaxWarning: invalid escape sequence on
import, and FunASR runs the 3.10-3.12 CI matrix. Escaping the
backslashes leaves every string value identical (ruff W605-style).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014HBQNzAf5C2E3MiJC48HQw
---
 fun_text_processing/text_normalization/normalize.py |  2 +-
 funasr/models/fsmn_kws/encoder.py                   | 12 ++++++------
 funasr/models/fun_asr_nano/tools/format5res.py      |  4 ++--
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/fun_text_processing/text_normalization/normalize.py b/fun_text_processing/text_normalization/normalize.py
index c1cebdd097..7b10b65e6d 100644
--- a/fun_text_processing/text_normalization/normalize.py
+++ b/fun_text_processing/text_normalization/normalize.py
@@ -351,7 +351,7 @@ def split_text_into_sentences(self, text: str) -> List[str]:
             upper_case_unicode = "\u0410-\u042F"
 
         # Read and split transcript by utterance (roughly, sentences)
-        split_pattern = f"(?<!\w\.\w.)(?<![A-Z{upper_case_unicode}][a-z{lower_case_unicode}]+\.)(?<![A-Z{upper_case_unicode}]\.)(?<=\.|\?|\!|\.”|\?”\!”)\s(?![0-9]+[a-z]*\.)"
+        split_pattern = f"(?<!\\w\\.\\w.)(?<![A-Z{upper_case_unicode}][a-z{lower_case_unicode}]+\\.)(?<![A-Z{upper_case_unicode}]\\.)(?<=\\.|\\?|\\!|\\.”|\\?”\\!”)\\s(?![0-9]+[a-z]*\\.)"
 
         sentences = regex.split(split_pattern, text)
         return sentences
diff --git a/funasr/models/fsmn_kws/encoder.py b/funasr/models/fsmn_kws/encoder.py
index ff4e2892dd..407576babe 100755
--- a/funasr/models/fsmn_kws/encoder.py
+++ b/funasr/models/fsmn_kws/encoder.py
@@ -84,7 +84,7 @@ def to_pytorch_net(self, fread):
                                   dtype=torch.float32)
         for i in range(self.output_dim):
             line = fread.readline()
-            splits = line.strip().strip('\[\]').strip().split()
+            splits = line.strip().strip('\\[\\]').strip().split()
             assert len(splits) == self.input_dim
             cols = torch.tensor([float(item) for item in splits],
                                 dtype=torch.float32)
@@ -160,7 +160,7 @@ def to_pytorch_net(self, fread):
                                   dtype=torch.float32)
         for i in range(self.output_dim):
             line = fread.readline()
-            splits = line.strip().strip('\[\]').strip().split()
+            splits = line.strip().strip('\\[\\]').strip().split()
             assert len(splits) == self.input_dim
             cols = torch.tensor([float(item) for item in splits],
                                 dtype=torch.float32)
@@ -171,7 +171,7 @@ def to_pytorch_net(self, fread):
         linear_bias = self.state_dict()['linear.bias']
         #print(linear_bias.shape)
         bias_line = fread.readline()
-        splits = bias_line.strip().strip('\[\]').strip().split()
+        splits = bias_line.strip().strip('\\[\\]').strip().split()
         assert len(splits) == self.output_dim
         new_bias = torch.tensor([float(item) for item in splits],
                                 dtype=torch.float32)
@@ -331,7 +331,7 @@ def to_pytorch_net(self, fread):
         self.dim = int(fsmn_split[1])
 
         params_line = fread.readline()
-        params_split = params_line.strip().strip('\[\]').strip().split()
+        params_split = params_line.strip().strip('\\[\\]').strip().split()
         assert len(params_split) == 12
         assert params_split[0] == '<LearnRateCoef>'
         assert params_split[2] == '<LOrder>'
@@ -352,7 +352,7 @@ def to_pytorch_net(self, fread):
         for i in range(self.lorder):
             print('read conv_left weight -- %d' % i)
             line = fread.readline()
-            splits = line.strip().strip('\[\]').strip().split()
+            splits = line.strip().strip('\\[\\]').strip().split()
             assert len(splits) == self.dim
             cols = torch.tensor([float(item) for item in splits],
                                 dtype=torch.float32)
@@ -375,7 +375,7 @@ def to_pytorch_net(self, fread):
             for i in range(self.rorder):
                 print('read conv_right weight -- %d' % i)
                 line = fread.readline()
-                splits = line.strip().strip('\[\]').strip().split()
+                splits = line.strip().strip('\\[\\]').strip().split()
                 assert len(splits) == self.dim
                 cols = torch.tensor([float(item) for item in splits],
                                     dtype=torch.float32)
diff --git a/funasr/models/fun_asr_nano/tools/format5res.py b/funasr/models/fun_asr_nano/tools/format5res.py
index 3b11bb83a7..619b9f67cb 100644
--- a/funasr/models/fun_asr_nano/tools/format5res.py
+++ b/funasr/models/fun_asr_nano/tools/format5res.py
@@ -320,8 +320,8 @@ def all_convert(content):
                     continue
         name = tmp[0]
         content = tmp[1]
-        name = re.sub("\.pcm", "", name)
-        name = re.sub("\.wav", "", name)
+        name = re.sub("\\.pcm", "", name)
+        name = re.sub("\\.wav", "", name)
         content = recoformat(content)
         content = numbersingle(content)
         content = ch_number2digit(content)