Fix building (#1343)

Author: csukuangfj

.github/workflows/c-api-test-loading-tokens-hotwords-from-memory.yaml .github/workflows/c-api-from-buffer.yaml

@@ -1,4 +1,4 @@
-name: c-api-test-loading-tokens-hotwords-from-memory
+name: c-api-from-memory
 
 on:
   push:
@@ -7,7 +7,7 @@ on:
     tags:
       - 'v[0-9]+.[0-9]+.[0-9]+*'
     paths:
-      - '.github/workflows/c-api.yaml'
+      - '.github/workflows/c-api-from-buffer.yaml'
       - 'CMakeLists.txt'
       - 'cmake/**'
       - 'sherpa-onnx/csrc/*'
@@ -18,7 +18,7 @@ on:
     branches:
       - master
     paths:
-      - '.github/workflows/c-api.yaml'
+      - '.github/workflows/c-api-from-buffer.yaml'
       - 'CMakeLists.txt'
       - 'cmake/**'
       - 'sherpa-onnx/csrc/*'
@@ -29,11 +29,11 @@ on:
   workflow_dispatch:
 
 concurrency:
-  group: c-api-${{ github.ref }}
+  group: c-api-from-buffer-${{ github.ref }}
   cancel-in-progress: true
 
 jobs:
-  c_api:
+  c_api_from_buffer:
     name: ${{ matrix.os }}
     runs-on: ${{ matrix.os }}
     strategy:
@@ -106,16 +106,17 @@ jobs:
           curl -SL -O https://huggingface.co/desh2608/icefall-asr-librispeech-pruned-transducer-stateless7-streaming-small/blob/main/data/lang_bpe_500/bpe.model
           cp bpe.model sherpa-onnx-streaming-zipformer-en-20M-2023-02-17/
           rm bpe.model
-          
+
           printf "▁A ▁T ▁P :1.5\n▁A ▁B ▁C :3.0" > hotwords.txt
+          mv hotwords.txt ./sherpa-onnx-streaming-zipformer-en-20M-2023-02-17
 
           ls -lh sherpa-onnx-streaming-zipformer-en-20M-2023-02-17
           echo "---"
           ls -lh sherpa-onnx-streaming-zipformer-en-20M-2023-02-17/test_wavs
 
           export LD_LIBRARY_PATH=$PWD/build/install/lib:$LD_LIBRARY_PATH
           export DYLD_LIBRARY_PATH=$PWD/build/install/lib:$DYLD_LIBRARY_PATH
-          
+
           ./streaming-zipformer-buffered-tokens-hotwords-c-api
-          
+
           rm -rf sherpa-onnx-streaming-zipformer-*

c-api-examples/streaming-zipformer-buffered-tokens-hotwords-c-api.c

@@ -5,8 +5,8 @@
 
 //
 // This file demonstrates how to use streaming Zipformer with sherpa-onnx's C
-// and with tokens and hotwords loaded from buffered strings instead of from external
-// files API.
+// and with tokens and hotwords loaded from buffered strings instead of from
+// external files API.
 // clang-format off
 // 
 // wget https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-streaming-zipformer-en-20M-2023-02-17.tar.bz2
@@ -22,7 +22,7 @@
 #include "sherpa-onnx/c-api/c-api.h"
 
 static size_t ReadFile(const char *filename, const char **buffer_out) {
-  FILE *file = fopen(filename, "rb");
+  FILE *file = fopen(filename, "r");
   if (file == NULL) {
     fprintf(stderr, "Failed to open %s\n", filename);
     return -1;
@@ -39,7 +39,7 @@ static size_t ReadFile(const char *filename, const char **buffer_out) {
   size_t read_bytes = fread(*buffer_out, 1, size, file);
   if (read_bytes != size) {
     printf("Errors occured in reading the file %s\n", filename);
-    free(*buffer_out);
+    free((void *)*buffer_out);
     *buffer_out = NULL;
     fclose(file);
     return -1;
@@ -80,14 +80,14 @@ int32_t main() {
   size_t token_buf_size = ReadFile(tokens_filename, &tokens_buf);
   if (token_buf_size < 1) {
     fprintf(stderr, "Please check your tokens.txt!\n");
-    free(tokens_buf);
+    free((void *)tokens_buf);
     return -1;
   }
   const char *hotwords_buf;
   size_t hotwords_buf_size = ReadFile(hotwords_filename, &hotwords_buf);
   if (hotwords_buf_size < 1) {
     fprintf(stderr, "Please check your hotwords.txt!\n");
-    free(hotwords_buf);
+    free((void *)hotwords_buf);
     return -1;
   }
 
@@ -119,9 +119,9 @@ int32_t main() {
   SherpaOnnxOnlineRecognizer *recognizer =
       SherpaOnnxCreateOnlineRecognizer(&recognizer_config);
 
-  free(tokens_buf);
+  free((void *)tokens_buf);
   tokens_buf = NULL;
-  free(hotwords_buf);
+  free((void *)hotwords_buf);
   hotwords_buf = NULL;
 
   if (recognizer == NULL) {
@@ -199,4 +199,4 @@ int32_t main() {
   fprintf(stderr, "\n");
 
   return 0;
-}
+}

flutter/sherpa_onnx/lib/src/sherpa_onnx_bindings.dart

@@ -234,6 +234,11 @@ final class SherpaOnnxOnlineModelConfig extends Struct {
   external Pointer<Utf8> modelingUnit;
 
   external Pointer<Utf8> bpeVocab;
+
+  external Pointer<Utf8> tokensBuf;
+
+  @Int32()
+  external int tokensBufSize;
 }
 
 final class SherpaOnnxOnlineCtcFstDecoderConfig extends Struct {
@@ -275,6 +280,11 @@ final class SherpaOnnxOnlineRecognizerConfig extends Struct {
 
   @Float()
   external double blankPenalty;
+
+  external Pointer<Utf8> hotwordsBuf;
+
+  @Int32()
+  external int hotwordsBufSize;
 }
 
 final class SherpaOnnxSileroVadModelConfig extends Struct {

scripts/dotnet/OnlineModelConfig.cs

@@ -22,6 +22,8 @@ public OnlineModelConfig()
             ModelType = "";
             ModelingUnit = "cjkchar";
             BpeVocab = "";
+            TokensBuf = "";
+            TokensBufSize = 0;
         }
 
         public OnlineTransducerModelConfig Transducer;
@@ -48,6 +50,11 @@ public OnlineModelConfig()
 
         [MarshalAs(UnmanagedType.LPStr)]
         public string BpeVocab;
+
+        [MarshalAs(UnmanagedType.LPStr)]
+        public string TokensBuf;
+
+        public int TokensBufSize;
     }
 
-}
+}

scripts/dotnet/OnlineRecognizerConfig.cs

@@ -26,6 +26,8 @@ public OnlineRecognizerConfig()
             RuleFsts = "";
             RuleFars = "";
             BlankPenalty = 0.0F;
+            HotwordsBuf = "";
+            HotwordsBufSize = 0;
         }
         public FeatureConfig FeatConfig;
         public OnlineModelConfig ModelConfig;
@@ -72,5 +74,10 @@ public OnlineRecognizerConfig()
         public string RuleFars;
 
         public float BlankPenalty;
+
+        [MarshalAs(UnmanagedType.LPStr)]
+        public string HotwordsBuf;
+
+        public int HotwordsBufSize;
     }
 }

scripts/go/sherpa_onnx.go

@@ -89,6 +89,8 @@ type OnlineModelConfig struct {
 	ModelType     string // Optional. You can specify it for faster model initialization
 	ModelingUnit  string // Optional. cjkchar, bpe, cjkchar+bpe
 	BpeVocab      string // Optional.
+	TokensBuf     string // Optional.
+	TokensBufSize int    // Optional.
 }
 
 // Configuration for the feature extractor
@@ -133,6 +135,8 @@ type OnlineRecognizerConfig struct {
 	CtcFstDecoderConfig     OnlineCtcFstDecoderConfig
 	RuleFsts                string
 	RuleFars                string
+	HotwordsBuf             string
+	HotwordsBufSize         int
 }
 
 // It contains the recognition result for a online stream.
@@ -184,6 +188,11 @@ func NewOnlineRecognizer(config *OnlineRecognizerConfig) *OnlineRecognizer {
 	c.model_config.tokens = C.CString(config.ModelConfig.Tokens)
 	defer C.free(unsafe.Pointer(c.model_config.tokens))
 
+	c.model_config.tokens_buf = C.CString(config.ModelConfig.TokensBuf)
+	defer C.free(unsafe.Pointer(c.model_config.tokens_buf))
+
+	c.model_config.tokens_buf_size = C.int(config.ModelConfig.TokensBufSize)
+
 	c.model_config.num_threads = C.int(config.ModelConfig.NumThreads)
 
 	c.model_config.provider = C.CString(config.ModelConfig.Provider)
@@ -212,6 +221,11 @@ func NewOnlineRecognizer(config *OnlineRecognizerConfig) *OnlineRecognizer {
 	c.hotwords_file = C.CString(config.HotwordsFile)
 	defer C.free(unsafe.Pointer(c.hotwords_file))
 
+	c.hotwords_buf = C.CString(config.HotwordsBuf)
+	defer C.free(unsafe.Pointer(c.hotwords_buf))
+
+	c.hotwords_buf_size = C.int(config.HotwordsBufSize)
+
 	c.hotwords_score = C.float(config.HotwordsScore)
 	c.blank_penalty = C.float(config.BlankPenalty)
 

scripts/node-addon-api/src/streaming-asr.cc

@@ -120,6 +120,8 @@ SherpaOnnxOnlineModelConfig GetOnlineModelConfig(Napi::Object obj) {
   SHERPA_ONNX_ASSIGN_ATTR_STR(model_type, modelType);
   SHERPA_ONNX_ASSIGN_ATTR_STR(modeling_unit, modelingUnit);
   SHERPA_ONNX_ASSIGN_ATTR_STR(bpe_vocab, bpeVocab);
+  SHERPA_ONNX_ASSIGN_ATTR_STR(tokens_buf, tokensBuf);
+  SHERPA_ONNX_ASSIGN_ATTR_INT32(tokens_buf_size, tokensBufSize);
 
   return c;
 }
@@ -192,6 +194,8 @@ static Napi::External<SherpaOnnxOnlineRecognizer> CreateOnlineRecognizerWrapper(
   SHERPA_ONNX_ASSIGN_ATTR_STR(rule_fsts, ruleFsts);
   SHERPA_ONNX_ASSIGN_ATTR_STR(rule_fars, ruleFars);
   SHERPA_ONNX_ASSIGN_ATTR_FLOAT(blank_penalty, blankPenalty);
+  SHERPA_ONNX_ASSIGN_ATTR_STR(hotwords_buf, hotwordsBuf);
+  SHERPA_ONNX_ASSIGN_ATTR_INT32(hotwords_buf_size, hotwordsBufSize);
 
   c.ctc_fst_decoder_config = GetCtcFstDecoderConfig(o);
 
@@ -241,6 +245,10 @@ static Napi::External<SherpaOnnxOnlineRecognizer> CreateOnlineRecognizerWrapper(
     delete[] c.model_config.bpe_vocab;
   }
 
+  if (c.model_config.tokens_buf) {
+    delete[] c.model_config.tokens_buf;
+  }
+
   if (c.decoding_method) {
     delete[] c.decoding_method;
   }
@@ -257,6 +265,10 @@ static Napi::External<SherpaOnnxOnlineRecognizer> CreateOnlineRecognizerWrapper(
     delete[] c.rule_fars;
   }
 
+  if (c.hotwords_buf) {
+    delete[] c.hotwords_buf;
+  }
+
   if (c.ctc_fst_decoder_config.graph) {
     delete[] c.ctc_fst_decoder_config.graph;
   }

sherpa-onnx/c-api/c-api.h

@@ -91,7 +91,7 @@ SHERPA_ONNX_API typedef struct SherpaOnnxOnlineModelConfig {
   /// if non-null, loading the tokens from the buffered string directly in
   /// prioriy
   const char *tokens_buf;
-  /// byte size excluding the tailing '\0'
+  /// byte size excluding the trailing '\0'
   int32_t tokens_buf_size;
 } SherpaOnnxOnlineModelConfig;
 

sherpa-onnx/csrc/offline-stream.cc

@@ -4,6 +4,8 @@
 
 #include "sherpa-onnx/csrc/offline-stream.h"
 
+#include <math.h>
+
 #include <algorithm>
 #include <cassert>
 #include <cmath>
@@ -245,7 +247,7 @@ class OfflineStream::Impl {
     for (int32_t i = 0; i != n; ++i) {
       float x = p[i];
       x = (x > amin) ? x : amin;
-      x = std::log10f(x) * multiplier;
+      x = log10f(x) * multiplier;
 
       max_x = (x > max_x) ? x : max_x;
       p[i] = x;

sherpa-onnx/csrc/online-recognizer-transducer-impl.h

@@ -372,7 +372,8 @@ class OnlineRecognizerTransducerImpl : public OnlineRecognizerImpl {
       // segment is incremented only when the last
       // result is not empty, contains non-blanks and longer than context_size)
       const auto &r = s->GetResult();
-      if (!r.tokens.empty() && r.tokens.back() != 0 && r.tokens.size() > context_size) {
+      if (!r.tokens.empty() && r.tokens.back() != 0 &&
+          r.tokens.size() > context_size) {
         s->GetCurrentSegment() += 1;
       }
     }
@@ -392,7 +393,8 @@ class OnlineRecognizerTransducerImpl : public OnlineRecognizerImpl {
     // if last result is not empty, then
     // preserve last tokens as the context for next result
     if (static_cast<int32_t>(last_result.tokens.size()) > context_size) {
-      std::vector<int64_t> context(last_result.tokens.end() - context_size, last_result.tokens.end());
+      std::vector<int64_t> context(last_result.tokens.end() - context_size,
+                                   last_result.tokens.end());
 
       Hypotheses context_hyp({{context, 0}});
       r.hyps = std::move(context_hyp);

sherpa-onnx/pascal-api/sherpa_onnx.pas

@@ -145,6 +145,8 @@   TSherpaOnnxOnlineModelConfig = record
     ModelType: AnsiString;
     ModelingUnit: AnsiString;
     BpeVocab: AnsiString;
+    TokensBuf: AnsiString;
+    TokensBufSize: Integer;
     function ToString: AnsiString;
     class operator Initialize({$IFDEF FPC}var{$ELSE}out{$ENDIF} Dest: TSherpaOnnxOnlineModelConfig);
   end;
@@ -178,6 +180,8 @@   TSherpaOnnxOnlineRecognizerConfig = record
     RuleFsts: AnsiString;
     RuleFars: AnsiString;
     BlankPenalty: Single;
+    HotwordsBuf: AnsiString;
+    HotwordsBufSize: Integer;
     function ToString: AnsiString;
     class operator Initialize({$IFDEF FPC}var{$ELSE}out{$ENDIF} Dest: TSherpaOnnxOnlineRecognizerConfig);
   end;
@@ -490,6 +494,8 @@   SherpaOnnxOnlineModelConfig= record
     ModelType: PAnsiChar;
     ModelingUnit: PAnsiChar;
     BpeVocab: PAnsiChar;
+    TokensBuf: PAnsiChar;
+    TokensBufSize: cint32;
   end;
   SherpaOnnxFeatureConfig = record
     SampleRate: cint32;
@@ -514,6 +520,8 @@   SherpaOnnxOnlineRecognizerConfig = record
     RuleFsts: PAnsiChar;
     RuleFars: PAnsiChar;
     BlankPenalty: cfloat;
+    HotwordsBuf: PAnsiChar;
+    HotwordsBufSize: cint32;
   end;
 
   PSherpaOnnxOnlineRecognizerConfig = ^SherpaOnnxOnlineRecognizerConfig;