fix jpn ん && eng ' -

Author: wolfgitpr

plugins/G2P/multig2p/internal/V1/TaskImpl.cpp

@@ -8,11 +8,13 @@
 
 #include <stdcorelib/str.h>
 
+#include <synthrt/Core/Support/Logging.h>
 #include <synthrt/Core/Tensor/Tensor.h>
 #include <synthrt/G2P/Task/Task.h>
 #include <synthrt/G2P/Task/G2pTask.h>
 
 namespace srt::g2p::plugins::Multig2p::Internal::V1 {
+    static srt::LogCategory Log("Multig2p");
     namespace InferenceHelper {
         // ============== 预处理 ==============
 
@@ -110,12 +112,12 @@ namespace srt::g2p::plugins::Multig2p::Internal::V1 {
 
             // 构造 [B, maxLen] int64 张量（pad_idx 填充） + [B, maxLen] bool 张量
             std::vector<int64_t> srcData(B * maxLen, vocab.padIdx);
-            std::vector<uint8_t> maskData(B * maxLen, 1); // true = pad
+            srt::core::Tensor::Container maskData(B * maxLen, std::byte{1}); // true = pad
             for (size_t i = 0; i < B; ++i) {
                 const auto &seq = sequences[i];
                 for (size_t j = 0; j < seq.size(); ++j) {
                     srcData[i * maxLen + j] = seq[j];
-                    maskData[i * maxLen + j] = 0; // false = 非pad
+                    maskData[i * maxLen + j] = std::byte{0}; // false = 非pad
                 }
             }
 
@@ -124,7 +126,8 @@ namespace srt::g2p::plugins::Multig2p::Internal::V1 {
             if (!srcExp) return srcExp.takeError();
 
             const std::vector<int64_t> maskShape{static_cast<int64_t>(B), static_cast<int64_t>(maxLen)};
-            auto maskExp = srt::core::Tensor::createFromView<uint8_t>(maskShape, stdc::array_view<uint8_t>{maskData});
+            auto maskExp = srt::core::Tensor::createFromRawData(
+                srt::core::ITensor::Bool, maskShape, std::move(maskData));
             if (!maskExp) return maskExp.takeError();
 
             std::vector<int64_t> langIdsData(B);
@@ -853,6 +856,8 @@ namespace srt::g2p::plugins::Multig2p::Internal::V1 {
         }
         if (!decodeExp) {
             // 解码失败，按 fallback 处理（保留原 lyric）
+            Log.srtWarning("Multig2p: decode failed for %1 word(s): %2",
+                           g2pInput->g2pInput.size(), decodeExp.error().message());
             auto g2pResult = srt::core::NO<srt::g2p::G2pResultV1>::create();
             g2pResult->g2pResult.reserve(g2pInput->g2pInput.size());
             for (const auto &lyric : g2pInput->g2pInput) {

resources/G2pPackages/Phonetic-Suite-Eng/modules/ChainG2p-Eng/config.json

@@ -1 +1 @@
-{"$version": "1.0", "level": 1, "mode": "standard", "schema": {}, "configuration": {"steps": [{"step": "tagAndValidate", "params": {"tagger": [{"type": "regex", "value": ["([A-Z]+)"], "tag": "uppercase", "action": "copy"}, {"type": "regex", "value": ["(?i)([a-z]+)"], "tag": "mixed", "action": "convert"}]}}, {"step": "dict", "params": {"enabled": true, "file": "ds_cmudict-07b.txt"}}, {"step": "model", "params": {"enabled": true, "id": "g2p-multig2p-multi-official", "langRef": "eng/default", "batchSize": 20}}, {"step": "fallback", "params": {"useOriginal": true, "markFailed": true}}]}}
+{"$version": "1.0", "level": 1, "mode": "standard", "schema": {}, "configuration": {"steps": [{"step": "tagAndValidate", "params": {"tagger": [{"type": "regex", "value": ["([A-Z]+)"], "tag": "uppercase", "action": "copy"}, {"type": "regex", "value": ["(?i)([a-z'\\-]+)"], "tag": "mixed", "action": "convert"}]}}, {"step": "dict", "params": {"enabled": true, "file": "ds_cmudict-07b.txt"}}, {"step": "model", "params": {"enabled": true, "id": "g2p-multig2p-multi-official", "langRef": "eng/default", "batchSize": 20}}, {"step": "fallback", "params": {"useOriginal": true, "markFailed": true}}]}}

resources/G2pPackages/Phonetic-Suite-Jpn/modules/ChainG2p-Jpn/config.json

@@ -1 +1 @@
-{"$version": "1.0", "level": 1, "mode": "standard", "schema": {}, "configuration": {"steps": [{"step": "tagAndValidate", "params": {"tagger": [{"type": "array", "value": ["SP", "AP", "EP"], "tag": "stopword", "action": "copy"}, {"type": "regex", "value": ["([\\x{3040}-\\x{309F}\\x{30A0}-\\x{30FF}[ゃゅょぁぃぅぇぉャュョァィゥェォ]?)"], "tag": "kana", "action": "convert"}, {"type": "regex", "value": ["([A-Za-z]+)"], "tag": "latin", "action": "copy"}]}}, {"step": "dict", "params": {"enabled": true, "file": "kana2romaji.txt"}}, {"step": "fallback", "params": {"useOriginal": true, "markFailed": true}}]}}
+{"$version": "1.0", "level": 1, "mode": "standard", "schema": {}, "configuration": {"steps": [{"step": "tagAndValidate", "params": {"tagger": [{"type": "array", "value": ["SP", "AP", "EP"], "tag": "stopword", "action": "copy"}, {"type": "regex", "value": ["([\\x{3040}-\\x{309F}\\x{30A0}-\\x{30FF}][ゃゅょぁぃぅぇぉャュョァィゥェォ]?)"], "tag": "kana", "action": "convert"}, {"type": "regex", "value": ["([A-Za-z]+)"], "tag": "latin", "action": "copy"}]}}, {"step": "dict", "params": {"enabled": true, "file": "kana2romaji.txt"}}, {"step": "fallback", "params": {"useOriginal": true, "markFailed": true}}]}}

resources/G2pPackages/Phonetic-Suite-Jpn/modules/ChainG2p-Jpn/kana2romaji.txt

@@ -168,7 +168,7 @@
 ろ	ro
 わ	wa
 を	o
-ん	N
+ん	n
 ゔぁ	va
 ゔぃ	vi
 ゔ	vu

tests/g2p/CMakeLists.txt

@@ -16,22 +16,24 @@ target_include_directories(synthrt-test-g2p-langref PRIVATE
     ${SYNTHRT_SOURCE_DIR}/include
 )
 
-# L2 end-to-end test: Multig2p ONNX inference covering all 12 languages
-# supported by Phonetic-Suite-Multi. Gated behind SYNTHRT_ENABLE_L2_TESTS
-# (requires real ONNX Runtime + Multig2p plugin DLL + model files).
-if(SYNTHRT_ENABLE_L2_TESTS)
-    synthrt_add_cross_test(synthrt-test-multig2p-languages tst_multig2p_languages.cpp)
-    target_link_libraries(synthrt-test-multig2p-languages PRIVATE
-        srt::g2p
-        srt::core
-        srt::driver
-    )
-    target_include_directories(synthrt-test-multig2p-languages PRIVATE
-        ${SYNTHRT_SOURCE_DIR}/include
-    )
-    target_compile_definitions(synthrt-test-multig2p-languages PRIVATE
-        SYNTHRT_ENABLE_L2_TESTS
-        SYNTHRT_TEST_SOURCE_DIR="${SYNTHRT_SOURCE_DIR}"
-        SYNTHRT_TEST_BINARY_DIR="${CMAKE_BINARY_DIR}"
-    )
-endif()
+# End-to-end test: Manager + plugin DLLs + real G2P packages (Multig2p ONNX
+# inference over all supported languages, plus the ChainG2p Eng/Jpn package
+# data regressions). Like every test in the tree it is unified under
+# SYNTHRT_BUILD_TESTS (see root CMakeLists note: no separate L2 level).
+# Requires real ONNX Runtime + plugin DLLs + package files at TEST time;
+# when they are missing the fixture reports setupError and all tests SKIP
+# gracefully instead of failing.
+synthrt_add_cross_test(synthrt-test-multig2p-languages tst_multig2p_languages.cpp)
+target_link_libraries(synthrt-test-multig2p-languages PRIVATE
+    srt::g2p
+    srt::core
+    srt::driver
+    srt::svs
+)
+target_include_directories(synthrt-test-multig2p-languages PRIVATE
+    ${SYNTHRT_SOURCE_DIR}/include
+)
+target_compile_definitions(synthrt-test-multig2p-languages PRIVATE
+    SYNTHRT_TEST_SOURCE_DIR="${SYNTHRT_SOURCE_DIR}"
+    SYNTHRT_TEST_BINARY_DIR="${CMAKE_BINARY_DIR}"
+)