From 9ca117be75a78b17440a4ac195f9bd8273b1a8ef Mon Sep 17 00:00:00 2001
From: zhousha <736730048@qq.com>
Date: Thu, 14 Aug 2025 10:02:15 +0800
Subject: [PATCH] Add as subdirectories

---
 .DS_Store                                     | Bin 6148 -> 6148 bytes
 README.md                                     | 119 +++++++++++++++++-
 Dockerfile_f5 => bi_v100-f5-tts/Dockerfile_f5 |   0
 bi_v100-f5-tts/README.md                      |   3 +
 .../constraints_f5.txt                        |   0
 f5_server.py => bi_v100-f5-tts/f5_server.py   |   0
 launch_f5.sh => bi_v100-f5-tts/launch_f5.sh   |   0
 .../requirements_f5.txt                       |   0
 bi_v100-gpt-sovits/.DS_Store                  | Bin 0 -> 6148 bytes
 .../Dockerfile_gsv                            |   0
 bi_v100-gpt-sovits/README.md                  |   2 +
 .../constraints_gsv.txt                       |   0
 .../gsv_server.py                             |   0
 .../launch_gsv.sh                             |   0
 .../Dockerfile_kokoro                         |   0
 bi_v100-kokoro/README.md                      |   3 +
 .../constraints_kokoro.txt                    |   0
 .../kokoro_server.py                          |   0
 .../launch_kokoro.sh                          |   0
 .../requirements_kokoro.txt                   |   0
 20 files changed, 122 insertions(+), 5 deletions(-)
 rename Dockerfile_f5 => bi_v100-f5-tts/Dockerfile_f5 (100%)
 create mode 100644 bi_v100-f5-tts/README.md
 rename constraints_f5.txt => bi_v100-f5-tts/constraints_f5.txt (100%)
 rename f5_server.py => bi_v100-f5-tts/f5_server.py (100%)
 rename launch_f5.sh => bi_v100-f5-tts/launch_f5.sh (100%)
 rename requirements_f5.txt => bi_v100-f5-tts/requirements_f5.txt (100%)
 create mode 100644 bi_v100-gpt-sovits/.DS_Store
 rename Dockerfile_gsv => bi_v100-gpt-sovits/Dockerfile_gsv (100%)
 create mode 100644 bi_v100-gpt-sovits/README.md
 rename constraints_gsv.txt => bi_v100-gpt-sovits/constraints_gsv.txt (100%)
 rename gsv_server.py => bi_v100-gpt-sovits/gsv_server.py (100%)
 rename launch_gsv.sh => bi_v100-gpt-sovits/launch_gsv.sh (100%)
 rename Dockerfile_kokoro => bi_v100-kokoro/Dockerfile_kokoro (100%)
 create mode 100644 bi_v100-kokoro/README.md
 rename constraints_kokoro.txt => bi_v100-kokoro/constraints_kokoro.txt (100%)
 rename kokoro_server.py => bi_v100-kokoro/kokoro_server.py (100%)
 rename launch_kokoro.sh => bi_v100-kokoro/launch_kokoro.sh (100%)
 rename requirements_kokoro.txt => bi_v100-kokoro/requirements_kokoro.txt (100%)

diff --git a/.DS_Store b/.DS_Store
index 39566dad8ade06b196960b793c1d556fbebf11bf..42c25544edda977551969df1fc3e5eaab6ca1f1c 100644
GIT binary patch
delta 214
zcmZoMXfc=|#>B`mu~2NHo}wr-0|Nsi1A_nqgCT<f5b83dF_=!=xV#=D!N-uqkjW6w
zP=+iCR8j(j#VN(fIZ65XIY4y`LKq6u847@kbQy{n@`2hj(KTS2oDEb6!bJ@EAd^`@
rMsM6$$+nrDgP#NF-i;f-Gf(Ch(d7Zz2Xrw*0}yZa5!t~!v4I5uCOa#Q

delta 119
zcmZoMXfc=|#>CJzu~2NHo}wrt0|NsP3otMQGvqUbF?ceBFa%F5R96H^a51<u1ONqe
zktI`#lXH^t^K&*&WGrKwynylPW_AvK4xouZf$z+d`9*X&L3$4WF~ejV9_h_7B1@P7
D%F`Nb

diff --git a/README.md b/README.md
index 46f7f6a..1a5639c 100644
--- a/README.md
+++ b/README.md
@@ -1,11 +1,120 @@
-# tiangai100-f5-tts
-# tiangai100-kokoro-tts
-# tiangai100-gpt-sovits
+# tiangai100-f5-tts & tiangai100-kokoro-tts & tiangai100-gpt-sovits
+
+## 天数智芯 天垓100 语音合成
+该模型测试框架在天垓100加速卡上，适配了 Kokoro, F5-TTS, GPT-SoVITS 这3个模型，将语音信号转换为文本。
+
+GPT-SoVITS 模型是一个集成了语音转换和文本转语音功能的先进 AI 系统，基于 GPT 和 SoVITS 技术构建。
+Kokoro 是由 hexgrad 团队开发并开源的轻量级、高性能文本转语音（TTS）模型。
+F5-TTS 模型由上海交通大学团队发布，是基于扩散 Transformer 和 ConvNeXt V2 的文本转语音（TTS）模型。
+
+## GPT-SoVITS 模型测试服务原理
+
+使用 GPT-SoVITS 框架内置的类 TTS 和 TTS_Config。
+其中 TTS_Config 封装了 GPT-SoVITS 模型运行所需的全部配置参数；TTS 类是 GPT-SoVITS 模型的高层封装，整合了模型加载、文本处理、语音生成等全流程逻辑。
+
+初始化时会加载 GPT-SoVITS 的 模型路径（如 bert_base_path、vits_weights_path）、运行设备（device: "cuda"）、精度（is_half）等参数，传入 TTS_Config 配置实例进行初始化，最终传给 TTS 类初始化管道，通过 TTS 类直接调用模型进行语音合成，将生成的音频片段打包后返回给客户端
+
+此外，服务中重写了 PyTorch 中 Conv1d 和 ConvTranspose1d 的前向传播方法，可能用于优化模型在特定设备 CUDA 上的运行效率
+
+## F5-TTS 模型测试服务原理
+通过 infer_batch_process 函数（from f5_tts.infer.utils_infer）调用 F5-TTS 模型，infer_batch_process封装了模型推理的核心细节。核心逻辑为：
+```python
+for gen_audio, gen_sr in infer_batch_process(
+    (audio, sr),  # 参考音频数据和采样率
+    ref_text,     # 参考文本
+    gen_text_batches,  # 分块后的目标文本
+    ema_model,    # 加载好的F5-TTS模型
+    vocoder,      # 声码器
+    device=device,
+    streaming=True,  # 流式输出（边生成边返回）
+    chunk_size=int(24e6),
+):
+    yield gen_audio.tobytes()  # 以字节流返回生成的音频
+```
+生成的音频数据以 WAV 格式的字节流通过 StreamingResponse 返回给客户端，实现实时语音输出。
+
+## GPT-SoVITS 和 F5-TTS 模型测试服务请求示例
+```python
+import requests
+
+# 服务地址（根据实际部署情况修改IP和端口,端口为80）
+url = f"{sut_url}/generate"
+#url = "http://localhost:80/generate"
+
+# 构造请求数据（multipart/form-data）
+files = {
+    "ref_audio": open("/path/to/reference.wav", "rb")  # 参考音频文件
+}
+data = {
+    "ref_text": "这是参考音频对应的文本",  # 参考文本
+    "text": "这是需要合成语音的目标文本",    # 目标文本
+    "lang": "zh"                          # 语言
+}
+
+# 发送请求并保存结果
+response = requests.post(url, files=files, data=data, stream=True)
+if response.status_code == 200:
+    with open("output.wav", "wb") as f:
+        for chunk in response.iter_content(chunk_size=1024):
+            if chunk:
+                f.write(chunk)
+    print("合成成功，音频已保存为 output.wav")
+else:
+    print(f"请求失败，状态码：{response.status_code}")
 
 
+#服务会根据参考音频（ref_audio）和参考文本（ref_text）提取语音风格，然后将目标文本（text）合成为具有相同风格的语音。
+#响应为流式音频，客户端需要按流的方式接收并保存（如示例中使用 stream=True 和迭代读取 chunk）。
+#若需检查服务是否可用，可发送 GET 请求到 http://localhost:80/health 或 http://localhost:80/ready，正常会返回 {"status": "ok"}。
+```
 
-【天垓100的模型适配情况】
-在天垓100上对部分语音合成模型进行适配，测试方式为在Nvidia A100和智铠100加速卡上对同一段text进行语音合成任务，获取运行时间
+## Kokoro 模型测试服务原理
+使用 kokoro 模块内置的 KModel 和 KPipeline 类。
+
+KModel 是封装 kokoro 模型的类，负责加载模型结构和权重。
+KPipeline 是连接文本输入和模型推理的核心管道类，主要职责包括：处理特定语言的文本（如英文音标转换、中文拼音处理等）和 调用 KModel（kokoro 模型）进行语音生成，关联模型与语言处理规则（如英文音标转换、语速调整等）。
+
+向 KPipeline 传入语言标识（lang_code）、模型实例（model）等参数，用于区分不同语言的处理逻辑；kokoro 模型通过 KModel 类加载，由 KPipeline 管道封装调用逻辑，最终调用 kokoro 模型根据输入文本和语言类型生成对应语言的语音,以流式音频形式返回结果
+
+## Kokoro 模型测试服务请求示例
+```python
+import requests
+
+# 服务器地址（根据实际部署修改）
+url = "http://localhost:80/tts"
+
+# SSML 输入（支持中文/英文，通过 xml:lang 指定语言）
+ssml = """
+<speak>
+  <voice xml:lang="zh">你好，这是 Kokoro 模型生成的中文语音。</voice>
+</speak>
+"""
+# 英文示例：<voice xml:lang="en">Hello, this is English speech generated by Kokoro.</voice>
+
+# 发送 POST 请求
+response = requests.post(
+    url,
+    data=ssml,
+    headers={"Content-Type": "text/plain"},  # 直接发送纯文本 SSML
+    stream=True  # 启用流式响应
+)
+
+# 保存音频到文件（PCM 格式，可通过播放器直接播放或转成 WAV）
+if response.status_code == 200:
+    with open("output.pcm", "wb") as f:
+        for chunk in response.iter_content(chunk_size=1024):
+            if chunk:
+                f.write(chunk)
+    print("音频保存成功：output.pcm")
+else:
+    print(f"请求失败：{response.status_code}，{response.text}")
+```
+
+## 如何使用语音合成模型测试框架
+代码实现了一个接收音频数据并返回识别文本的语音识别 HTTP 服务，并基于 corex:3.2.1 基础镜像，将该 HTTP 服务重新打包成 docker 镜像，通过 k8s 集群sut容器去请求这个 HTTP 服务。
+
+## 天垓100上语音合成模型运行测试结果
+在天垓100上对部分语音合成模型进行适配，测试方式为在 Nvidia A100 和 智铠100 加速卡上对同一段text进行语音合成任务，获取运行时间
 
 | 模型名称   | 模型类型               | 适配状态 | 天垓100运行时间/s | Nvidia A100运行时间/s |
 | ---------- | ---------------------- | -------- | ----------------- | --------------------- |
diff --git a/Dockerfile_f5 b/bi_v100-f5-tts/Dockerfile_f5
similarity index 100%
rename from Dockerfile_f5
rename to bi_v100-f5-tts/Dockerfile_f5
diff --git a/bi_v100-f5-tts/README.md b/bi_v100-f5-tts/README.md
new file mode 100644
index 0000000..4896c69
--- /dev/null
+++ b/bi_v100-f5-tts/README.md
@@ -0,0 +1,3 @@
+# tiangai100-f5-tts
+
+【语音合成】
\ No newline at end of file
diff --git a/constraints_f5.txt b/bi_v100-f5-tts/constraints_f5.txt
similarity index 100%
rename from constraints_f5.txt
rename to bi_v100-f5-tts/constraints_f5.txt
diff --git a/f5_server.py b/bi_v100-f5-tts/f5_server.py
similarity index 100%
rename from f5_server.py
rename to bi_v100-f5-tts/f5_server.py
diff --git a/launch_f5.sh b/bi_v100-f5-tts/launch_f5.sh
similarity index 100%
rename from launch_f5.sh
rename to bi_v100-f5-tts/launch_f5.sh
diff --git a/requirements_f5.txt b/bi_v100-f5-tts/requirements_f5.txt
similarity index 100%
rename from requirements_f5.txt
rename to bi_v100-f5-tts/requirements_f5.txt
diff --git a/bi_v100-gpt-sovits/.DS_Store b/bi_v100-gpt-sovits/.DS_Store
new file mode 100644
index 0000000000000000000000000000000000000000..f2bdc65e7d226dcc365c58a892b5e3d1c980c78c
GIT binary patch
literal 6148
zcmeHKJxfDD5S-N%L7SA8_XjNe2S<pAoe)VKG-cowV*jfAuKcvjeiVj>U=f=z3wv|7
zw{yo$@%AkMTYm4JfCYd#-4O>Lrsn(Z6T7L15$QZ*jR70H;+?a<nRD;(f;Vy=fAV)<
zpZkwt*n5IZ3P=GdAO)m=6u6iIRa)ow7duf0NdYNvEeiPeq0t?C;g}eo4i3=*5N8aB
zaUQ({v3Y>l3&%u8XqHrBQmtAHOFHAN@_ONzm~>b*A67S8bto3M^ZXX+u%4(<3P^!-
z1#WY>@cREu|E2#wCut=Gq`*}vV6)YFwd9kkwoV@BwYJf>bk8}`-8c^lhbYIyD92oQ
fIev(w%xga9elHvogU)!+iTW9EU1U<=zZEzEzoZls

literal 0
HcmV?d00001

diff --git a/Dockerfile_gsv b/bi_v100-gpt-sovits/Dockerfile_gsv
similarity index 100%
rename from Dockerfile_gsv
rename to bi_v100-gpt-sovits/Dockerfile_gsv
diff --git a/bi_v100-gpt-sovits/README.md b/bi_v100-gpt-sovits/README.md
new file mode 100644
index 0000000..7640f33
--- /dev/null
+++ b/bi_v100-gpt-sovits/README.md
@@ -0,0 +1,2 @@
+# tiangai100-gpt-sovits
+
diff --git a/constraints_gsv.txt b/bi_v100-gpt-sovits/constraints_gsv.txt
similarity index 100%
rename from constraints_gsv.txt
rename to bi_v100-gpt-sovits/constraints_gsv.txt
diff --git a/gsv_server.py b/bi_v100-gpt-sovits/gsv_server.py
similarity index 100%
rename from gsv_server.py
rename to bi_v100-gpt-sovits/gsv_server.py
diff --git a/launch_gsv.sh b/bi_v100-gpt-sovits/launch_gsv.sh
similarity index 100%
rename from launch_gsv.sh
rename to bi_v100-gpt-sovits/launch_gsv.sh
diff --git a/Dockerfile_kokoro b/bi_v100-kokoro/Dockerfile_kokoro
similarity index 100%
rename from Dockerfile_kokoro
rename to bi_v100-kokoro/Dockerfile_kokoro
diff --git a/bi_v100-kokoro/README.md b/bi_v100-kokoro/README.md
new file mode 100644
index 0000000..e351bcd
--- /dev/null
+++ b/bi_v100-kokoro/README.md
@@ -0,0 +1,3 @@
+# tiangai100-kokoro-tts
+
+【语音合成】
\ No newline at end of file
diff --git a/constraints_kokoro.txt b/bi_v100-kokoro/constraints_kokoro.txt
similarity index 100%
rename from constraints_kokoro.txt
rename to bi_v100-kokoro/constraints_kokoro.txt
diff --git a/kokoro_server.py b/bi_v100-kokoro/kokoro_server.py
similarity index 100%
rename from kokoro_server.py
rename to bi_v100-kokoro/kokoro_server.py
diff --git a/launch_kokoro.sh b/bi_v100-kokoro/launch_kokoro.sh
similarity index 100%
rename from launch_kokoro.sh
rename to bi_v100-kokoro/launch_kokoro.sh
diff --git a/requirements_kokoro.txt b/bi_v100-kokoro/requirements_kokoro.txt
similarity index 100%
rename from requirements_kokoro.txt
rename to bi_v100-kokoro/requirements_kokoro.txt