commit 8c4e98688a0f39fbf2a22a019e9ea2bdc97cc852
Author: zhousha <736730048@qq.com>
Date:   Tue Aug 12 15:03:58 2025 +0800

    merge code repo for sensevoice and whisper

diff --git a/.DS_Store b/.DS_Store
new file mode 100644
index 0000000..4dbf4df
Binary files /dev/null and b/.DS_Store differ
diff --git a/Dockerfile_sensevoice b/Dockerfile_sensevoice
new file mode 100644
index 0000000..eb4c345
--- /dev/null
+++ b/Dockerfile_sensevoice
@@ -0,0 +1,7 @@
+FROM corex:3.2.1
+
+WORKDIR /workspace
+COPY requirements_sensevoice.txt constraints_sensevoice.txt server_sensevoice.py launch_sensevoice.sh /workspace/
+RUN pip install -r requirements_sensevoice.txt -c constraints_sensevoice.txt
+
+ENTRYPOINT ["/bin/bash", "launch_sensevoice.sh"]
\ No newline at end of file
diff --git a/Dockerfile_whisper b/Dockerfile_whisper
new file mode 100644
index 0000000..ab82865
--- /dev/null
+++ b/Dockerfile_whisper
@@ -0,0 +1,10 @@
+FROM zibo.harbor.iluvatar.com.cn:30000/saas/bi100-3.2.1-x86-ubuntu20.04-py3.10-poc-llm-infer:v1.2.2
+
+RUN mkdir /workspace
+WORKDIR /workspace/
+
+COPY asr_server_whisper.py /workspace/asr_server_whisper.py
+COPY ./launch_service /workspace/launch_service
+COPY whisper-tiny /model
+
+ENTRYPOINT ["./launch_service"]
diff --git a/README.md b/README.md
new file mode 100644
index 0000000..a3c0803
--- /dev/null
+++ b/README.md
@@ -0,0 +1,6 @@
+# tiangai100-sensevoice-funasr
+# tiangai100-whisper
+
+【语音识别】
+
+funasr可正常照nvidia方式使用。
diff --git a/asr_server_whisper.py b/asr_server_whisper.py
new file mode 100644
index 0000000..73b2531
--- /dev/null
+++ b/asr_server_whisper.py
@@ -0,0 +1,34 @@
+import torch
+from transformers import pipeline
+from flask import Flask, request, jsonify
+from threading import Lock
+
+asr_pipeline = pipeline(
+    task="automatic-speech-recognition",
+    model="/model",
+    torch_dtype=torch.float16,
+    device="cuda:0"
+)
+
+app = Flask(__name__)
+pipeline_lock = Lock()
+
+@app.route('/recognition', methods=['POST'])
+def predict():
+    audio_data = request.data
+    params = request.params
+    params_json = json.loads(params)
+    lang = params_json.get('language', 'en')
+    with pipeline_lock:
+        res = asr_pipeline(inputs=audio_data, generate_kwargs={"language": lang})
+    text = res['text'].strip()
+    return jsonify({'RecognitionStatus': 'Success', "DisplayText": text})
+
+@app.route('/ready', methods=['GET'])
+@app.route('/health', methods=['GET'])
+@app.route('/health_check', methods=['GET'])
+def health():
+    return jsonify({'status': 'ok'})
+
+if __name__ == '__main__':
+    app.run(host='0.0.0.0', port=80)
diff --git a/constraints_sensevoice.txt b/constraints_sensevoice.txt
new file mode 100644
index 0000000..65d66ac
--- /dev/null
+++ b/constraints_sensevoice.txt
@@ -0,0 +1 @@
+torch==2.1.0+corex.3.2.1
\ No newline at end of file
diff --git a/launch_sensevoice.sh b/launch_sensevoice.sh
new file mode 100755
index 0000000..e75b371
--- /dev/null
+++ b/launch_sensevoice.sh
@@ -0,0 +1,4 @@
+#!/bin/bash
+
+python3 asr_server.py
+
diff --git a/launch_service b/launch_service
new file mode 100755
index 0000000..3ad80ba
--- /dev/null
+++ b/launch_service
@@ -0,0 +1,9 @@
+#!/bin/bash
+
+data
+cat /proc/cpuinfo
+ixsmi
+export
+date
+
+python3 asr.py
diff --git a/requirements_sensevoice.txt b/requirements_sensevoice.txt
new file mode 100644
index 0000000..8c9e4a6
--- /dev/null
+++ b/requirements_sensevoice.txt
@@ -0,0 +1,3 @@
+funasr
+fastapi
+uvicorn[standard]
\ No newline at end of file
diff --git a/server_sensevoice.py b/server_sensevoice.py
new file mode 100644
index 0000000..a87e43c
--- /dev/null
+++ b/server_sensevoice.py
@@ -0,0 +1,41 @@
+from fastapi import FastAPI, Request
+from contextlib import asynccontextmanager
+from funasr import AutoModel
+from funasr.utils.postprocess_utils import rich_transcription_postprocess
+import os
+import torch
+
+model_dir = os.getenv("MODEL_DIR", "/model/iic/SenseVoiceSmall")
+model = None
+
+@asynccontextmanager
+async def lifespan(app: FastAPI):
+    global model
+    if model is None:
+        device = "cuda" if torch.cuda.is_available() else "cpu"
+        model = AutoModel(model=model_dir, disable_update=True, device=device)
+    yield
+    pass
+
+app = FastAPI(lifespan=lifespan)
+
+@app.post("/recognition")
+async def asr(request: Request, language: str = "auto"):
+    audio_data = await request.body()
+    res = model.generate(input=audio_data, use_itn=True, ban_emo_unk=True)
+    text = rich_transcription_postprocess(res[0]["text"])
+    return {
+        "RecognitionStatus": "Success",
+        "DisplayText": text
+    }
+
+@app.get("/health")
+@app.get("/ready")
+def ready():
+    return {"status": "ok"}
+
+
+if __name__ == "__main__":
+    import uvicorn
+    uvicorn.run(app, host="0.0.0.0", port=80)
+
diff --git a/whisper-tiny/.msc b/whisper-tiny/.msc
new file mode 100644
index 0000000..850289a
Binary files /dev/null and b/whisper-tiny/.msc differ
diff --git a/whisper-tiny/.mv b/whisper-tiny/.mv
new file mode 100644
index 0000000..2b25642
--- /dev/null
+++ b/whisper-tiny/.mv
@@ -0,0 +1 @@
+Revision:master,CreatedAt:1736300537
\ No newline at end of file