bugfix

app/f5tts.py

@@ -1,5 +1,4 @@
 import sys
-sys.path.append('./F5TTS')
 import argparse
 import codecs
 import re
@@ -7,11 +6,11 @@ from pathlib import Path
 
 import numpy as np
 import soundfile as sf
-import tomli
+# import tomli
 from cached_path import cached_path
 
-from model import DiT, UNetT
+from f5_tts.model import DiT, UNetT
-from model.utils_infer import (
+from f5_tts.model.utils_infer import (
 	load_vocoder,
 	load_model,
 	preprocess_ref_audio_text,
@@ -23,9 +22,11 @@ import os
 import json
 from time import time
 from appPublic.dictObject import DictObject
-from appPublic.zmq_reqrep import ZmqReplier
 from appPublic.folderUtils import temp_file
 from appPublic.jsonConfig import getConfig
+from appPublic.worker import awaitify
+from ahserver.webapp import webapp
+from ahserver.serverEnv import ServerEnv
 
 n_mel_channels = 100
 hop_length = 256
@@ -39,17 +40,10 @@ speed = 1.0
 class F5TTS:
 	def __init__(self):
 		self.config = getConfig()
-		self.zmq_url = self.config.zmq_url
-		self.replier = ZmqReplier(self.config.zmq_url, self.generate)
 		# self.vocos = load_vocoder(is_local=True, local_path="../checkpoints/charactr/vocos-mel-24khz")
 		self.load_model()
 		self.setup_voice()
 
-	def run(self):
-		print(f'running {self.zmq_url}')
-		self.replier._run()
-		print('ended ...')
-
 	def load_model(self):
 		# load models
 		ckpt_file = ''
@@ -75,37 +69,6 @@ class F5TTS:
 						self.config.vocab_file)
 		self.model = self.model.to(self.config.device)
 
-	def generate(self, d):
-		msg= d.decode('utf-8')
-		data = DictObject(**json.loads(msg))
-		print(data)
-		t1 = time()
-		if data.stream:
-			for wav in self.inference_stream(data.prompt, stream=data.stream):
-				d = {
-					"reqid":data.reqid,
-					"b64wave":b64str(wav),
-					"finish":False
-				}
-				self.replier.send(json.dumps(d))
-			t2 = time()
-			d = {
-				"reqid":data.reqid,
-				"time_cost":t2 - t1,
-				"finish":True
-			}
-			return json.dumps(d)
-		else:
-			audio_fn = self.inference(data.prompt)
-			t2 = time()
-			d = {
-				"reqid":data.reqid,
-				"audio_file":audio_fn,
-				"time_cost":t2 - t1
-			}
-			print(f'{d}')
-			return json.dumps(d)
-
 	def setup_voice(self):
 		main_voice = {"ref_audio": self.config.ref_audio_fn, 
 					"ref_text": self.config.ref_text}
@@ -123,7 +86,7 @@ class F5TTS:
 			print("Ref_text:", voices[voice]["ref_text"])
 		self.voices = voices
 
-	def inference_stream(self, prompt):
+	def _inference_stream(self, prompt):
 		text_gen = prompt
 		remove_silence = False
 		generated_audio_segments = []
@@ -157,11 +120,11 @@ class F5TTS:
 			'finish':True
 		}
 
-	def inference(self, prompt):
+	def _inference(self, prompt):
 		generated_audio_segments = []
 		remove_silence = self.config.remove_silence or False
 		final_sample_rate = 24000
-		for d in self.inference_stream(prompt):
+		for d in self._inference_stream(prompt):
 			if not d['finish']:
 				audio = d['audio'] 
 				final_sample_rate = d['sample_rate']
@@ -177,20 +140,11 @@ class F5TTS:
 					remove_silence_for_generated_wav(f.name)
 			return fn
 
+def init():
+	g = ServerEnv()
+	f5 = F5TTS()
+	g.infer_stream = awaitify(f5._inference_stream)
+	g.infer = awaitify(f5._inference)
+
 if __name__ == '__main__':
-	workdir = os.getcwd()
+	webapp(init)
-	config = getConfig(workdir, {'workdir':workdir})
-	print(config.ref_audio_fn)
-	tts = F5TTS()
-	print('here')
-	tts.run()
-	"""
-	while True:
-		print('prompt:')
-		p = input()
-		if p != '':
-			t1 = time()
-			f = tts.inference(p)
-			t2 = time()
-			print(f'{f}, cost {t2-t1} seconds')
-	"""

app/w4a2wav.py

@@ -0,0 +1,14 @@
+import os
+import sys
+from pydub import AudioSegment
+
+if len(sys.argv) < 2:
+	exit(1)
+
+m4afn = sys.argv[1]
+wavfn = m4afn[:-3] + 'wav'
+# Load the m4a file
+audio = AudioSegment.from_file(m4afn, format="m4a")
+
+# Export the audio as a wav file
+audio.export(wavfn, format="wav")

conf/config.json

@@ -1,5 +1,4 @@
 {
-	"zmq_url" : "tcp://127.0.0.1:10003",
 	"sample_rate":16000,
 	"vocab_file":"",
 	"remove_silence":false,
@@ -7,6 +6,61 @@
 	"device":"cuda:0",
 	"ref_audio_fn":"$[workdir]$/samples/ttt.wav",
 	"ref_text":"快点吃饭，上课要迟到了。",
-	"cross_fade_duration":0
+	"cross_fade_duration":0,
+	"website":{
+		"paths":[
+			["$[workdir]$/wwwroot",""]
+		],
+		"client_max_size":10000,
+		"host":"0.0.0.0",
+		"port":10099,
+		"coding":"utf-8",
+		"ssl_gg":{
+			"crtfile":"$[workdir]$/conf/www.bsppo.com.pem",
+			"keyfile":"$[workdir]$/conf/www.bsppo.com.key"
+		},
+		"indexes":[
+			"index.html",
+			"index.tmpl",
+			"index.ui",
+			"index.dspy",
+			"index.md"
+		],
+		"startswiths":[
+			{
+				"leading":"/idfile",
+				"registerfunction":"idFileDownload"
+			}
+		],
+		"processors":[
+			[".ws","ws"],
+			[".xterm","xterm"],
+			[".proxy","proxy"],
+			[".llm", "llm"],
+			[".llms", "llms"],
+			[".llma", "llma"],
+			[".xlsxds","xlsxds"],
+			[".sqlds","sqlds"],
+			[".tmpl.js","tmpl"],
+			[".tmpl.css","tmpl"],
+			[".html.tmpl","tmpl"],
+			[".bcrud", "bricks_crud"],
+			[".tmpl","tmpl"],
+			[".app","app"],
+			[".bui","bui"],
+			[".ui","bui"],
+			[".dspy","dspy"],
+			[".md","md"]
+		],
+		"rsakey":{
+			"privatekey":"$[workdir]$/conf/rsa_private_key.pem",
+			"publickey":"$[workdir]$/conf/rsa_public_key.pem"
+		},
+		"session_max_time":3000,
+		"session_issue_time":2500,
+		"session_redis_notuse":{
+			"url":"redis://127.0.0.1:6379"
+		}
+	}
 }
 

install

@@ -0,0 +1,79 @@
+#!/usr/bin/python3
+
+import os
+import sys
+import codecs
+
+if len(sys.argv) < 3:
+	print(f'Usage:\n{sys.argv[0]} venvname')
+	sys.exit(1)
+
+user = os.getlogin()
+home = os.environ.get('HOME')
+try:
+	os.mkdir(f'{home}/ve')
+except:
+	pass
+
+venv = sys.argv[1]
+port = int(sys.argv[2])
+if not os.path.exists(f'{home}/{venv}'):
+	os.system(f'python3 -m venv ~/{venv}')
+pwd = os.getcwd()
+name = os.path.basename(pwd)
+if os.path.exists(f'./app/{name}.py'):
+	print('env exists')
+	sys.exit(1)
+
+dirs = f'./app ./conf ./files ./wwwroot ./script ./logs'.split(' ')
+for d in dirs:
+	try:
+		os.mkdir(d)
+	except:
+		pass
+files=f'app/{name}.py conf/config.json files/README.md wwwroot/index.dspy'.split(' ')
+for f in files:
+	os.system(f'touch {f}')
+
+service = f"""[Unit]
+Description={name} service
+Wants=systemd-networkd.service
+Requires=nginx.service
+
+[Service]
+Type=forking
+ExecStart=su - {user} -c "{pwd}/script/{name}.sh"
+ExecStop=su - ymq "{home}/bin/killname {name}.py"
+[Install]
+WantedBy=multi-user.target
+"""
+
+with codecs.open(f'./script/{name}.service', 'w', 'utf-8') as f:
+	f.write(service)
+
+with codecs.open(f'./script/{name}.sh', 'w', 'utf-8') as f:
+	f.write(f"""#!/usr/bin/bash
+
+killname {pwd}/app/{name}.py
+{home}/{venv}/bin/python {pwd}/app/{name}.py -w {pwd} > {pwd}/logs/stderr.log 2>&1 &
+exit 0
+""")
+
+with codecs.open(f'./script/install.sh', 'w', 'utf-8') as f:
+	f.write(f"""#!/usr/bin/bash
+sudo cp {name}.service /etc/systemd/system
+sudo systemctl enable {name}.service
+sudo systemctl start {name}
+""")
+
+if not os.path.exists(f'{home}/bin'):
+	os.mkdir(f'{home}/bin')
+if not os.path.exists(f'{home}/bin/killname'):
+	with codecs.open(f'{home}/bin/killname', 'w', 'utf-8') as f:
+		f.write("""#!/usr/bin/bash
+
+ps -ef|grep "$1"|grep -v grep|awk '{print("kill -9", $2)}'|sh
+""")
+os.system(f'chmod +x {pwd}/bin/*')
+os.system(f'{pwd}/script/install.sh')
+

run.sh

@@ -1,4 +0,0 @@
-#!/bin/sh
-
-r=$HOME/ve/f5tts/bin/python
-$r $*

script/f5tts.service

@@ -0,0 +1,13 @@
+[Unit]
+Description=f5tts service
+Documention=f5tts service to control f5tts service start or stoop
+Wants=systemd-networkd.service
+Requires=nginx.service
+
+[Service]
+Type=forking
+ExecStart=su - ymq -c "/d/ymq/py/f5tts/run.sh f5tts.py"
+ExecStop=su - ymq "/d/ymq/bin/killname f5tts.py"
+[Install]
+WantedBy=multi-user.target
+

wwwroot/api/inference/index.dspy

@@ -0,0 +1,15 @@
+async def g():
+	async for d in infer_stream(params_kw.prompt):
+		nd = d.get('audio')
+		if nd:
+			yield nd
+try:
+	resp = await stream_response(request, g, content_type='audio/mpeg')
+	return resp
+except Exception as e:
+	exc = format_exc()
+	exception(f'{e=}\n{exc}')
+	return {
+	"status":"error",
+	"message":str(e)
+}

zmq_client.py

@@ -1,44 +0,0 @@
-import sys
-import json
-import os
-
-from appPublic.dictObject import DictObject
-from appPublic.zmq_reqrep import ZmqRequester
-from appPublic.jsonConfig import getConfig
-from appPublic.uniqueID import getID
-
-zmq_url = "tcp://127.0.0.1:9999"
-from time import time
-
-class F5TTSClient:
-    def __init__(self, zmq_url):
-        self.zmq_url = zmq_url
-        self.requester = ZmqRequester(self.zmq_url)
-
-    def generate(self, prompt):
-        d = {
-            "prompt":prompt,
-            "reqid":getID()
-        }
-        msg = json.dumps(d)
-        resp = self.requester.send(msg)
-        if resp != None:
-            ret = json.loads(resp)
-            print(f'response={ret}')
-        else:
-            print(f'response is None')
-
-    def run(self):
-        print(f'running {self.zmq_url}')
-        while True:
-            print('input audio_file:')
-            af = input()
-            if len(af) > 0:
-                self.generate(af)
-        print('ended ...')
-
-if __name__ == '__main__':
-    workdir = os.getcwd()
-    config = getConfig(workdir)
-    asr = ASRClient(config.zmq_url or zmq_url)
-    asr.run()