Spaces:
Paused
Paused
Update app.py
Browse files
app.py
CHANGED
@@ -36,6 +36,7 @@ NEW_BLOCK = 128257
|
|
36 |
EOS_TOKEN = 128258
|
37 |
AUDIO_BASE = 128266
|
38 |
AUDIO_SPAN = 4096 * 7 # 28672 Codes
|
|
|
39 |
# Create AUDIO_IDS on the correct device later in load_models
|
40 |
AUDIO_IDS_CPU = torch.arange(AUDIO_BASE, AUDIO_BASE + AUDIO_SPAN)
|
41 |
|
@@ -45,113 +46,117 @@ class AudioMask(LogitsProcessor):
|
|
45 |
super().__init__()
|
46 |
# Allow NEW_BLOCK and all valid audio tokens initially
|
47 |
self.allow = torch.cat([
|
48 |
-
torch.tensor([new_block_token_id], device=audio_ids.device),
|
49 |
audio_ids
|
50 |
], dim=0)
|
51 |
-
self.eos = torch.tensor([eos_token_id], device=audio_ids.device)
|
52 |
-
self.allow_with_eos = torch.cat([self.allow, self.eos], dim=0)
|
53 |
self.sent_blocks = 0 # State: Number of audio blocks sent
|
54 |
|
55 |
def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor) -> torch.FloatTensor:
|
56 |
-
# Determine which tokens are allowed based on whether blocks have been sent
|
57 |
current_allow = self.allow_with_eos if self.sent_blocks > 0 else self.allow
|
58 |
-
|
59 |
-
# Create a mask initialized to negative infinity
|
60 |
mask = torch.full_like(scores, float("-inf"))
|
61 |
-
# Set allowed token scores to 0 (effectively allowing them)
|
62 |
mask[:, current_allow] = 0
|
63 |
-
# Apply the mask to the scores
|
64 |
return scores + mask
|
65 |
|
66 |
def reset(self):
|
67 |
-
"""Resets the state for a new generation request."""
|
68 |
self.sent_blocks = 0
|
69 |
|
70 |
# 3) StoppingCriteria für EOS ---------------------------------------
|
71 |
-
# generate() needs explicit stopping criteria when using a streamer
|
72 |
class EosStoppingCriteria(StoppingCriteria):
|
73 |
def __init__(self, eos_token_id: int):
|
74 |
self.eos_token_id = eos_token_id
|
75 |
|
76 |
def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor, **kwargs) -> bool:
|
77 |
-
# Check if the *last* generated token is the EOS token
|
78 |
if input_ids.shape[1] > 0 and input_ids[:, -1] == self.eos_token_id:
|
79 |
-
# print("StoppingCriteria: EOS detected.")
|
80 |
return True
|
81 |
return False
|
82 |
|
83 |
# 4) Benutzerdefinierter AudioStreamer -------------------------------
|
84 |
class AudioStreamer(BaseStreamer):
|
85 |
-
# --- Updated __init__ to accept target_device ---
|
86 |
def __init__(self, ws: WebSocket, snac_decoder: SNAC, audio_mask: AudioMask, loop: asyncio.AbstractEventLoop, target_device: str):
|
87 |
self.ws = ws
|
88 |
self.snac = snac_decoder
|
89 |
-
self.masker = audio_mask
|
90 |
-
self.loop = loop
|
91 |
-
# --- Use the passed target_device ---
|
92 |
self.device = target_device
|
93 |
-
self.buf: list[int] = []
|
94 |
-
self.tasks = set()
|
95 |
|
96 |
def _decode_block(self, block7: list[int]) -> bytes:
|
97 |
"""
|
98 |
Decodes a block of 7 audio token values (AUDIO_BASE subtracted) into audio bytes.
|
99 |
-
NOTE:
|
100 |
-
|
101 |
-
|
102 |
-
Ensure this mapping is correct for the specific model!
|
103 |
"""
|
104 |
if len(block7) != 7:
|
105 |
print(f"Streamer Warning: _decode_block received {len(block7)} tokens, expected 7. Skipping.")
|
106 |
-
return b""
|
107 |
|
108 |
-
# --- Mapping derived from previous user version (indices [0], [1,4], [2,3,5,6]) ---
|
109 |
-
# This seems more likely to be correct for Kartoffel_Orpheus if the previous version worked.
|
110 |
try:
|
111 |
-
|
112 |
-
|
113 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
114 |
except IndexError:
|
115 |
print(f"Streamer Error: Index out of bounds during token mapping. Block: {block7}")
|
116 |
return b""
|
|
|
|
|
|
|
117 |
|
118 |
-
# ---
|
119 |
-
|
120 |
-
|
121 |
-
|
122 |
-
|
123 |
-
|
124 |
-
|
125 |
-
|
126 |
-
|
127 |
-
|
128 |
-
|
129 |
-
# Convert lists to tensors on the correct device
|
130 |
-
# Use self.device which was set correctly in __init__
|
131 |
-
codes_l1 = torch.tensor(l1, dtype=torch.long, device=self.device).unsqueeze(0)
|
132 |
-
codes_l2 = torch.tensor(l2, dtype=torch.long, device=self.device).unsqueeze(0)
|
133 |
-
codes_l3 = torch.tensor(l3, dtype=torch.long, device=self.device).unsqueeze(0)
|
134 |
-
codes = [codes_l1, codes_l2, codes_l3] # List of tensors for SNAC
|
135 |
-
|
136 |
-
# Decode using SNAC
|
137 |
-
with torch.no_grad():
|
138 |
-
# self.snac should already be on self.device from load_models_startup
|
139 |
-
audio = self.snac.decode(codes)[0] # Decode expects list of tensors, result might have batch dim
|
140 |
|
141 |
-
#
|
142 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
143 |
|
144 |
-
#
|
145 |
-
|
146 |
-
|
|
|
|
|
|
|
|
|
|
|
147 |
|
148 |
async def _send_audio_bytes(self, data: bytes):
|
149 |
"""Coroutine to send bytes over WebSocket."""
|
150 |
-
if not data:
|
151 |
return
|
152 |
try:
|
153 |
await self.ws.send_bytes(data)
|
154 |
-
# print(f"Streamer: Sent {len(data)} audio bytes.")
|
155 |
except WebSocketDisconnect:
|
156 |
print("Streamer: WebSocket disconnected during send.")
|
157 |
except Exception as e:
|
@@ -159,67 +164,43 @@ class AudioStreamer(BaseStreamer):
|
|
159 |
|
160 |
def put(self, value: torch.LongTensor):
|
161 |
"""
|
162 |
-
Receives new token IDs (Tensor) from generate()
|
163 |
-
Processes tokens, decodes full blocks, and schedules sending
|
164 |
"""
|
165 |
-
# Ensure value is on CPU and flatten to a list of ints
|
166 |
if value.numel() == 0:
|
167 |
return
|
168 |
new_token_ids = value.squeeze().tolist()
|
169 |
-
if isinstance(new_token_ids, int):
|
170 |
new_token_ids = [new_token_ids]
|
171 |
|
172 |
for t in new_token_ids:
|
173 |
if t == EOS_TOKEN:
|
174 |
-
|
175 |
-
# EOS is handled by StoppingCriteria, no action needed here except maybe logging.
|
176 |
-
break # Stop processing this batch if EOS is found
|
177 |
-
|
178 |
if t == NEW_BLOCK:
|
179 |
-
# print("Streamer: NEW_BLOCK token encountered.")
|
180 |
-
# NEW_BLOCK indicates the start of audio, might reset buffer if needed
|
181 |
self.buf.clear()
|
182 |
-
continue
|
183 |
-
|
184 |
-
# Check if token is within the expected audio range
|
185 |
if AUDIO_BASE <= t < AUDIO_BASE + AUDIO_SPAN:
|
186 |
-
# Store value relative to base
|
187 |
-
|
188 |
-
else:
|
189 |
-
# Log unexpected tokens (like START_TOKEN or others if generation goes wrong)
|
190 |
# print(f"Streamer Warning: Ignoring unexpected token {t}")
|
191 |
-
pass # Ignore tokens outside the audio range
|
192 |
|
193 |
-
# If buffer has 7 tokens, decode and send
|
194 |
if len(self.buf) == 7:
|
195 |
audio_bytes = self._decode_block(self.buf)
|
196 |
-
self.buf.clear()
|
197 |
|
198 |
-
if audio_bytes:
|
199 |
-
# Schedule the async send function to run on the main event loop
|
200 |
future = asyncio.run_coroutine_threadsafe(self._send_audio_bytes(audio_bytes), self.loop)
|
201 |
self.tasks.add(future)
|
202 |
-
# Optional: Remove completed tasks to prevent memory leak if generation is very long
|
203 |
future.add_done_callback(self.tasks.discard)
|
204 |
|
205 |
-
# Allow EOS only after the first full block has been processed and scheduled for sending
|
206 |
if self.masker.sent_blocks == 0:
|
207 |
-
|
208 |
-
self.masker.sent_blocks = 1 # Update state in the mask
|
209 |
-
|
210 |
-
# Note: No need to explicitly wait for tasks here. put() should return quickly.
|
211 |
|
212 |
def end(self):
|
213 |
"""Called by generate() when generation finishes."""
|
214 |
-
# Handle any remaining tokens in the buffer (optional, here we discard them)
|
215 |
if len(self.buf) > 0:
|
216 |
print(f"Streamer: End of generation with incomplete block ({len(self.buf)} tokens). Discarding.")
|
217 |
self.buf.clear()
|
218 |
-
|
219 |
-
# Optional: Wait briefly for any outstanding send tasks to complete?
|
220 |
-
# This is tricky because end() is sync. A robust solution might involve
|
221 |
-
# signaling the WebSocket handler to wait before closing.
|
222 |
-
# For simplicity, we rely on FastAPI/Uvicorn's graceful shutdown handling.
|
223 |
# print(f"Streamer: Generation finished. Pending send tasks: {len(self.tasks)}")
|
224 |
pass
|
225 |
|
@@ -227,7 +208,7 @@ class AudioStreamer(BaseStreamer):
|
|
227 |
app = FastAPI()
|
228 |
|
229 |
@app.on_event("startup")
|
230 |
-
async def load_models_startup():
|
231 |
global tok, model, snac, masker, stopping_criteria, device, AUDIO_IDS_CPU
|
232 |
|
233 |
print(f"🚀 Starting up on device: {device}")
|
@@ -236,41 +217,32 @@ async def load_models_startup(): # Make startup async if needed for future async
|
|
236 |
tok = AutoTokenizer.from_pretrained(REPO)
|
237 |
print("Tokenizer loaded.")
|
238 |
|
239 |
-
# Load SNAC first (usually smaller)
|
240 |
snac = SNAC.from_pretrained("hubertsiuzdak/snac_24khz").to(device)
|
241 |
-
# --- FIXED Print statement ---
|
242 |
print(f"SNAC loaded to {device}.") # Use the global device variable
|
243 |
|
244 |
-
|
245 |
-
# Determine appropriate dtype based on device and support
|
246 |
-
model_dtype = torch.float32 # Default to float32 for CPU
|
247 |
if device == "cuda":
|
248 |
if torch.cuda.is_bf16_supported():
|
249 |
model_dtype = torch.bfloat16
|
250 |
print("Using bfloat16 for model.")
|
251 |
else:
|
252 |
-
model_dtype = torch.float16
|
253 |
print("Using float16 for model.")
|
254 |
|
255 |
model = AutoModelForCausalLM.from_pretrained(
|
256 |
REPO,
|
257 |
-
device_map={"": 0} if device == "cuda" else None,
|
258 |
torch_dtype=model_dtype,
|
259 |
-
low_cpu_mem_usage=True,
|
260 |
)
|
261 |
-
model.config.pad_token_id = model.config.eos_token_id
|
262 |
print(f"Model loaded to {model.device} with dtype {model.dtype}.")
|
263 |
-
|
264 |
-
# Ensure model is in evaluation mode
|
265 |
model.eval()
|
266 |
|
267 |
-
# Initialize AudioMask (needs AUDIO_IDS on the correct device)
|
268 |
audio_ids_device = AUDIO_IDS_CPU.to(device)
|
269 |
masker = AudioMask(audio_ids_device, NEW_BLOCK, EOS_TOKEN)
|
270 |
print("AudioMask initialized.")
|
271 |
|
272 |
-
# Initialize StoppingCriteria
|
273 |
-
# IMPORTANT: Create the list and add the criteria instance
|
274 |
stopping_criteria = StoppingCriteriaList([EosStoppingCriteria(EOS_TOKEN)])
|
275 |
print("StoppingCriteria initialized.")
|
276 |
|
@@ -283,18 +255,15 @@ def hello():
|
|
283 |
# 6) Helper zum Prompt Bauen -------------------------------------------
|
284 |
def build_prompt(text: str, voice: str) -> tuple[torch.Tensor, torch.Tensor]:
|
285 |
"""Builds the input_ids and attention_mask for the model."""
|
286 |
-
# Format: <START> <VOICE>: <TEXT> <NEW_BLOCK>
|
287 |
prompt_text = f"{voice}: {text}"
|
288 |
prompt_ids = tok(prompt_text, return_tensors="pt").input_ids.to(device)
|
289 |
|
290 |
-
# Construct input_ids tensor
|
291 |
input_ids = torch.cat([
|
292 |
-
torch.tensor([[START_TOKEN]], device=device, dtype=torch.long),
|
293 |
-
prompt_ids,
|
294 |
-
torch.tensor([[NEW_BLOCK]], device=device, dtype=torch.long)
|
295 |
], dim=1)
|
296 |
|
297 |
-
# Create attention mask (all ones)
|
298 |
attention_mask = torch.ones_like(input_ids)
|
299 |
return input_ids, attention_mask
|
300 |
|
@@ -303,49 +272,37 @@ def build_prompt(text: str, voice: str) -> tuple[torch.Tensor, torch.Tensor]:
|
|
303 |
async def tts(ws: WebSocket):
|
304 |
await ws.accept()
|
305 |
print("🔌 Client connected")
|
306 |
-
streamer = None
|
307 |
-
main_loop = asyncio.get_running_loop()
|
308 |
|
309 |
try:
|
310 |
-
# Receive configuration
|
311 |
req_text = await ws.receive_text()
|
312 |
print(f"Received request: {req_text}")
|
313 |
req = json.loads(req_text)
|
314 |
-
text = req.get("text", "Hallo Welt, wie geht es dir heute?")
|
315 |
-
voice = req.get("voice", "Jakob")
|
316 |
|
317 |
if not text:
|
318 |
print("⚠️ Request text is empty.")
|
319 |
-
await ws.close(code=1003, reason="Text cannot be empty")
|
320 |
return
|
321 |
|
322 |
print(f"Generating audio for: '{text}' with voice '{voice}'")
|
323 |
-
|
324 |
-
# Prepare prompt
|
325 |
ids, attn = build_prompt(text, voice)
|
326 |
-
|
327 |
-
# --- Reset stateful components ---
|
328 |
-
masker.reset() # CRITICAL: Reset the mask state for the new request
|
329 |
-
|
330 |
-
# --- Create Streamer Instance ---
|
331 |
-
# --- Pass the global 'device' variable ---
|
332 |
streamer = AudioStreamer(ws, snac, masker, main_loop, device)
|
333 |
|
334 |
-
# --- Run model.generate in a separate thread ---
|
335 |
-
# This prevents blocking the main FastAPI event loop
|
336 |
print("Starting generation in background thread...")
|
337 |
await asyncio.to_thread(
|
338 |
model.generate,
|
339 |
input_ids=ids,
|
340 |
attention_mask=attn,
|
341 |
-
max_new_tokens=1500,
|
342 |
logits_processor=[masker],
|
343 |
stopping_criteria=stopping_criteria,
|
344 |
-
do_sample=False, #
|
345 |
-
# do_sample=True, temperature=0.7, top_p=0.95, # Or use sampling
|
346 |
use_cache=True,
|
347 |
-
streamer=streamer
|
348 |
-
# No need to manage past_key_values manually
|
349 |
)
|
350 |
print("Generation thread finished.")
|
351 |
|
@@ -358,32 +315,26 @@ async def tts(ws: WebSocket):
|
|
358 |
except Exception as e:
|
359 |
error_details = traceback.format_exc()
|
360 |
print(f"❌ WS‑Error: {e}\n{error_details}", flush=True)
|
361 |
-
# Try to send an error message before closing, if possible
|
362 |
error_payload = json.dumps({"error": str(e)})
|
363 |
try:
|
364 |
if ws.client_state.name == "CONNECTED":
|
365 |
-
await ws.send_text(error_payload)
|
366 |
except Exception:
|
367 |
-
pass
|
368 |
-
# Close with internal server error code
|
369 |
if ws.client_state.name == "CONNECTED":
|
370 |
-
await ws.close(code=1011)
|
371 |
finally:
|
372 |
-
# Ensure streamer's end method is called if it exists
|
373 |
if streamer:
|
374 |
try:
|
375 |
-
# print("Calling streamer.end()")
|
376 |
streamer.end()
|
377 |
except Exception as e_end:
|
378 |
print(f"Error during streamer.end(): {e_end}")
|
379 |
|
380 |
-
# Ensure WebSocket is closed
|
381 |
print("Closing connection.")
|
382 |
if ws.client_state.name == "CONNECTED":
|
383 |
try:
|
384 |
-
await ws.close(code=1000)
|
385 |
except RuntimeError as e_close:
|
386 |
-
# Can happen if connection is already closing/closed
|
387 |
print(f"Runtime error closing websocket: {e_close}")
|
388 |
except Exception as e_close_final:
|
389 |
print(f"Error closing websocket: {e_close_final}")
|
@@ -395,6 +346,4 @@ async def tts(ws: WebSocket):
|
|
395 |
if __name__ == "__main__":
|
396 |
import uvicorn
|
397 |
print("Starting Uvicorn server...")
|
398 |
-
|
399 |
-
# Consider adding --workers 1 if you experience issues with multiple workers and global state/GPU memory
|
400 |
-
uvicorn.run("app:app", host="0.0.0.0", port=7860, log_level="info") #, reload=True)
|
|
|
36 |
EOS_TOKEN = 128258
|
37 |
AUDIO_BASE = 128266
|
38 |
AUDIO_SPAN = 4096 * 7 # 28672 Codes
|
39 |
+
CODEBOOK_SIZE = 4096 # Explicitly define the codebook size
|
40 |
# Create AUDIO_IDS on the correct device later in load_models
|
41 |
AUDIO_IDS_CPU = torch.arange(AUDIO_BASE, AUDIO_BASE + AUDIO_SPAN)
|
42 |
|
|
|
46 |
super().__init__()
|
47 |
# Allow NEW_BLOCK and all valid audio tokens initially
|
48 |
self.allow = torch.cat([
|
49 |
+
torch.tensor([new_block_token_id], device=audio_ids.device, dtype=torch.long),
|
50 |
audio_ids
|
51 |
], dim=0)
|
52 |
+
self.eos = torch.tensor([eos_token_id], device=audio_ids.device, dtype=torch.long)
|
53 |
+
self.allow_with_eos = torch.cat([self.allow, self.eos], dim=0)
|
54 |
self.sent_blocks = 0 # State: Number of audio blocks sent
|
55 |
|
56 |
def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor) -> torch.FloatTensor:
|
|
|
57 |
current_allow = self.allow_with_eos if self.sent_blocks > 0 else self.allow
|
|
|
|
|
58 |
mask = torch.full_like(scores, float("-inf"))
|
|
|
59 |
mask[:, current_allow] = 0
|
|
|
60 |
return scores + mask
|
61 |
|
62 |
def reset(self):
|
|
|
63 |
self.sent_blocks = 0
|
64 |
|
65 |
# 3) StoppingCriteria für EOS ---------------------------------------
|
|
|
66 |
class EosStoppingCriteria(StoppingCriteria):
|
67 |
def __init__(self, eos_token_id: int):
|
68 |
self.eos_token_id = eos_token_id
|
69 |
|
70 |
def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor, **kwargs) -> bool:
|
|
|
71 |
if input_ids.shape[1] > 0 and input_ids[:, -1] == self.eos_token_id:
|
|
|
72 |
return True
|
73 |
return False
|
74 |
|
75 |
# 4) Benutzerdefinierter AudioStreamer -------------------------------
|
76 |
class AudioStreamer(BaseStreamer):
|
|
|
77 |
def __init__(self, ws: WebSocket, snac_decoder: SNAC, audio_mask: AudioMask, loop: asyncio.AbstractEventLoop, target_device: str):
|
78 |
self.ws = ws
|
79 |
self.snac = snac_decoder
|
80 |
+
self.masker = audio_mask
|
81 |
+
self.loop = loop
|
|
|
82 |
self.device = target_device
|
83 |
+
self.buf: list[int] = []
|
84 |
+
self.tasks = set()
|
85 |
|
86 |
def _decode_block(self, block7: list[int]) -> bytes:
|
87 |
"""
|
88 |
Decodes a block of 7 audio token values (AUDIO_BASE subtracted) into audio bytes.
|
89 |
+
NOTE: Extracts base code value (0-4095) using modulo, assuming
|
90 |
+
input values represent (slot_offset + code_value).
|
91 |
+
Maps extracted values using the structure potentially correct for Kartoffel_Orpheus.
|
|
|
92 |
"""
|
93 |
if len(block7) != 7:
|
94 |
print(f"Streamer Warning: _decode_block received {len(block7)} tokens, expected 7. Skipping.")
|
95 |
+
return b""
|
96 |
|
|
|
|
|
97 |
try:
|
98 |
+
# --- Extract base code value (0 to CODEBOOK_SIZE-1) for each slot using modulo ---
|
99 |
+
code_val_0 = block7[0] % CODEBOOK_SIZE
|
100 |
+
code_val_1 = block7[1] % CODEBOOK_SIZE
|
101 |
+
code_val_2 = block7[2] % CODEBOOK_SIZE
|
102 |
+
code_val_3 = block7[3] % CODEBOOK_SIZE
|
103 |
+
code_val_4 = block7[4] % CODEBOOK_SIZE
|
104 |
+
code_val_5 = block7[5] % CODEBOOK_SIZE
|
105 |
+
code_val_6 = block7[6] % CODEBOOK_SIZE
|
106 |
+
|
107 |
+
# --- Map the extracted code values to the SNAC codebooks (l1, l2, l3) ---
|
108 |
+
# Using the structure from the user's previous version, believed to be correct
|
109 |
+
l1 = [code_val_0]
|
110 |
+
l2 = [code_val_1, code_val_4]
|
111 |
+
l3 = [code_val_2, code_val_3, code_val_5, code_val_6]
|
112 |
+
|
113 |
except IndexError:
|
114 |
print(f"Streamer Error: Index out of bounds during token mapping. Block: {block7}")
|
115 |
return b""
|
116 |
+
except Exception as e_map: # Catch potential issues with modulo/mapping
|
117 |
+
print(f"Streamer Error: Exception during code value extraction/mapping: {e_map}. Block: {block7}")
|
118 |
+
return b""
|
119 |
|
120 |
+
# --- Convert lists to tensors on the correct device ---
|
121 |
+
try:
|
122 |
+
codes_l1 = torch.tensor(l1, dtype=torch.long, device=self.device).unsqueeze(0)
|
123 |
+
codes_l2 = torch.tensor(l2, dtype=torch.long, device=self.device).unsqueeze(0)
|
124 |
+
codes_l3 = torch.tensor(l3, dtype=torch.long, device=self.device).unsqueeze(0)
|
125 |
+
codes = [codes_l1, codes_l2, codes_l3]
|
126 |
+
except Exception as e_tensor:
|
127 |
+
print(f"Streamer Error: Exception during tensor conversion: {e_tensor}. l1={l1}, l2={l2}, l3={l3}")
|
128 |
+
return b""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
129 |
|
130 |
+
# --- Decode using SNAC ---
|
131 |
+
try:
|
132 |
+
with torch.no_grad():
|
133 |
+
# self.snac should already be on self.device from load_models_startup
|
134 |
+
audio = self.snac.decode(codes)[0] # Decode expects list of tensors, result might have batch dim
|
135 |
+
except Exception as e_decode:
|
136 |
+
# Add more detailed logging here if it fails again
|
137 |
+
print(f"Streamer Error: Exception during snac.decode: {e_decode}")
|
138 |
+
print(f"Input codes shapes: {[c.shape for c in codes]}")
|
139 |
+
print(f"Input codes dtypes: {[c.dtype for c in codes]}")
|
140 |
+
print(f"Input codes devices: {[c.device for c in codes]}")
|
141 |
+
# Avoid printing potentially huge lists, maybe just check min/max?
|
142 |
+
print(f"Input code values (min/max): L1({min(l1)}/{max(l1)}) L2({min(l2)}/{max(l2)}) L3({min(l3)}/{max(l3)})")
|
143 |
+
return b""
|
144 |
|
145 |
+
# --- Post-processing ---
|
146 |
+
try:
|
147 |
+
audio_np = audio.squeeze().detach().cpu().numpy()
|
148 |
+
audio_bytes = (audio_np * 32767).astype("int16").tobytes()
|
149 |
+
return audio_bytes
|
150 |
+
except Exception as e_post:
|
151 |
+
print(f"Streamer Error: Exception during post-processing: {e_post}. Audio tensor shape: {audio.shape}")
|
152 |
+
return b""
|
153 |
|
154 |
async def _send_audio_bytes(self, data: bytes):
|
155 |
"""Coroutine to send bytes over WebSocket."""
|
156 |
+
if not data:
|
157 |
return
|
158 |
try:
|
159 |
await self.ws.send_bytes(data)
|
|
|
160 |
except WebSocketDisconnect:
|
161 |
print("Streamer: WebSocket disconnected during send.")
|
162 |
except Exception as e:
|
|
|
164 |
|
165 |
def put(self, value: torch.LongTensor):
|
166 |
"""
|
167 |
+
Receives new token IDs (Tensor) from generate().
|
168 |
+
Processes tokens, decodes full blocks, and schedules sending.
|
169 |
"""
|
|
|
170 |
if value.numel() == 0:
|
171 |
return
|
172 |
new_token_ids = value.squeeze().tolist()
|
173 |
+
if isinstance(new_token_ids, int):
|
174 |
new_token_ids = [new_token_ids]
|
175 |
|
176 |
for t in new_token_ids:
|
177 |
if t == EOS_TOKEN:
|
178 |
+
break
|
|
|
|
|
|
|
179 |
if t == NEW_BLOCK:
|
|
|
|
|
180 |
self.buf.clear()
|
181 |
+
continue
|
|
|
|
|
182 |
if AUDIO_BASE <= t < AUDIO_BASE + AUDIO_SPAN:
|
183 |
+
self.buf.append(t - AUDIO_BASE) # Store value relative to base
|
184 |
+
# else: # Optionally log ignored tokens
|
|
|
|
|
185 |
# print(f"Streamer Warning: Ignoring unexpected token {t}")
|
|
|
186 |
|
|
|
187 |
if len(self.buf) == 7:
|
188 |
audio_bytes = self._decode_block(self.buf)
|
189 |
+
self.buf.clear()
|
190 |
|
191 |
+
if audio_bytes:
|
|
|
192 |
future = asyncio.run_coroutine_threadsafe(self._send_audio_bytes(audio_bytes), self.loop)
|
193 |
self.tasks.add(future)
|
|
|
194 |
future.add_done_callback(self.tasks.discard)
|
195 |
|
|
|
196 |
if self.masker.sent_blocks == 0:
|
197 |
+
self.masker.sent_blocks = 1
|
|
|
|
|
|
|
198 |
|
199 |
def end(self):
|
200 |
"""Called by generate() when generation finishes."""
|
|
|
201 |
if len(self.buf) > 0:
|
202 |
print(f"Streamer: End of generation with incomplete block ({len(self.buf)} tokens). Discarding.")
|
203 |
self.buf.clear()
|
|
|
|
|
|
|
|
|
|
|
204 |
# print(f"Streamer: Generation finished. Pending send tasks: {len(self.tasks)}")
|
205 |
pass
|
206 |
|
|
|
208 |
app = FastAPI()
|
209 |
|
210 |
@app.on_event("startup")
|
211 |
+
async def load_models_startup():
|
212 |
global tok, model, snac, masker, stopping_criteria, device, AUDIO_IDS_CPU
|
213 |
|
214 |
print(f"🚀 Starting up on device: {device}")
|
|
|
217 |
tok = AutoTokenizer.from_pretrained(REPO)
|
218 |
print("Tokenizer loaded.")
|
219 |
|
|
|
220 |
snac = SNAC.from_pretrained("hubertsiuzdak/snac_24khz").to(device)
|
|
|
221 |
print(f"SNAC loaded to {device}.") # Use the global device variable
|
222 |
|
223 |
+
model_dtype = torch.float32
|
|
|
|
|
224 |
if device == "cuda":
|
225 |
if torch.cuda.is_bf16_supported():
|
226 |
model_dtype = torch.bfloat16
|
227 |
print("Using bfloat16 for model.")
|
228 |
else:
|
229 |
+
model_dtype = torch.float16
|
230 |
print("Using float16 for model.")
|
231 |
|
232 |
model = AutoModelForCausalLM.from_pretrained(
|
233 |
REPO,
|
234 |
+
device_map={"": 0} if device == "cuda" else None,
|
235 |
torch_dtype=model_dtype,
|
236 |
+
low_cpu_mem_usage=True,
|
237 |
)
|
238 |
+
model.config.pad_token_id = model.config.eos_token_id
|
239 |
print(f"Model loaded to {model.device} with dtype {model.dtype}.")
|
|
|
|
|
240 |
model.eval()
|
241 |
|
|
|
242 |
audio_ids_device = AUDIO_IDS_CPU.to(device)
|
243 |
masker = AudioMask(audio_ids_device, NEW_BLOCK, EOS_TOKEN)
|
244 |
print("AudioMask initialized.")
|
245 |
|
|
|
|
|
246 |
stopping_criteria = StoppingCriteriaList([EosStoppingCriteria(EOS_TOKEN)])
|
247 |
print("StoppingCriteria initialized.")
|
248 |
|
|
|
255 |
# 6) Helper zum Prompt Bauen -------------------------------------------
|
256 |
def build_prompt(text: str, voice: str) -> tuple[torch.Tensor, torch.Tensor]:
|
257 |
"""Builds the input_ids and attention_mask for the model."""
|
|
|
258 |
prompt_text = f"{voice}: {text}"
|
259 |
prompt_ids = tok(prompt_text, return_tensors="pt").input_ids.to(device)
|
260 |
|
|
|
261 |
input_ids = torch.cat([
|
262 |
+
torch.tensor([[START_TOKEN]], device=device, dtype=torch.long),
|
263 |
+
prompt_ids,
|
264 |
+
torch.tensor([[NEW_BLOCK]], device=device, dtype=torch.long)
|
265 |
], dim=1)
|
266 |
|
|
|
267 |
attention_mask = torch.ones_like(input_ids)
|
268 |
return input_ids, attention_mask
|
269 |
|
|
|
272 |
async def tts(ws: WebSocket):
|
273 |
await ws.accept()
|
274 |
print("🔌 Client connected")
|
275 |
+
streamer = None
|
276 |
+
main_loop = asyncio.get_running_loop()
|
277 |
|
278 |
try:
|
|
|
279 |
req_text = await ws.receive_text()
|
280 |
print(f"Received request: {req_text}")
|
281 |
req = json.loads(req_text)
|
282 |
+
text = req.get("text", "Hallo Welt, wie geht es dir heute?")
|
283 |
+
voice = req.get("voice", "Jakob")
|
284 |
|
285 |
if not text:
|
286 |
print("⚠️ Request text is empty.")
|
287 |
+
await ws.close(code=1003, reason="Text cannot be empty")
|
288 |
return
|
289 |
|
290 |
print(f"Generating audio for: '{text}' with voice '{voice}'")
|
|
|
|
|
291 |
ids, attn = build_prompt(text, voice)
|
292 |
+
masker.reset()
|
|
|
|
|
|
|
|
|
|
|
293 |
streamer = AudioStreamer(ws, snac, masker, main_loop, device)
|
294 |
|
|
|
|
|
295 |
print("Starting generation in background thread...")
|
296 |
await asyncio.to_thread(
|
297 |
model.generate,
|
298 |
input_ids=ids,
|
299 |
attention_mask=attn,
|
300 |
+
max_new_tokens=1500,
|
301 |
logits_processor=[masker],
|
302 |
stopping_criteria=stopping_criteria,
|
303 |
+
do_sample=False, # Using greedy decoding
|
|
|
304 |
use_cache=True,
|
305 |
+
streamer=streamer
|
|
|
306 |
)
|
307 |
print("Generation thread finished.")
|
308 |
|
|
|
315 |
except Exception as e:
|
316 |
error_details = traceback.format_exc()
|
317 |
print(f"❌ WS‑Error: {e}\n{error_details}", flush=True)
|
|
|
318 |
error_payload = json.dumps({"error": str(e)})
|
319 |
try:
|
320 |
if ws.client_state.name == "CONNECTED":
|
321 |
+
await ws.send_text(error_payload)
|
322 |
except Exception:
|
323 |
+
pass
|
|
|
324 |
if ws.client_state.name == "CONNECTED":
|
325 |
+
await ws.close(code=1011)
|
326 |
finally:
|
|
|
327 |
if streamer:
|
328 |
try:
|
|
|
329 |
streamer.end()
|
330 |
except Exception as e_end:
|
331 |
print(f"Error during streamer.end(): {e_end}")
|
332 |
|
|
|
333 |
print("Closing connection.")
|
334 |
if ws.client_state.name == "CONNECTED":
|
335 |
try:
|
336 |
+
await ws.close(code=1000)
|
337 |
except RuntimeError as e_close:
|
|
|
338 |
print(f"Runtime error closing websocket: {e_close}")
|
339 |
except Exception as e_close_final:
|
340 |
print(f"Error closing websocket: {e_close_final}")
|
|
|
346 |
if __name__ == "__main__":
|
347 |
import uvicorn
|
348 |
print("Starting Uvicorn server...")
|
349 |
+
uvicorn.run("app:app", host="0.0.0.0", port=7860, log_level="info")
|
|
|
|