app.py

# app.py ──────────────────────────────────────────────────────────────
import os, json, torch, asyncio
from fastapi import FastAPI, WebSocket, WebSocketDisconnect
from huggingface_hub import login
from transformers import AutoTokenizer, AutoModelForCausalLM, LogitsProcessor
from transformers.generation.utils import Cache
from snac import SNAC

# 0 · Auth & Device ---------------------------------------------------
HF_TOKEN = os.getenv("HF_TOKEN")
if HF_TOKEN:
    login(HF_TOKEN)

device = "cuda" if torch.cuda.is_available() else "cpu"
torch.backends.cuda.enable_flash_sdp(False)            # SDP‑Assert fix

# 1 · Konstanten ------------------------------------------------------
REPO         = "SebastianBodza/Kartoffel_Orpheus-3B_german_natural-v0.1"
CHUNK_TOKENS = 50
START_TOKEN  = 128259
NEW_BLOCK    = 128257
EOS_TOKEN    = 128258
AUDIO_BASE   = 128266
AUDIO_SPAN   = 4096 * 7                                # 28 672 Codes
VALID_AUDIO  = torch.arange(AUDIO_BASE, AUDIO_BASE + AUDIO_SPAN)

# 2 · Logit‑Masker ----------------------------------------------------
class DynamicMask(LogitsProcessor):
    def __init__(self, audio_ids: torch.Tensor, min_blocks:int=1):
        super().__init__()
        self.audio_ids = audio_ids
        self.ctrl_ids  = torch.tensor([NEW_BLOCK], device=audio_ids.device)
        self.blocks    = 0
        self.min_blk   = min_blocks
    def __call__(self, inp_ids, scores):
        allow = torch.cat([self.audio_ids, self.ctrl_ids])
        if self.blocks >= self.min_blk:
            allow = torch.cat([allow,
                               torch.tensor([EOS_TOKEN], device=scores.device)])
        mask = torch.full_like(scores, float("-inf"))
        mask[:, allow] = 0
        return scores + mask

# 3 · FastAPI‑App -----------------------------------------------------
app = FastAPI()

@app.get("/")
async def root():
    return {"msg": "Orpheus‑TTS online"}

@app.on_event("startup")
async def load():
    global tok, model, snac, masker
    print("⏳ Lade Modelle …")
    tok   = AutoTokenizer.from_pretrained(REPO)
    snac  = SNAC.from_pretrained("hubertsiuzdak/snac_24khz").to(device)
    model = AutoModelForCausalLM.from_pretrained(
        REPO,
        low_cpu_mem_usage=True,
        device_map={"":0} if device=="cuda" else None,
        torch_dtype=torch.bfloat16 if device=="cuda" else None,
    )
    model.config.pad_token_id = model.config.eos_token_id
    model.config.use_cache    = True
    masker = DynamicMask(VALID_AUDIO.to(device))
    print("✅ Modelle geladen")

# 4 · Hilfsfunktionen -------------------------------------------------
def build_inputs(text:str, voice:str):
    prompt = f"{voice}: {text}"
    ids = tok(prompt, return_tensors="pt").input_ids.to(device)
    ids = torch.cat([torch.tensor([[START_TOKEN]], device=device),
                     ids,
                     torch.tensor([[128009,128260]], device=device)],1)
    return ids, torch.ones_like(ids)

def decode_block(b):
    l1,l2,l3=[],[],[]
    l1.append(b[0])
    l2.append(b[1]-4096)
    l3 += [b[2]-8192,  b[3]-12288]
    l2.append(b[4]-16384)
    l3 += [b[5]-20480, b[6]-24576]
    codes=[torch.tensor(x,device=device).unsqueeze(0) for x in (l1,l2,l3)]
    audio=snac.decode(codes).squeeze().cpu().numpy()
    return (audio*32767).astype("int16").tobytes()

# 5 · WebSocket‑Endpoint ---------------------------------------------
@app.websocket("/ws/tts")
async def tts(ws: WebSocket):
    await ws.accept()
    try:
        req   = json.loads(await ws.receive_text())
        ids, attn = build_inputs(req.get("text",""), req.get("voice","Jakob"))
        past, last_tok, buf = None, None, []
        prompt_len = ids.shape[1]

        while True:
            print(f"DEBUG: Before generate - past is None: {past is None}") # Added logging
            out = model.generate(
                input_ids      = ids if past is None else torch.tensor([[last_tok]], device=device),
                attention_mask = attn if past is None else None,
                past_key_values= past,
                max_new_tokens = CHUNK_TOKENS,
                logits_processor=[masker],
                do_sample=True, temperature=0.7, top_p=0.95,
                use_cache=True, return_dict_in_generate=True,
                return_legacy_cache=True)
            print(f"DEBUG: After generate - type of out.past_key_values: {type(out.past_key_values)}") # Added logging
            pkv = out.past_key_values
            print(f"DEBUG: After getting pkv - type of pkv: {type(pkv)}") # Added logging
            if isinstance(pkv, Cache): pkv = pkv.to_legacy()
            past = pkv
            print(f"DEBUG: After cache handling - past is None: {past is None}") # Added logging

            seq  = out.sequences[0].tolist()
            new  = seq[prompt_len:];  prompt_len = len(seq)
            print("new tokens:", new[:25])

            if not new: raise StopIteration
            for t in new:
                last_tok = t
                if t == EOS_TOKEN: raise StopIteration
                if t == NEW_BLOCK: buf.clear(); continue
                buf.append(t - AUDIO_BASE)
                if len(buf) == 7:
                    await ws.send_bytes(decode_block(buf))
                    buf.clear()
                    masker.blocks += 1

            ids, attn = None, None                  # ab jetzt 1‑Token‑Step

    except (StopIteration, WebSocketDisconnect):
        pass
    except Exception as e:
        print("❌ WS‑Error:", e)
        if ws.client_state.name != "DISCONNECTED":
            await ws.close(code=1011)
    finally:
        if ws.client_state.name != "DISCONNECTED":
            try: await ws.close()
            except RuntimeError: pass

# 6 · Local run -------------------------------------------------------
if __name__ == "__main__":
    import uvicorn
    uvicorn.run("app:app", host="0.0.0.0", port=7860)