Update handler.py

handler.py

@@ -1,8 +1,9 @@
 # -*- coding: utf-8 -*-
-# handler.py — Rapid_ECG / PULSE-7B — Stabil ve DEBUG'li sürüm (local/hub + vision tower fix)
-# - HF Endpoint uyumlu (EndpointHandler.load().__call__)
-# - Yerel klasörden (HF_MODEL_DIR) veya hub'dan (HF_MODEL_ID) yükleme
-# - Görsel sadece .preprocess() ile işlenir
+# handler.py — Rapid_ECG / PULSE-7B — Startup-load, Stabil ve DEBUG'li sürüm
+# - Sunucu açılır açılmaz model yüklenir (cold start only once)
+# - HF Endpoint sözleşmesi (EndpointHandler.load().__call__)
+# - Yerel (HF_MODEL_DIR) → Hub (HF_MODEL_ID) yükleme sırası
+# - Görsel sadece .preprocess() ile işlenir (process_images yok)
 # - Vision tower kontrolü: mm_vision_tower veya vision_tower
 # - IMAGE_TOKEN_INDEX kullanımı ve kapsamlı [DEBUG] logları
 
@@ -17,7 +18,8 @@ import torch
 from PIL import Image
 import requests
 
-# ===== LLaVA kütüphanesi (gerekirse kur) =====
+
+# ===== LLaVA kütüphanesini garantiye al =====
 def _ensure_llava(tag: str = "v1.2.0"):
     try:
         import llava  # noqa
@@ -86,16 +88,23 @@ def _load_image_from_any(image_input: Any) -> Image.Image:
     if isinstance(image_input, str):
         s = image_input.strip()
         if s.startswith("data:image"):
-            _, b64 = s.split(",", 1)
-            data = base64.b64decode(b64)
-            return Image.open(io.BytesIO(data)).convert("RGB")
+            try:
+                _, b64 = s.split(",", 1)
+                data = base64.b64decode(b64)
+                return Image.open(io.BytesIO(data)).convert("RGB")
+            except Exception as e:
+                raise ValueError(f"Bad data URL: {e}")
         if _is_probably_base64(s) and not s.startswith(("http://", "https://")):
-            data = base64.b64decode(s)
-            return Image.open(io.BytesIO(data)).convert("RGB")
+            try:
+                data = base64.b64decode(s)
+                return Image.open(io.BytesIO(data)).convert("RGB")
+            except Exception as e:
+                raise ValueError(f"Bad base64 image: {e}")
         if s.startswith(("http://", "https://")):
             resp = requests.get(s, timeout=20)
             resp.raise_for_status()
             return Image.open(io.BytesIO(resp.content)).convert("RGB")
+        # local path
         return Image.open(s).convert("RGB")
     raise ValueError(f"Unsupported image input type: {type(image_input)}")
 
@@ -114,7 +123,7 @@ def _get_conv_mode(model_name: str) -> str:
     return "llava_v0"
 
 def _build_prompt_with_image(prompt: str, model_cfg) -> str:
-    # Kullanıcı prompt'a image token eklediyse yeniden eklemeyelim
+    # Kullanıcı image token eklediyse yeniden eklemeyelim
     if DEFAULT_IMAGE_TOKEN in prompt or DEFAULT_IM_START_TOKEN in prompt:
         return prompt
     if getattr(model_cfg, "mm_use_im_start_end", False):
@@ -123,7 +132,7 @@ def _build_prompt_with_image(prompt: str, model_cfg) -> str:
     return f"{DEFAULT_IMAGE_TOKEN}\n{prompt}"
 
 def _resolve_model_path(model_dir_hint: Optional[str], default_dir: str = "/repository") -> str:
-    # Öncelik sırası: HF_MODEL_DIR (yerel) -> verilen model_dir_hint -> default_dir
+    # Öncelik: HF_MODEL_DIR (yerel) -> ctor'dan gelen model_dir_hint -> default_dir
     p = _get_env("HF_MODEL_DIR") or model_dir_hint or default_dir
     p = os.path.abspath(p)
     print(f"[DEBUG] resolved model path: {p}")
@@ -133,22 +142,44 @@ def _resolve_model_path(model_dir_hint: Optional[str], default_dir: str = "/repo
 # ---------- Endpoint Handler ----------
 class EndpointHandler:
     def __init__(self, model_dir: Optional[str] = None):
-        print(f"[DEBUG] EndpointHandler.__init__ model_dir={model_dir}")
+        # DEBUG banner
+        print("🚀 Starting up PULSE-7B handler (startup load)...")
+        print("📝 Enhanced by Kefstacks")
+        print(f"🔧 Python: {sys.version}")
+        print(f"🔧 PyTorch: {torch.__version__}")
+        try:
+            import transformers
+            print(f"🔧 Transformers: {transformers.__version__}")
+        except Exception as e:
+            print(f"[DEBUG] transformers import failed: {e}")
+
         self.model_dir = model_dir
+        self.device = _pick_device()
+        self.dtype = _pick_dtype(self.device)
+
+        # Ortam ayarları (flash attn ipucu, zarar vermez)
+        os.environ.setdefault("ATTN_IMPLEMENTATION", "flash_attention_2")
+        os.environ.setdefault("FLASH_ATTENTION", "1")
+        print(f"[DEBUG] ATTN_IMPLEMENTATION={os.getenv('ATTN_IMPLEMENTATION')} FLASH_ATTENTION={os.getenv('FLASH_ATTENTION')}")
+
+        # Model/Tokenizer/ImageProcessor konteynerleri
         self.model = None
         self.tokenizer = None
         self.image_processor = None
         self.context_len = None
-        self.device = _pick_device()
-        self.dtype = _pick_dtype(self.device)
         self.model_name = None
 
-    def load(self):
-        """
-        Yükleme stratejisi:
-        - Eğer HF_MODEL_DIR set edilmişse veya repo kökünde ağırlıklar varsa: YERELDEN yükle.
-        - Aksi halde HF_MODEL_ID ile hub'dan yükle.
-        """
+        # ---- Modeli burada (startup’ta) yükle ----
+        try:
+            self._startup_load_model()
+            print("✅ Model loaded & ready in __init__")
+        except Exception as e:
+            # Kritik hata: init'te patladıysa endpoint zaten ayağa kalkamaz
+            print(f"💥 CRITICAL: model startup load failed: {e}")
+            raise
+
+    def _startup_load_model(self):
+        # Yerel dizin varsa onu kullan, yoksa hub
         local_path = _resolve_model_path(self.model_dir)
         use_local = os.path.isdir(local_path) and any(
             os.path.exists(os.path.join(local_path, f))
@@ -156,9 +187,6 @@ class EndpointHandler:
         )
         model_base = _get_env("HF_MODEL_BASE", None)
 
-        os.environ.setdefault("ATTN_IMPLEMENTATION", "flash_attention_2")
-        os.environ.setdefault("FLASH_ATTENTION", "1")
-
         if use_local:
             model_path = local_path
             print(f"[DEBUG] loading model LOCALLY from: {model_path}")
@@ -166,7 +194,6 @@ class EndpointHandler:
             model_path = _get_env("HF_MODEL_ID", "PULSE-ECG/PULSE-7B")
             print(f"[DEBUG] loading model from HUB: {model_path} (HF_MODEL_BASE={model_base})")
 
-        # Modeli yükle
         print("[DEBUG] calling load_pretrained_model ...")
         self.tokenizer, self.model, self.image_processor, self.context_len = load_pretrained_model(
             model_path=model_path,
@@ -179,7 +206,7 @@ class EndpointHandler:
         self.model_name = getattr(self.model.config, "name_or_path", str(model_path))
         print(f"[DEBUG] model loaded: name={self.model_name}")
 
-        # ---- Vision tower kontrolü: mm_vision_tower veya vision_tower
+        # Vision tower kontrolü (yeni/eskı alan adları)
         vt = (
             getattr(self.model.config, "mm_vision_tower", None)
             or getattr(self.model.config, "vision_tower", None)
@@ -188,26 +215,29 @@ class EndpointHandler:
         if self.image_processor is None or vt is None:
             raise RuntimeError(
                 "[ERROR] Vision tower not loaded (mm_vision_tower/vision_tower None). "
-                "Bu model multimodal değil veya yanlış checkpoint yüklendi. "
-                "Yerelden yükleyecekseniz HF_MODEL_DIR doğru klasörü göstermeli; "
-                "hub'dan yükleyecekseniz HF_MODEL_ID olarak PULSE/LLaVA tabanlı bir model verin (örn: 'PULSE-ECG/PULSE-7B')."
+                "Yerel yükleme için HF_MODEL_DIR doğru klasörü göstermeli; "
+                "Hub için HF_MODEL_ID PULSE/LLaVA tabanlı olmalı (örn: 'PULSE-ECG/PULSE-7B')."
             )
 
-        # tokenizer güvenliği
+        # Tokenizer güvenliği
         try:
             self.tokenizer.padding_side = "left"
-            if self.tokenizer.pad_token_id is None:
+            if getattr(self.tokenizer, "pad_token_id", None) is None:
                 self.tokenizer.pad_token_id = self.tokenizer.eos_token_id
-        except Exception:
-            pass
+        except Exception as e:
+            print(f"[DEBUG] tokenizer safety patch failed: {e}")
 
         self.model.eval()
-        print("[DEBUG] model.eval() done")
+
+    # HF inference toolkit load() yine çağıracağı için no-op
+    def load(self):
+        print("[DEBUG] load(): model is already initialized in __init__")
         return True
 
     @torch.inference_mode()
     def __call__(self, inputs: Dict[str, Any]) -> Dict[str, Any]:
         print(f"[DEBUG] __call__ inputs keys={list(inputs.keys()) if hasattr(inputs,'keys') else 'N/A'}")
+        # HF {"inputs": {...}} sarmasını aç
         if "inputs" in inputs and isinstance(inputs["inputs"], dict):
             inputs = inputs["inputs"]
 
@@ -218,13 +248,14 @@ class EndpointHandler:
         if not isinstance(prompt, str) or not prompt.strip():
             return {"error": "Missing 'query'/'prompt' text"}
 
+        # Üretim parametreleri
         temperature = float(inputs.get("temperature", 0.0))
         top_p = float(inputs.get("top_p", 0.9))
         max_new = int(inputs.get("max_new_tokens", inputs.get("max_tokens", 512)))
         repetition_penalty = float(inputs.get("repetition_penalty", 1.0))
         conv_mode_override = inputs.get("conv_mode") or _get_env("CONV_MODE", None)
 
-        # ---- image load + preprocess
+        # ---- Görsel yükle + preprocess
         try:
             image = _load_image_from_any(image_in)
             print(f"[DEBUG] loaded image size={image.size}")
@@ -237,31 +268,39 @@ class EndpointHandler:
         try:
             out = self.image_processor.preprocess(image, return_tensors="pt")
             images_tensor = out["pixel_values"].to(self.device, dtype=self.dtype)
-            image_sizes = [image.size]  # bazı LLaVA sürümleri image_sizes ister
+            image_sizes = [image.size]
             print(f"[DEBUG] preprocess OK; images_tensor.shape={images_tensor.shape}")
         except Exception as e:
             return {"error": f"Image preprocessing failed: {e}"}
 
-        # ---- conversation + prompt
+        # ---- Konuşma + prompt
         mode = conv_mode_override or _get_conv_mode(self.model_name)
         conv = (conv_templates.get(mode) or conv_templates[list(conv_templates.keys())[0]]).copy()
         conv.append_message(conv.roles[0], _build_prompt_with_image(prompt.strip(), self.model.config))
         conv.append_message(conv.roles[1], None)
         full_prompt = conv.get_prompt()
+        print(f"[DEBUG] conv_mode={mode}; full_prompt_len={len(full_prompt)}")
 
-        # ---- tokenization (IMAGE_TOKEN_INDEX ile)
+        # ---- Tokenization (IMAGE_TOKEN_INDEX ile)
         try:
             input_ids = tokenizer_image_token(
                 full_prompt, self.tokenizer, image_token_index=IMAGE_TOKEN_INDEX, return_tensors="pt"
             ).unsqueeze(0).to(self.device)
-        except Exception:
-            toks = self.tokenizer([full_prompt], return_tensors="pt", padding=True, truncation=True)
-            input_ids = toks["input_ids"].to(self.device)
+            print(f"[DEBUG] tokenizer_image_token OK; input_ids.shape={input_ids.shape}")
+        except Exception as e:
+            print(f"[DEBUG] tokenizer_image_token failed: {e}; fallback to plain tokenizer")
+            try:
+                toks = self.tokenizer([full_prompt], return_tensors="pt", padding=True, truncation=True)
+                input_ids = toks["input_ids"].to(self.device)
+                print(f"[DEBUG] plain tokenizer OK; input_ids.shape={input_ids.shape}")
+            except Exception as e2:
+                return {"error": f"Tokenization failed: {e} / {e2}"}
 
         attention_mask = torch.ones_like(input_ids, device=self.device)
 
-        # ---- generate
+        # ---- Generate
         try:
+            print(f"[DEBUG] generate(max_new_tokens={max_new}, temp={temperature}, top_p={top_p}, rep={repetition_penalty})")
             gen_ids = self.model.generate(
                 input_ids=input_ids,
                 attention_mask=attention_mask,
@@ -274,12 +313,17 @@ class EndpointHandler:
                 repetition_penalty=repetition_penalty,
                 use_cache=True,
             )
+            print(f"[DEBUG] generate OK; gen_ids.shape={gen_ids.shape}")
         except Exception as e:
             return {"error": f"Generation failed: {e}"}
 
-        # ---- decode (sadece yeni tokenlar)
-        new_tokens = gen_ids[0, input_ids.shape[1]:]
-        text = self.tokenizer.decode(new_tokens, skip_special_tokens=True).strip()
+        # ---- Decode (sadece yeni tokenlar)
+        try:
+            new_tokens = gen_ids[0, input_ids.shape[1]:]
+            text = self.tokenizer.decode(new_tokens, skip_special_tokens=True).strip()
+            print(f"[DEBUG] decoded_text_len={len(text)}")
+        except Exception as e:
+            return {"error": f"Decode failed: {e}"}
 
         return {
             "generated_text": text,