Upload app.py with huggingface_hub

app.py

@@ -1,6 +1,7 @@
 from __future__ import annotations
 
 import os
+import threading
 import time
 from typing import Any
 
@@ -37,6 +38,7 @@ app.add_middleware(
 _model: Any | None = None
 _tokenizer: Any | None = None
 _loaded_at: float | None = None
+_load_lock = threading.Lock()
 
 
 def load_runtime() -> tuple[Any, Any]:
@@ -44,26 +46,30 @@ def load_runtime() -> tuple[Any, Any]:
     if _model is not None and _tokenizer is not None:
         return _model, _tokenizer
 
-    from transformers import AutoModelForCausalLM, AutoTokenizer
-
-    started = time.time()
-    tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, token=HF_TOKEN)
-    model = AutoModelForCausalLM.from_pretrained(
-        MODEL_ID,
-        token=HF_TOKEN,
-        trust_remote_code=True,
-    )
-    if DEVICE.type == "cuda":
-        model = model.to(device=DEVICE, dtype=torch.float16)
-    else:
-        model = model.to(DEVICE)
-    model.eval()
-    model.attach_tokenizer(tokenizer)
-
-    _model = model
-    _tokenizer = tokenizer
-    _loaded_at = started
-    return model, tokenizer
+    with _load_lock:
+        if _model is not None and _tokenizer is not None:
+            return _model, _tokenizer
+
+        from transformers import AutoModelForCausalLM, AutoTokenizer
+
+        started = time.time()
+        tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, token=HF_TOKEN)
+        model = AutoModelForCausalLM.from_pretrained(
+            MODEL_ID,
+            token=HF_TOKEN,
+            trust_remote_code=True,
+        )
+        if DEVICE.type == "cuda":
+            model = model.to(device=DEVICE, dtype=torch.float16)
+        else:
+            model = model.to(DEVICE)
+        model.eval()
+        model.attach_tokenizer(tokenizer)
+
+        _model = model
+        _tokenizer = tokenizer
+        _loaded_at = started
+        return model, tokenizer
 
 
 def format_prompt(value: str) -> str: