Upload folder using huggingface_hub

handler.py

@@ -1,14 +1,24 @@
 from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
 import torch
+import os
 
 class EndpointHandler:
     def __init__(self, path=""):
-        self.tokenizer = AutoTokenizer.from_pretrained(path)
+        # Explicitly prevent sentence-transformers auto-detection
+        os.environ["TRANSFORMERS_OFFLINE"] = "1"
+        
+        print(f"Loading T5Gemma model from: {path}")
+        self.tokenizer = AutoTokenizer.from_pretrained(
+            path,
+            trust_remote_code=True
+        )
         self.model = AutoModelForSeq2SeqLM.from_pretrained(
             path, 
             torch_dtype=torch.bfloat16,
-            trust_remote_code=True  # Allow loading of custom model architectures
+            trust_remote_code=True,
+            device_map="auto"
         )
+        print("T5Gemma model loaded successfully")
         
     def __call__(self, data):
         inputs = data.pop("inputs", data)

inference.py

@@ -0,0 +1,6 @@
+# This file explicitly tells HF this is a custom text generation model
+# and prevents auto-detection of sentence-transformers
+
+TASK = "text-generation"
+FRAMEWORK = "transformers"
+MODEL_TYPE = "t5gemma"

requirements.txt

@@ -1,4 +1,4 @@
-torch>=2.6.0
-transformers==4.54.1
-sentencepiece>=0.1.99
-accelerate>=0.21.0
+torch>=2.6.0
+transformers==4.54.1
+sentencepiece>=0.1.99
+accelerate>=0.21.0

temp_check/requirements.txt

@@ -0,0 +1,4 @@
+torch>=2.4.0
+transformers==4.54.1
+sentencepiece>=0.1.99
+accelerate>=0.21.0