Update app.py

app.py

@@ -32,11 +32,25 @@ if missing_vars:
 # Replace with your actual model name
 model_name = "jacksonstrut/tinyllama-1.1B-chat"  # Update this with your model's name
 
-# Load the tokenizer and model with use_fast=False
+from transformers import AutoTokenizer
+
+model_name = "jacksonstrut/tinyllama-1.1B-chat"
+HUGGINGFACE_API_TOKEN = os.getenv('HUGGINGFACE_API_TOKEN')
+
+# Disable tokenizer parallelism
+import os
+os.environ["TOKENIZERS_PARALLELISM"] = "false"
+
+# Load the tokenizer with use_fast=False
 tokenizer = AutoTokenizer.from_pretrained(
     model_name,
     token=HUGGINGFACE_API_TOKEN,
-    use_fast=False  # Use the slow tokenizer to avoid conversion issues
+    use_fast=False
+)
+
+# Ensure pad_token is set
+if tokenizer.pad_token is None:
+    tokenizer.pad_token = tokenizer.eos_token
 )
 config = AutoConfig.from_pretrained(model_name)
 model = AutoModelForCausalLM.from_pretrained(