Update app.py

app.py

@@ -7,8 +7,8 @@ import torch
 tokenizer = AutoTokenizer.from_pretrained("unsloth/Qwen3-1.7B")
 base_model = AutoModelForCausalLM.from_pretrained(
     "unsloth/Qwen3-1.7B",
-    torch_dtype=torch.float16,
-    device_map="auto"
+    torch_dtype=torch.float32,
+    device_map="cpu"
 )
 model = PeftModel.from_pretrained(base_model, "khazarai/BioGenesis-ToT")
 
@@ -26,12 +26,12 @@ def generate_response(user_input, chat_history):
     )
 
     # Tokenize and send to GPU
-    inputs = tokenizer(text, return_tensors="pt").to("cuda")
+    inputs = tokenizer(text, return_tensors="pt").to("cpu")
 
     # Generate
     output_tokens = model.generate(
         **inputs,
-        max_new_tokens=1200,
+        max_new_tokens=2200,
         temperature=0.6,
         top_p=0.95,
         top_k=20,