7

app.py

@@ -10,8 +10,8 @@ from datasets import load_dataset
 
 huggingface_hub.login(os.getenv('HF_TOKEN'))
 #peft_model_id = "debisoft/DeepSeek-R1-Distill-Qwen-7B-thinking-function_calling-quant-V0"
-#peft_model_id = "debisoft/Qwen2.5-VL-7B-Instruct-thinking-function_calling-quant-V0"
-peft_model_id = "debisoft/Qwen2.5-VL-3B-Instruct-thinking-function_calling-V0"
+peft_model_id = "debisoft/Qwen2.5-VL-7B-Instruct-thinking-function_calling-quant-V0"
+#peft_model_id = "debisoft/Qwen2.5-VL-3B-Instruct-thinking-function_calling-V0"
 
 bnb_config = BitsAndBytesConfig(
             load_in_4bit=True,
@@ -52,7 +52,7 @@ def sentience_check():
 
     with torch.no_grad():
         outputs = peft_model.generate(
-            **inputs, max_new_tokens=1024, pad_token_id = tokenizer.eos_token_id
+            **inputs, max_new_tokens=512, pad_token_id = tokenizer.eos_token_id
         )
 
     #peft_model.to(cpu_device)