Changed blip2 model to 2.7b

app.py

@@ -16,8 +16,8 @@ device_map = {
 }
 
 # Load the Blip2 model
-preprocessor_blip2_8_bit = AutoProcessor.from_pretrained("Salesforce/blip2-opt-6.7b")
-model_blip2_8_bit = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-opt-6.7b", device_map=device_map)
+preprocessor_blip2_8_bit = AutoProcessor.from_pretrained("Salesforce/blip2-opt-2.7b")
+model_blip2_8_bit = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-opt-2.7b", device_map="auto", load_in_8bit=True)
 
 # Load the Blip base model
 # preprocessor_blip_base = AutoProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
@@ -79,19 +79,22 @@ def generate_caption(
 	if use_float_16:
 		inputs = inputs.to(torch.float16)
 
-	generated_ids = model.generate(
-		pixel_values=inputs.pixel_values,
-		# attention_mask=inputs.attention_mask,
-		max_length=32,
-		use_cache=True,
-	)
-
-	if tokenizer is None:
-		generated_caption = preprocessor.batch_decode(generated_ids, skip_special_tokens=True)[0]
-	else:
-		generated_caption = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
+	# generated_ids = model.generate(
+	# 	pixel_values=inputs.pixel_values,
+	# 	# attention_mask=inputs.attention_mask,
+	# 	max_length=32,
+	# 	use_cache=True,
+	# )
+
+	# if tokenizer is None:
+	# 	generated_caption = preprocessor.batch_decode(generated_ids, skip_special_tokens=True)[0]
+	# else:
+	# 	generated_caption = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
+
+	generated_ids = model.generate(**inputs, max_new_tokens=32)
+	generated_text = preprocessor.batch_decode(generated_ids, skip_special_tokens=True)[0].strip()	
 		
-	return generated_caption
+	return generated_text
 
 
 def generate_captions_clip(