Attempt to fix `InstructBlipForConditionalGeneration`

app.py

@@ -21,7 +21,7 @@ blip2_processor = AutoProcessor.from_pretrained("Salesforce/blip2-opt-6.7b")
 blip2_model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-opt-6.7b", device_map="auto", torch_dtype=torch.float16)
 
 instructblip_processor = AutoProcessor.from_pretrained("Salesforce/instructblip-vicuna-7b")
-instructblip_model = InstructBlipForConditionalGeneration.from_pretrained("Salesforce/instructblip-vicuna-7b", device_map="auto", torch_dtype=torch.float16)
+instructblip_model = InstructBlipForConditionalGeneration.from_pretrained("Salesforce/instructblip-vicuna-7b").to(device)
 
 def generate_caption(processor, model, image, tokenizer=None, use_float_16=False):
     inputs = processor(images=image, return_tensors="pt").to(device)
@@ -45,7 +45,8 @@ def generate_caption_blip2(processor, model, image, replace_token=False):
     inputs = processor(images=image, text=prompt, return_tensors="pt").to(device=device, dtype=torch.float16)
 
     generated_ids = model.generate(pixel_values=inputs.pixel_values,
-                                   num_beams=5, max_length=50, min_length=1, top_p=0.9, repetition_penalty=1.5, length_penalty=1.0, temperature=1)
+                                   num_beams=5, max_length=50, min_length=1, top_p=0.9,
+                                   repetition_penalty=1.5, length_penalty=1.0, temperature=1)
     if replace_token:
         # TODO remove once https://github.com/huggingface/transformers/pull/24492 is merged
         generated_ids[generated_ids == 0] = 2