Update image_text_model.py

image_text_model.py

@@ -0,0 +1,37 @@
+import gradio as gr
+from transformers import AutoProcessor, BlipForConditionalGeneration, AutoTokenizer
+import librosa
+import numpy as np
+import torch
+import open_clip
+
+# Carga el modelo de clasificación de imagen a texto
+blip_processor_large = AutoProcessor.from_pretrained("Salesforce/blip-image-captioning-large")
+blip_model_large = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-large")
+
+device = "cuda" if torch.cuda.is_available() else "cpu"
+blip_model_large.to(device)
+
+##### IMAGE MODEL TO TEXT, MODEL 1
+def generate_caption(processor, model, image, tokenizer=None, use_float_16=False):
+    inputs = processor(images=image, return_tensors="pt").to(device)
+
+    if use_float_16:
+        inputs = inputs.to(torch.float16)
+    
+    generated_ids = model.generate(pixel_values=inputs.pixel_values, max_length=50)
+
+    if tokenizer is not None:
+        generated_caption = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
+    else:
+        generated_caption = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
+   
+    return generated_caption
+
+
+def generate_caption_coca(model, transform, image):
+    im = transform(image).unsqueeze(0).to(device)
+    with torch.no_grad(), torch.cuda.amp.autocast():
+        generated = model.generate(im, seq_len=20)
+    return open_clip.decode(generated[0].detach()).split("<end_of_text>")[0].replace("<start_of_text>", "")
+#####END IMAGE MODEL TO TEXT