pix2struct

app.py

@@ -145,8 +145,8 @@ def m3(que, image):
     
 
 def m4(que, image):
-    processor3 = Pix2StructProcessor.from_pretrained('google/matcha-plotqa-v2')
-    model3 = Pix2StructForConditionalGeneration.from_pretrained('google/matcha-plotqa-v2')
+    processor3 = Pix2StructProcessor.from_pretrained('google/matcha-plotqa-v1')
+    model3 = Pix2StructForConditionalGeneration.from_pretrained('google/matcha-plotqa-v1')
 
     inputs = processor3(images=image, text=que, return_tensors="pt")
     predictions = model3.generate(**inputs, max_new_tokens=512)
@@ -154,8 +154,8 @@ def m4(que, image):
 
 def m5(que, image):
 
-    processor3 = AutoProcessor.from_pretrained("google/pix2struct-ocrvqa-large")
-    model3 = AutoModelForSeq2SeqLM.from_pretrained("google/pix2struct-ocrvqa-large")
+    processor3 = Pix2StructProcessor.from_pretrained("google/pix2struct-ocrvqa-large")
+    model3 = Pix2StructForConditionalGeneration.from_pretrained("google/pix2struct-ocrvqa-large")
 
     inputs = processor3(images=image, text=que, return_tensors="pt")
 
@@ -170,8 +170,8 @@ def m6(que, image):
 
     # predictions = model3.generate(**inputs)
     # return processor3.decode(predictions[0], skip_special_tokens=True)
-    processor3 = Pix2StructProcessor.from_pretrained('google/matcha-plotqa-v2')
-    model3 = Pix2StructForConditionalGeneration.from_pretrained('google/matcha-plotqa-v2')
+    processor3 = Pix2StructProcessor.from_pretrained('google/matcha-plotqa-v1')
+    model3 = Pix2StructForConditionalGeneration.from_pretrained('google/matcha-plotqa-v1')
 
     inputs = processor3(images=image, text=que, return_tensors="pt")
     predictions = model3.generate(**inputs, max_new_tokens=512)