initial commit
Browse files
app.py
CHANGED
|
@@ -1,12 +1,12 @@
|
|
| 1 |
import gradio as gr
|
| 2 |
from transformers import ViltProcessor, ViltForQuestionAnswering
|
| 3 |
-
from transformers import AutoProcessor, AutoModelForVisualQuestionAnswering
|
| 4 |
from PIL import Image
|
| 5 |
import torch
|
| 6 |
|
|
|
|
|
|
|
| 7 |
dataset_name = "Multimodal-Fatima/OK-VQA_train"
|
| 8 |
-
model_name = "microsoft/git-base-vqav2"
|
| 9 |
-
model_path = "git-base-vqav2"
|
| 10 |
|
| 11 |
questions = ["What can happen the objects shown are thrown on the ground?",
|
| 12 |
"What was the machine beside the bowl used for?",
|
|
|
|
| 1 |
import gradio as gr
|
| 2 |
from transformers import ViltProcessor, ViltForQuestionAnswering
|
| 3 |
+
from transformers import AutoProcessor, AutoModelForVisualQuestionAnswering, AutoModelForCausalLM
|
| 4 |
from PIL import Image
|
| 5 |
import torch
|
| 6 |
|
| 7 |
+
processor = AutoProcessor.from_pretrained("microsoft/git-base-vqav2")
|
| 8 |
+
model = AutoModelForCausalLM.from_pretrained("microsoft/git-base-vqav2")
|
| 9 |
dataset_name = "Multimodal-Fatima/OK-VQA_train"
|
|
|
|
|
|
|
| 10 |
|
| 11 |
questions = ["What can happen the objects shown are thrown on the ground?",
|
| 12 |
"What was the machine beside the bowl used for?",
|