Spaces:
Build error
Build error
Commit
•
c7d234c
1
Parent(s):
8028287
Update app.py
Browse files
app.py
CHANGED
@@ -41,20 +41,7 @@ def query(payload):
|
|
41 |
#model = AutoModelForCausalLM.from_pretrained("EleutherAI/gpt-j-6B")
|
42 |
|
43 |
|
44 |
-
config = ViTConfig(num_hidden_layers=12, hidden_size=768)
|
45 |
-
model = ViTForImageClassification(config)
|
46 |
|
47 |
-
#print(config)
|
48 |
-
|
49 |
-
feature_extractor = ViTFeatureExtractor()
|
50 |
-
# or, to load one that corresponds to a checkpoint on the hub:
|
51 |
-
#feature_extractor = ViTFeatureExtractor.from_pretrained("google/vit-base-patch16-224")
|
52 |
-
|
53 |
-
#the following gets called by classify_image()
|
54 |
-
feature_extractor = PerceiverFeatureExtractor.from_pretrained("deepmind/vision-perceiver-conv")
|
55 |
-
model = PerceiverForImageClassificationConvProcessing.from_pretrained("deepmind/vision-perceiver-conv")
|
56 |
-
#google/vit-base-patch16-224, deepmind/vision-perceiver-conv
|
57 |
-
image_pipe = ImageClassificationPipeline(model=model, feature_extractor=feature_extractor)
|
58 |
|
59 |
|
60 |
|
@@ -168,12 +155,30 @@ def self_caption(image):
|
|
168 |
print(pred_dictionary)
|
169 |
#return(pred_dictionary)
|
170 |
preds = ' '.join(preds)
|
171 |
-
|
172 |
-
story =
|
|
|
|
|
173 |
return story
|
174 |
|
175 |
|
176 |
def classify_image(image):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
177 |
results = image_pipe(image)
|
178 |
|
179 |
print("RESULTS")
|
|
|
41 |
#model = AutoModelForCausalLM.from_pretrained("EleutherAI/gpt-j-6B")
|
42 |
|
43 |
|
|
|
|
|
44 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
45 |
|
46 |
|
47 |
|
|
|
155 |
print(pred_dictionary)
|
156 |
#return(pred_dictionary)
|
157 |
preds = ' '.join(preds)
|
158 |
+
#inference(input_sentence, max_length, sample_or_greedy, seed=42)
|
159 |
+
story = inference(preds, 32, "Sample", 42)
|
160 |
+
#story = create_story(preds)
|
161 |
+
#story = ' '.join(story)
|
162 |
return story
|
163 |
|
164 |
|
165 |
def classify_image(image):
|
166 |
+
config = ViTConfig(num_hidden_layers=12, hidden_size=768)
|
167 |
+
model = ViTForImageClassification(config)
|
168 |
+
|
169 |
+
#print(config)
|
170 |
+
|
171 |
+
feature_extractor = ViTFeatureExtractor()
|
172 |
+
# or, to load one that corresponds to a checkpoint on the hub:
|
173 |
+
#feature_extractor = ViTFeatureExtractor.from_pretrained("google/vit-base-patch16-224")
|
174 |
+
|
175 |
+
#the following gets called by classify_image()
|
176 |
+
feature_extractor = PerceiverFeatureExtractor.from_pretrained("deepmind/vision-perceiver-conv")
|
177 |
+
model = PerceiverForImageClassificationConvProcessing.from_pretrained("deepmind/vision-perceiver-conv")
|
178 |
+
#google/vit-base-patch16-224, deepmind/vision-perceiver-conv
|
179 |
+
image_pipe = ImageClassificationPipeline(model=model, feature_extractor=feature_extractor)
|
180 |
+
|
181 |
+
|
182 |
results = image_pipe(image)
|
183 |
|
184 |
print("RESULTS")
|