Spaces:
Runtime error
Runtime error
added optional image
Browse files
app.py
CHANGED
@@ -238,11 +238,15 @@ def inference(img, text, is_translate):
|
|
238 |
use_beam_search = True
|
239 |
if is_translate:
|
240 |
# encode text
|
|
|
|
|
241 |
text = clip_model.tokenize([text]).to(device)
|
242 |
with torch.no_grad():
|
243 |
prefix = clip_model.encode_text(text).to(device, dtype=torch.float32)
|
244 |
|
245 |
else:
|
|
|
|
|
246 |
image = io.imread(img.name)
|
247 |
pil_image = PIL.Image.fromarray(image)
|
248 |
image = preprocess(pil_image).unsqueeze(0).to(device)
|
@@ -251,6 +255,7 @@ def inference(img, text, is_translate):
|
|
251 |
prefix = clip_model.encode_image(image).to(device, dtype=torch.float32)
|
252 |
|
253 |
prefix_embed = model.clip_project(prefix).reshape(1, prefix_length, -1)
|
|
|
254 |
if use_beam_search:
|
255 |
generated_text_prefix = generate_beam(model, tokenizer, embed=prefix_embed)[0]
|
256 |
else:
|
@@ -262,7 +267,7 @@ description = "You can translate English sentence to Vietnamese sentence or gene
|
|
262 |
examples=[["drug.jpg","", False], ["", "What is your name?", True]]
|
263 |
|
264 |
inputs = [
|
265 |
-
gr.inputs.Image(type="file", label="Image to generate Vietnamese caption"),
|
266 |
gr.inputs.Textbox(lines=2, placeholder="English sentence for translation"),
|
267 |
gr.inputs.Checkbox()
|
268 |
]
|
|
|
238 |
use_beam_search = True
|
239 |
if is_translate:
|
240 |
# encode text
|
241 |
+
if text is None:
|
242 |
+
return "No text provided"
|
243 |
text = clip_model.tokenize([text]).to(device)
|
244 |
with torch.no_grad():
|
245 |
prefix = clip_model.encode_text(text).to(device, dtype=torch.float32)
|
246 |
|
247 |
else:
|
248 |
+
if img is None:
|
249 |
+
return "No image"
|
250 |
image = io.imread(img.name)
|
251 |
pil_image = PIL.Image.fromarray(image)
|
252 |
image = preprocess(pil_image).unsqueeze(0).to(device)
|
|
|
255 |
prefix = clip_model.encode_image(image).to(device, dtype=torch.float32)
|
256 |
|
257 |
prefix_embed = model.clip_project(prefix).reshape(1, prefix_length, -1)
|
258 |
+
|
259 |
if use_beam_search:
|
260 |
generated_text_prefix = generate_beam(model, tokenizer, embed=prefix_embed)[0]
|
261 |
else:
|
|
|
267 |
examples=[["drug.jpg","", False], ["", "What is your name?", True]]
|
268 |
|
269 |
inputs = [
|
270 |
+
gr.inputs.Image(type="file", label="Image to generate Vietnamese caption", optional=True),
|
271 |
gr.inputs.Textbox(lines=2, placeholder="English sentence for translation"),
|
272 |
gr.inputs.Checkbox()
|
273 |
]
|