Spaces:
Sleeping
Sleeping
howard-hou
commited on
Commit
•
c25fbe0
1
Parent(s):
a9b31ad
Update app.py
Browse files
app.py
CHANGED
@@ -1,6 +1,7 @@
|
|
1 |
import gradio as gr
|
2 |
import os, gc
|
3 |
import torch
|
|
|
4 |
from transformers import CLIPImageProcessor
|
5 |
from huggingface_hub import hf_hub_download
|
6 |
|
@@ -33,7 +34,7 @@ image_processor = CLIPImageProcessor.from_pretrained(vision_tower_name)
|
|
33 |
##########################################################################
|
34 |
def generate_prompt(instruction):
|
35 |
instruction = instruction.strip().replace('\r\n','\n').replace('\n\n','\n')
|
36 |
-
return f"{instruction}\n\nAssistant:"
|
37 |
|
38 |
def generate(
|
39 |
ctx,
|
@@ -57,10 +58,8 @@ def generate(
|
|
57 |
for i in range(int(token_count)):
|
58 |
if i == 0:
|
59 |
input_ids = pipeline.encode(ctx)
|
60 |
-
print(input_ids)
|
61 |
text_embs = model.w['emb.weight'][input_ids]
|
62 |
-
input_embs = torch.cat((image_features, text_embs), dim=0)
|
63 |
-
print(input_embs.shape)
|
64 |
out, state = model.forward(embs=input_embs, state=None)
|
65 |
else:
|
66 |
input_ids = [token]
|
@@ -103,12 +102,18 @@ examples = [
|
|
103 |
"What are the things I should be cautious about when I visit here?",
|
104 |
]
|
105 |
]
|
|
|
106 |
def chatbot(image, question):
|
107 |
if image is None:
|
108 |
yield "Please upload an image."
|
109 |
return
|
110 |
image = image_processor(images=image.convert('RGB'), return_tensors='pt')['pixel_values']
|
111 |
image_features = visual_encoder.encode_images(image.unsqueeze(0)).squeeze(0) # [L, D]
|
|
|
|
|
|
|
|
|
|
|
112 |
input_text = generate_prompt(question)
|
113 |
for output in generate(input_text, image_features):
|
114 |
yield output
|
@@ -119,7 +124,7 @@ with gr.Blocks(title=title) as demo:
|
|
119 |
image = gr.Image(type='pil', label="Image")
|
120 |
with gr.Column():
|
121 |
prompt = gr.Textbox(lines=5, label="Prompt",
|
122 |
-
value="
|
123 |
with gr.Row():
|
124 |
submit = gr.Button("Submit", variant="primary")
|
125 |
clear = gr.Button("Clear", variant="secondary")
|
|
|
1 |
import gradio as gr
|
2 |
import os, gc
|
3 |
import torch
|
4 |
+
import torch.nn.functional as F
|
5 |
from transformers import CLIPImageProcessor
|
6 |
from huggingface_hub import hf_hub_download
|
7 |
|
|
|
34 |
##########################################################################
|
35 |
def generate_prompt(instruction):
|
36 |
instruction = instruction.strip().replace('\r\n','\n').replace('\n\n','\n')
|
37 |
+
return f"\n{instruction}\n\nAssistant:"
|
38 |
|
39 |
def generate(
|
40 |
ctx,
|
|
|
58 |
for i in range(int(token_count)):
|
59 |
if i == 0:
|
60 |
input_ids = pipeline.encode(ctx)
|
|
|
61 |
text_embs = model.w['emb.weight'][input_ids]
|
62 |
+
input_embs = torch.cat((image_features, text_embs), dim=0)[-ctx_limit:]
|
|
|
63 |
out, state = model.forward(embs=input_embs, state=None)
|
64 |
else:
|
65 |
input_ids = [token]
|
|
|
102 |
"What are the things I should be cautious about when I visit here?",
|
103 |
]
|
104 |
]
|
105 |
+
|
106 |
def chatbot(image, question):
|
107 |
if image is None:
|
108 |
yield "Please upload an image."
|
109 |
return
|
110 |
image = image_processor(images=image.convert('RGB'), return_tensors='pt')['pixel_values']
|
111 |
image_features = visual_encoder.encode_images(image.unsqueeze(0)).squeeze(0) # [L, D]
|
112 |
+
# apply layer norm to image feature, very important
|
113 |
+
image_features = F.layer_norm(image_features,
|
114 |
+
(image_features.shape[-1],),
|
115 |
+
weight=model.w['blocks.0.ln0.weight'],
|
116 |
+
bias=model.w['blocks.0.ln0.bias'])
|
117 |
input_text = generate_prompt(question)
|
118 |
for output in generate(input_text, image_features):
|
119 |
yield output
|
|
|
124 |
image = gr.Image(type='pil', label="Image")
|
125 |
with gr.Column():
|
126 |
prompt = gr.Textbox(lines=5, label="Prompt",
|
127 |
+
value="Render a clear and concise summary of the photo.")
|
128 |
with gr.Row():
|
129 |
submit = gr.Button("Submit", variant="primary")
|
130 |
clear = gr.Button("Clear", variant="secondary")
|