Spaces:
Running
on
Zero
Running
on
Zero
ZeroGPU
Browse files- app.py +5 -4
- requirements.txt +4 -3
app.py
CHANGED
@@ -1,23 +1,24 @@
|
|
1 |
import gradio as gr
|
2 |
import torch
|
|
|
3 |
from PIL import Image
|
4 |
from transformers import AutoModelForCausalLM, AutoTokenizer
|
5 |
|
6 |
model_name = "Lin-Chen/ShareCaptioner"
|
7 |
tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
|
8 |
model = AutoModelForCausalLM.from_pretrained(
|
9 |
-
model_name, device_map="cpu", trust_remote_code=True).eval()
|
10 |
model.tokenizer = tokenizer
|
11 |
|
12 |
model.cuda()
|
13 |
-
model.half()
|
14 |
|
15 |
seg1 = '<|User|>:'
|
16 |
seg2 = f'Analyze the image in a comprehensive and detailed manner.{model.eoh}\n<|Bot|>:'
|
17 |
-
seg_emb1 = model.encode_text(seg1, add_special_tokens=True)
|
18 |
-
seg_emb2 = model.encode_text(seg2, add_special_tokens=False)
|
19 |
|
20 |
|
|
|
21 |
def detailed_caption(img_path):
|
22 |
subs = []
|
23 |
image = Image.open(img_path).convert("RGB")
|
|
|
1 |
import gradio as gr
|
2 |
import torch
|
3 |
+
import spaces
|
4 |
from PIL import Image
|
5 |
from transformers import AutoModelForCausalLM, AutoTokenizer
|
6 |
|
7 |
model_name = "Lin-Chen/ShareCaptioner"
|
8 |
tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
|
9 |
model = AutoModelForCausalLM.from_pretrained(
|
10 |
+
model_name, device_map="cpu", torch_dtype=torch.float16, trust_remote_code=True).eval()
|
11 |
model.tokenizer = tokenizer
|
12 |
|
13 |
model.cuda()
|
|
|
14 |
|
15 |
seg1 = '<|User|>:'
|
16 |
seg2 = f'Analyze the image in a comprehensive and detailed manner.{model.eoh}\n<|Bot|>:'
|
17 |
+
seg_emb1 = model.encode_text(seg1, add_special_tokens=True).cuda()
|
18 |
+
seg_emb2 = model.encode_text(seg2, add_special_tokens=False).cuda()
|
19 |
|
20 |
|
21 |
+
@spaces.GPU
|
22 |
def detailed_caption(img_path):
|
23 |
subs = []
|
24 |
image = Image.open(img_path).convert("RGB")
|
requirements.txt
CHANGED
@@ -4,10 +4,11 @@ tiktoken==0.5.1
|
|
4 |
einops==0.7.0
|
5 |
transformers_stream_generator==0.0.4
|
6 |
scipy==1.11.3
|
7 |
-
|
|
|
8 |
pillow==10.0.1
|
9 |
matplotlib==3.8.0
|
10 |
-
gradio==3.50.2
|
11 |
sentencepiece
|
12 |
urllib3==1.26.18
|
13 |
-
timm==0.
|
|
|
|
4 |
einops==0.7.0
|
5 |
transformers_stream_generator==0.0.4
|
6 |
scipy==1.11.3
|
7 |
+
torch==2.1.2
|
8 |
+
torchvision==0.16.2
|
9 |
pillow==10.0.1
|
10 |
matplotlib==3.8.0
|
|
|
11 |
sentencepiece
|
12 |
urllib3==1.26.18
|
13 |
+
timm==1.0.3
|
14 |
+
spaces
|