Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,69 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: apache-2.0
|
3 |
+
language:
|
4 |
+
- en
|
5 |
+
pipeline_tag: image-text-to-text
|
6 |
+
---
|
7 |
+
# This Model is for Educational Research Purpose Only.
|
8 |
+
|
9 |
+
# Sample Code
|
10 |
+
|
11 |
+
```
|
12 |
+
%%capture
|
13 |
+
!pip install -U bitsandbytes
|
14 |
+
|
15 |
+
from transformers import AutoProcessor, AutoModelForVision2Seq
|
16 |
+
import torch
|
17 |
+
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
|
18 |
+
|
19 |
+
processor = AutoProcessor.from_pretrained("manifestasi/smolVLM-161M-q4-manifestasi")
|
20 |
+
model = AutoModelForVision2Seq.from_pretrained("manifestasi/smolVLM-161M-q4-manifestasi",
|
21 |
+
torch_dtype=torch.float16,
|
22 |
+
_attn_implementation="eager").to(DEVICE)
|
23 |
+
|
24 |
+
from PIL import Image
|
25 |
+
from transformers.image_utils import load_image
|
26 |
+
|
27 |
+
|
28 |
+
# Load images
|
29 |
+
# image1 = load_image("https://huggingface.co/spaces/HuggingFaceTB/SmolVLM/resolve/main/example_images/rococo.jpg")
|
30 |
+
image2 = load_image("/kaggle/input/bandaraaa/799269_1200.jpg")
|
31 |
+
|
32 |
+
# Create input messages
|
33 |
+
messages = [
|
34 |
+
{
|
35 |
+
"role": "user",
|
36 |
+
"content": [
|
37 |
+
# {"type": "image"},
|
38 |
+
{"type": "image"},
|
39 |
+
{"type": "text",
|
40 |
+
"text": """
|
41 |
+
Instructions :
|
42 |
+
you are visual assistant for blind people, please answer politely and short
|
43 |
+
under 100 words.
|
44 |
+
Prompt :
|
45 |
+
can you direct me to find toilet
|
46 |
+
"""}
|
47 |
+
]
|
48 |
+
},
|
49 |
+
]
|
50 |
+
|
51 |
+
# Prepare inputs
|
52 |
+
prompt = processor.apply_chat_template(messages, add_generation_prompt=True)
|
53 |
+
# inputs = processor(text=prompt, return_tensors="pt")
|
54 |
+
inputs = processor(text=prompt, images=[image2], return_tensors="pt")
|
55 |
+
inputs = inputs.to(DEVICE)
|
56 |
+
# Generate outputs
|
57 |
+
from time import time
|
58 |
+
|
59 |
+
tim1 = time()
|
60 |
+
generated_ids = model.generate(**inputs, max_new_tokens=120)
|
61 |
+
generated_texts = processor.batch_decode(
|
62 |
+
generated_ids,
|
63 |
+
skip_special_tokens=True,
|
64 |
+
)
|
65 |
+
tim2 = time()
|
66 |
+
print(f"{(tim2 - tim1)} detik")
|
67 |
+
print(generated_texts[0].split("Assistant: ")[1])
|
68 |
+
|
69 |
+
```
|