Update README.md
Browse files
README.md
CHANGED
|
@@ -6,10 +6,63 @@ widget:
|
|
| 6 |
- text: Hello!
|
| 7 |
example_title: Hello world
|
| 8 |
group: Python
|
|
|
|
|
|
|
| 9 |
---
|
| 10 |
|
| 11 |
This model is for debugging. It is randomly initialized using the config from [Qwen/Qwen2-VL-7B-Instruct](https://huggingface.co/Qwen/Qwen2-VL-7B-Instruct) but with smaller size.
|
| 12 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 13 |
Codes:
|
| 14 |
```python
|
| 15 |
import os
|
|
@@ -91,4 +144,4 @@ def try_inference():
|
|
| 91 |
|
| 92 |
|
| 93 |
try_inference()
|
| 94 |
-
```
|
|
|
|
| 6 |
- text: Hello!
|
| 7 |
example_title: Hello world
|
| 8 |
group: Python
|
| 9 |
+
base_model:
|
| 10 |
+
- Qwen/Qwen2-VL-7B-Instruct
|
| 11 |
---
|
| 12 |
|
| 13 |
This model is for debugging. It is randomly initialized using the config from [Qwen/Qwen2-VL-7B-Instruct](https://huggingface.co/Qwen/Qwen2-VL-7B-Instruct) but with smaller size.
|
| 14 |
|
| 15 |
+
Usage:
|
| 16 |
+
```python
|
| 17 |
+
from PIL import Image
|
| 18 |
+
import requests
|
| 19 |
+
import torch
|
| 20 |
+
from torchvision import io
|
| 21 |
+
from typing import Dict
|
| 22 |
+
from transformers import Qwen2VLForConditionalGeneration, AutoTokenizer, AutoProcessor
|
| 23 |
+
|
| 24 |
+
model_id = "yujiepan/qwen2-vl-tiny-random"
|
| 25 |
+
|
| 26 |
+
# Load the model in half-precision on the available device(s)
|
| 27 |
+
model = Qwen2VLForConditionalGeneration.from_pretrained(
|
| 28 |
+
model_id, torch_dtype="auto", device_map="auto"
|
| 29 |
+
)
|
| 30 |
+
processor = AutoProcessor.from_pretrained(model_id)
|
| 31 |
+
|
| 32 |
+
# Image
|
| 33 |
+
url = "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-VL/assets/demo.jpeg"
|
| 34 |
+
image = Image.open(requests.get(url, stream=True).raw)
|
| 35 |
+
conversation = [
|
| 36 |
+
{
|
| 37 |
+
"role": "user",
|
| 38 |
+
"content": [
|
| 39 |
+
{
|
| 40 |
+
"type": "image",
|
| 41 |
+
},
|
| 42 |
+
{"type": "text", "text": "Describe this image."},
|
| 43 |
+
],
|
| 44 |
+
}
|
| 45 |
+
]
|
| 46 |
+
text_prompt = processor.apply_chat_template(conversation, add_generation_prompt=True)
|
| 47 |
+
# Excepted output: '<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n<|im_start|>user\n<|vision_start|><|image_pad|><|vision_end|>Describe this image.<|im_end|>\n<|im_start|>assistant\n'
|
| 48 |
+
|
| 49 |
+
inputs = processor(
|
| 50 |
+
text=[text_prompt], images=[image], padding=True, return_tensors="pt"
|
| 51 |
+
)
|
| 52 |
+
inputs = inputs.to("cuda")
|
| 53 |
+
|
| 54 |
+
output_ids = model.generate(**inputs, max_new_tokens=128)
|
| 55 |
+
generated_ids = [
|
| 56 |
+
output_ids[len(input_ids) :]
|
| 57 |
+
for input_ids, output_ids in zip(inputs.input_ids, output_ids)
|
| 58 |
+
]
|
| 59 |
+
output_text = processor.batch_decode(
|
| 60 |
+
generated_ids, skip_special_tokens=True, clean_up_tokenization_spaces=True
|
| 61 |
+
)
|
| 62 |
+
print(output_text)
|
| 63 |
+
```
|
| 64 |
+
|
| 65 |
+
|
| 66 |
Codes:
|
| 67 |
```python
|
| 68 |
import os
|
|
|
|
| 144 |
|
| 145 |
|
| 146 |
try_inference()
|
| 147 |
+
```
|