Update README.md to match chat-v1.1 model
Browse files
README.md
CHANGED
|
@@ -29,9 +29,37 @@ model = AutoModelForCausalLM.from_pretrained(
|
|
| 29 |
trust_remote_code=True
|
| 30 |
).to('cuda').eval()
|
| 31 |
|
|
|
|
|
|
|
|
|
|
| 32 |
query = 'Describe this image'
|
| 33 |
image = Image.open(requests.get('https://github.com/THUDM/CogVLM/blob/main/examples/1.png?raw=true', stream=True).raw).convert('RGB')
|
| 34 |
-
inputs = model.build_conversation_input_ids(tokenizer, query=query, history=[], images=[image])
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 35 |
inputs = {
|
| 36 |
'input_ids': inputs['input_ids'].unsqueeze(0).to('cuda'),
|
| 37 |
'token_type_ids': inputs['token_type_ids'].unsqueeze(0).to('cuda'),
|
|
@@ -45,9 +73,7 @@ with torch.no_grad():
|
|
| 45 |
outputs = outputs[:, inputs['input_ids'].shape[1]:]
|
| 46 |
print(tokenizer.decode(outputs[0]))
|
| 47 |
|
| 48 |
-
#
|
| 49 |
-
# number 24 on it. He is holding a brown basketball. On the right side, there is another player wearing a blue and red jersey, blocking Kobe's
|
| 50 |
-
# movement. Behind them, there are many spectators watching the game.</s>
|
| 51 |
```
|
| 52 |
|
| 53 |
# 方法(Method)
|
|
|
|
| 29 |
trust_remote_code=True
|
| 30 |
).to('cuda').eval()
|
| 31 |
|
| 32 |
+
|
| 33 |
+
# chat example
|
| 34 |
+
|
| 35 |
query = 'Describe this image'
|
| 36 |
image = Image.open(requests.get('https://github.com/THUDM/CogVLM/blob/main/examples/1.png?raw=true', stream=True).raw).convert('RGB')
|
| 37 |
+
inputs = model.build_conversation_input_ids(tokenizer, query=query, history=[], images=[image]) # chat mode
|
| 38 |
+
inputs = {
|
| 39 |
+
'input_ids': inputs['input_ids'].unsqueeze(0).to('cuda'),
|
| 40 |
+
'token_type_ids': inputs['token_type_ids'].unsqueeze(0).to('cuda'),
|
| 41 |
+
'attention_mask': inputs['attention_mask'].unsqueeze(0).to('cuda'),
|
| 42 |
+
'images': [[inputs['images'][0].to('cuda').to(torch.bfloat16)]],
|
| 43 |
+
}
|
| 44 |
+
gen_kwargs = {"max_length": 2048, "do_sample": False}
|
| 45 |
+
|
| 46 |
+
with torch.no_grad():
|
| 47 |
+
outputs = model.generate(**inputs, **gen_kwargs)
|
| 48 |
+
outputs = outputs[:, inputs['input_ids'].shape[1]:]
|
| 49 |
+
print(tokenizer.decode(outputs[0]))
|
| 50 |
+
|
| 51 |
+
# This image captures a moment from a basketball game. Two players are prominently featured: one wearing a yellow jersey with the number
|
| 52 |
+
# 24 and the word 'Lakers' written on it, and the other wearing a navy blue jersey with the word 'Washington' and the number 34. The player
|
| 53 |
+
# in yellow is holding a basketball and appears to be dribbling it, while the player in navy blue is reaching out with his arm, possibly
|
| 54 |
+
# trying to block or defend. The background shows a filled stadium with spectators, indicating that this is a professional game.</s>
|
| 55 |
+
|
| 56 |
+
|
| 57 |
+
|
| 58 |
+
# vqa example
|
| 59 |
+
|
| 60 |
+
query = 'How many houses are there in this cartoon?'
|
| 61 |
+
image = Image.open(requests.get('https://github.com/THUDM/CogVLM/blob/main/examples/4.jpg?raw=true', stream=True).raw).convert('RGB')
|
| 62 |
+
inputs = model.build_conversation_input_ids(tokenizer, query=query, history=[], images=[image], template_version='vqa') # vqa mode
|
| 63 |
inputs = {
|
| 64 |
'input_ids': inputs['input_ids'].unsqueeze(0).to('cuda'),
|
| 65 |
'token_type_ids': inputs['token_type_ids'].unsqueeze(0).to('cuda'),
|
|
|
|
| 73 |
outputs = outputs[:, inputs['input_ids'].shape[1]:]
|
| 74 |
print(tokenizer.decode(outputs[0]))
|
| 75 |
|
| 76 |
+
# 4
|
|
|
|
|
|
|
| 77 |
```
|
| 78 |
|
| 79 |
# 方法(Method)
|