File size: 4,873 Bytes
b3bb0f6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
from transformers import Qwen2_5_VLForConditionalGeneration, AutoTokenizer, AutoProcessor
from qwen_vl_utils import process_vision_info
from smolagents import tool
import torch


@tool
def video_reasoner(file_path : str, query : str) -> str: 
    """
    This tool performs requested visual reasoning task on the provided video and returns the generated output.

    Args:
        file_path: Path of a local video file on which visual reasoning is to be done.
        query: visual reasoning that is to be done.
    """
    try:
        # default: Load the model on the available device(s)
        model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
            "Qwen/Qwen2.5-VL-7B-Instruct", torch_dtype="auto", device_map="auto"
        )

        # default processer
        processor = AutoProcessor.from_pretrained("Qwen/Qwen2.5-VL-7B-Instruct")

        messages = [
        {
        "role": "user",
        "content": [
            {
                "type": "video",
                "video": file_path,
                "max_pixels": 360 * 360,
                "fps": 0.3,
            },
            {"type": "text", "text": f"{query}\n\nAdditional instruction: Treat the two types of penguins as distinct species e.g. Adelie and Emperor Penguins are considered two different species of birds."},
            ],
        }      
        ]

        # Preparation for inference
        text = processor.apply_chat_template(
            messages, tokenize=False, add_generation_prompt=True
        )
        image_inputs, video_inputs = process_vision_info(messages)
        inputs = processor(
            text=[text],
            images=image_inputs,
            videos=video_inputs,
            padding=True,
            return_tensors="pt",
        )
        inputs = inputs.to("cuda")

        # Inference: Generation of the output
        generated_ids = model.generate(**inputs, max_new_tokens=512)
        generated_ids_trimmed = [
            out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
        ]
        output_text = processor.batch_decode(
            generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
        )

        import gc
        
        # After inference
        del image_inputs
        del video_inputs
        del inputs
        del model
        del processor
        gc.collect()                 # Force Python garbage collection
        torch.cuda.empty_cache()     # Clear cached memory
        
        return output_text
    
    except Exception as e:
        return f'error occured: {e}'
    

@tool
def image_reasoner(file_path : str, query : str) -> str: 
    """
    This tool performs requested visual reasoning task on the provided image and returns the generated output.

    Args:
        file_path: Path of a local image file on which visual reasoning is to be done.
        query: visual reasoning that is to be done.
    """
    try:
        # default: Load the model on the available device(s)
        model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
            "Qwen/Qwen2.5-VL-7B-Instruct", torch_dtype="auto", device_map="auto"
        )

        # default processer
        processor = AutoProcessor.from_pretrained("Qwen/Qwen2.5-VL-7B-Instruct")

        messages = [
        {
        "role": "user",
        "content": [
            {
                "type": "image",
                "image": file_path,
            },
            {"type": "text", "text": f"{query}\n\nAdditional instruction: Review your answer for correctness."},
            ],
        }      
        ]

        # Preparation for inference
        text = processor.apply_chat_template(
            messages, tokenize=False, add_generation_prompt=True
        )
        image_inputs, video_inputs = process_vision_info(messages)
        inputs = processor(
            text=[text],
            images=image_inputs,
            videos=video_inputs,
            padding=True,
            return_tensors="pt",
        )
        inputs = inputs.to("cuda")

        # Inference: Generation of the output
        generated_ids = model.generate(**inputs, max_new_tokens=512)
        generated_ids_trimmed = [
            out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
        ]
        output_text = processor.batch_decode(
            generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
        )

        import gc
        
        # After inference
        del image_inputs
        del video_inputs
        del inputs
        del model
        del processor
        gc.collect()                 # Force Python garbage collection
        torch.cuda.empty_cache()     # Clear cached memory
        
        return output_text
    
    except Exception as e:
        return f'error occured: {e}'