File size: 3,470 Bytes
2a4f1d3
9559a52
 
 
 
 
0410591
 
2a4f1d3
 
9559a52
 
 
2a4f1d3
 
9086f73
9559a52
0410591
 
 
 
2a4f1d3
0410591
2a4f1d3
0410591
2a4f1d3
0410591
 
2a4f1d3
 
0410591
9559a52
 
0410591
 
9559a52
35f3879
9559a52
 
 
2a4f1d3
35f3879
2a4f1d3
 
 
9559a52
2a4f1d3
 
 
 
 
9559a52
 
 
 
 
 
 
2a4f1d3
 
 
 
 
 
 
 
 
9559a52
2a4f1d3
9559a52
 
 
 
2a4f1d3
 
 
 
0f6bee9
 
9559a52
0f6bee9
9559a52
 
0f6bee9
 
 
9559a52
0f6bee9
 
9559a52
0f6bee9
9559a52
 
 
 
0f6bee9
 
 
 
9559a52
 
 
 
 
 
 
 
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
from fastapi import FastAPI, Query
from transformers import (
    AutoProcessor,
    AutoModelForCausalLM,
    AutoTokenizer,
)
from transformers import Qwen2_5_VLForConditionalGeneration

from qwen_vl_utils import process_vision_info
import torch
import logging

logging.basicConfig(level=logging.INFO)

app = FastAPI()
 


checkpoint = "Qwen/Qwen2.5-VL-3B-Instruct"
min_pixels = 256*28*28
max_pixels = 1280*28*28
processor = AutoProcessor.from_pretrained(
    checkpoint,
    min_pixels=min_pixels,
    max_pixels=max_pixels
)
qwen_model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
    checkpoint,
    torch_dtype=torch.bfloat16,
    device_map="auto",
    # attn_implementation="flash_attention_2",
)



# LLaMA Model Setup
llama_model_name = "Orenguteng/Llama-3.1-8B-Lexi-Uncensored-V2"
llama_tokenizer = AutoTokenizer.from_pretrained(llama_model_name)
llama_model = AutoModelForCausalLM.from_pretrained(
    llama_model_name, torch_dtype=torch.float16, device_map="auto"
)
 

@app.get("/")
def read_root():
    return {"message": "API is live. Use the /predict, /chat, or /llama_chat endpoints."}

@app.get("/predict")
def predict(image_url: str = Query(...), prompt: str = Query(...)):
    messages = [
        {"role": "system", "content": "You are a helpful assistant with vision abilities."},
        {
            "role": "user",
            "content": [
                {"type": "image", "image": image_url},
                {"type": "text", "text": prompt},
            ],
        },
    ]
    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",
    ).to(qwen_model.device)
    with torch.no_grad():
        generated_ids = qwen_model.generate(**inputs, max_new_tokens=128)
    generated_ids_trimmed = [
        out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs["input_ids"], generated_ids)
    ]
    output_texts = processor.batch_decode(
        generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
    )
    return {"response": output_texts[0]}

@app.get("/chat")
def chat(prompt: str = Query(...)):
    messages = [
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": [{"type": "text", "text": prompt}]},
    ]
    text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
    inputs = processor(
        text=[text],
        padding=True,
        return_tensors="pt",
    ).to(qwen_model.device)
    with torch.no_grad():
        generated_ids = qwen_model.generate(**inputs, max_new_tokens=128)
    generated_ids_trimmed = [
        out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs["input_ids"], generated_ids)
    ]
    output_texts = processor.batch_decode(
        generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
    )
    return {"response": output_texts[0]}

@app.get("/llama_chat")
def llama_chat(prompt: str = Query(...)):
    inputs = llama_tokenizer(prompt, return_tensors="pt").to(llama_model.device)
    with torch.no_grad():
        outputs = llama_model.generate(**inputs, max_new_tokens=128)
    response = llama_tokenizer.decode(outputs[0], skip_special_tokens=True)
    return {"response": response}