File size: 3,864 Bytes
4d6e8c2
fe4a4cb
 
 
ad46b79
 
 
fe4a4cb
3b09640
ad46b79
fe4a4cb
4d6e8c2
fe4a4cb
ad46b79
 
 
4d6e8c2
ad46b79
3b09640
 
 
ad46b79
4d6e8c2
 
ad46b79
1c33274
ad46b79
fe4a4cb
3b09640
1c33274
70f5f26
ad46b79
4d6e8c2
fe4a4cb
70f5f26
 
fe4a4cb
70f5f26
4d6e8c2
fe4a4cb
4d6e8c2
fe4a4cb
 
 
 
 
 
 
3b09640
ad46b79
 
 
3b09640
ad46b79
1431ab9
ad46b79
 
 
 
 
 
 
 
 
fe4a4cb
 
 
ad46b79
fe4a4cb
 
 
 
ad46b79
fe4a4cb
ad46b79
 
 
fe4a4cb
 
 
 
 
 
 
 
 
ad46b79
fe4a4cb
 
 
4d6e8c2
 
fe4a4cb
70f5f26
fe4a4cb
 
 
 
 
 
4d6e8c2
 
70f5f26
4d6e8c2
fe4a4cb
ad46b79
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
from fastapi import APIRouter
from datetime import datetime
from datasets import load_dataset
from sklearn.metrics import accuracy_score
from accelerate import Accelerator
from tqdm import tqdm
from torch.amp import autocast
import random
import os
import torch

from .utils.evaluation import AudioEvaluationRequest
from .utils.emissions import tracker, clean_emissions_data, get_space_info
from audio_utils import AudioClassifier, AudioDataset, Config, collate_fn, Evaluator
from transformers import AutoModelForImageClassification
from torch.utils.data import DataLoader

from loguru import logger
from dotenv import load_dotenv
load_dotenv()


router = APIRouter()

DESCRIPTION = "Audio pipeline to classify sounds."
ROUTE = "/audio"
device = "cuda"


@router.post(ROUTE, tags=["Audio Task"],
             description=DESCRIPTION)
async def evaluate_audio(request: AudioEvaluationRequest): #, model_path: str):
    """
    Evaluate audio classification for rainforest sound detection.
    
    Current Model: Random Baseline
    - Makes random predictions from the label space (0-1)
    - Used as a baseline for comparison
    """
    # Get space info
    username, space_url = get_space_info()

    # Define the label mapping
    LABEL_MAPPING = {
        "chainsaw": 0,
        "environment": 1
    }
    # Load and prepare the dataset
    # Because the dataset is gated, we need to use the HF_TOKEN environment variable to authenticate
    config = Config()
    accelerator = Accelerator()
    device = accelerator.device
    
    dataset = load_dataset(request.dataset_name,token=os.getenv("HF_TOKEN"))
    test_dataset = dataset["test"]
    test_dataset = test_dataset.filter(lambda x: len(x["audio"]["array"]) > 0)
    true_labels = test_dataset["label"]

    test_dataset = AudioDataset(test_dataset)
    test_loader = DataLoader(test_dataset, batch_size=2 * config.BATCH_SIZE, shuffle=False, collate_fn=collate_fn, num_workers=config.NUM_WORKERS, pin_memory=True)

    model = AudioClassifier(config.MODEL_NAME, config.MODEL_PATH)
    model, test_loader = accelerator.prepare(model, test_loader)

    # Start tracking emissions
    tracker.start()
    tracker.start_task("inference")

    #--------------------------------------------------------------------------------------------
    # YOUR MODEL INFERENCE CODE HERE
    # Update the code below to replace the random baseline by your model inference within the inference pass where the energy consumption and emissions are tracked.
    #--------------------------------------------------------------------------------------------   
    predictions = []
    
    logger.info("Running inference ...")
    evaluator = Evaluator(model, test_loader, device)
    predictions = evaluator.evaluate()
    
    #--------------------------------------------------------------------------------------------
    # YOUR MODEL INFERENCE STOPS HERE
    #--------------------------------------------------------------------------------------------   
    # Stop tracking emissions
    emissions_data = tracker.stop_task()
    
    # Calculate accuracy
    accuracy = accuracy_score(true_labels, predictions)
    print("accuracy", accuracy)
    
    # Prepare results dictionary
    results = {
        "username": username,
        "space_url": space_url,
        "submission_timestamp": datetime.now().isoformat(),
        "model_description": DESCRIPTION,
        "accuracy": float(accuracy),
        "energy_consumed_wh": emissions_data.energy_consumed * 1000,
        "emissions_gco2eq": emissions_data.emissions * 1000,
        "emissions_data": clean_emissions_data(emissions_data),
        "api_route": ROUTE,
        "dataset_config": {
            "dataset_name": request.dataset_name,
            "test_size": request.test_size,
            "test_seed": request.test_seed
        }
    }
    return results