File size: 6,177 Bytes
4d6e8c2
 
 
 
 
32590b1
a66df45
 
b134905
a66df45
 
4d6e8c2
 
 
 
a66df45
 
 
4d6e8c2
 
70f5f26
1c33274
70f5f26
f5ac2a0
 
 
 
 
 
 
 
 
 
 
93741cc
f5ac2a0
93741cc
 
 
 
 
 
 
 
f5ac2a0
 
93741cc
f5ac2a0
93741cc
 
 
f5ac2a0
93741cc
 
f5ac2a0
1c33274
70f5f26
4d6e8c2
 
70f5f26
 
 
 
 
4d6e8c2
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
76fccaf
 
4d6e8c2
 
 
 
70f5f26
 
 
 
 
4d6e8c2
 
92fa037
a66df45
f5ac2a0
a66df45
 
 
 
 
 
 
 
 
 
92fa037
 
 
a66df45
 
 
 
 
 
93741cc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
92fa037
 
a66df45
70f5f26
 
 
4d6e8c2
 
 
 
 
 
5518620
4d6e8c2
 
 
 
 
 
70f5f26
4d6e8c2
 
 
 
1c33274
4d6e8c2
 
 
 
 
 
5518620
 
4d6e8c2
92fa037
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
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
from fastapi import APIRouter
from datetime import datetime
from datasets import load_dataset
from sklearn.metrics import accuracy_score
import random
from transformers import pipeline, AutoConfig
import os
from concurrent.futures import ThreadPoolExecutor
from typing import List, Dict, Tuple
import numpy as np
import torch

from .utils.evaluation import TextEvaluationRequest
from .utils.emissions import tracker, clean_emissions_data, get_space_info

# Disable torch compile
os.environ["TORCH_COMPILE_DISABLE"] = "1"

router = APIRouter()

DESCRIPTION = "Random Baseline"
ROUTE = "/text"

class TextClassifier:
    def __init__(self):
        self.config = AutoConfig.from_pretrained("camillebrl/ModernBERT-envclaims-overfit")
        self.label2id = self.config.label2id
        self.classifier = pipeline(
            "text-classification",
            "camillebrl/ModernBERT-envclaims-overfit",
            device="cpu",
            batch_size=16
        )

    def process_batch(self, batch: List[str], batch_idx: int) -> Tuple[List[int], int]:
        """
        Process a batch of texts and return their predictions along with batch index
        
        Args:
            batch: List of texts to process
            batch_idx: Index of the current batch
            
        Returns:
            Tuple containing list of predictions and batch index
        """
        try:
            print(f"Processing batch {batch_idx} with {len(batch)} items")
            batch_preds = self.classifier(list(batch))
            predictions = [self.label2id[pred[0]["label"]] for pred in batch_preds]
            print(f"Completed batch {batch_idx} with {len(predictions)} predictions")
            return predictions, batch_idx
        except Exception as e:
            print(f"Error in batch {batch_idx}: {str(e)}")
            return [], batch_idx

@router.post(ROUTE, tags=["Text Task"], 
             description=DESCRIPTION)
async def evaluate_text(request: TextEvaluationRequest):
    """
    Evaluate text classification for climate disinformation detection.
    
    Current Model: Random Baseline
    - Makes random predictions from the label space (0-7)
    - Used as a baseline for comparison
    """
    # Get space info
    username, space_url = get_space_info()

    # Define the label mapping
    LABEL_MAPPING = {
        "0_not_relevant": 0,
        "1_not_happening": 1,
        "2_not_human": 2,
        "3_not_bad": 3,
        "4_solutions_harmful_unnecessary": 4,
        "5_science_unreliable": 5,
        "6_proponents_biased": 6,
        "7_fossil_fuels_needed": 7
    }

    # Load and prepare the dataset
    dataset = load_dataset(request.dataset_name)

    # Convert string labels to integers
    dataset = dataset.map(lambda x: {"label": LABEL_MAPPING[x["label"]]})

    # Split dataset
    train_test = dataset["train"]
    test_dataset = dataset["test"]
    
    # 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.
    #--------------------------------------------------------------------------------------------   
    
    true_labels = test_dataset["label"]
    
    # Initialize the model once
    classifier = TextClassifier()

    # Prepare batches
    batch_size = 32
    quotes = test_dataset["quote"]
    num_batches = len(quotes) // batch_size + (1 if len(quotes) % batch_size != 0 else 0)
    batches = [
        quotes[i * batch_size:(i + 1) * batch_size]
        for i in range(num_batches)
    ]

    # Initialize batch_results before parallel processing
    batch_results = [[] for _ in range(num_batches)]
    
    # Process batches in parallel
    max_workers = min(os.cpu_count(), 4)  # Limit to 4 workers or CPU count
    print(f"Processing with {max_workers} workers")
    
    with ThreadPoolExecutor(max_workers=max_workers) as executor:
        # Submit all batches for processing
        future_to_batch = {
            executor.submit(
                classifier.process_batch, 
                batch, 
                idx
            ): idx for idx, batch in enumerate(batches)
        }

        # Collect results in order
        for future in future_to_batch:
            batch_idx = future_to_batch[future]
            try:
                predictions, idx = future.result()
                batch_results[idx] = predictions
                print(f"Stored results for batch {idx}")
            except Exception as e:
                print(f"Failed to get results for batch {batch_idx}: {e}")
                batch_results[batch_idx] = []

    # Flatten predictions while maintaining order
    predictions = [pred for batch_preds in batch_results for pred in batch_preds]
    print(f"Total predictions collected: {len(predictions)}")

    #--------------------------------------------------------------------------------------------
    # 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
        }
    }

    print("results : ", results)
    
    return results