Spaces:
Sleeping
Sleeping
fix transformers
Browse files- tasks/text.py +47 -60
tasks/text.py
CHANGED
@@ -1,14 +1,13 @@
|
|
1 |
from fastapi import APIRouter
|
2 |
from datetime import datetime
|
|
|
3 |
from datasets import load_dataset
|
4 |
from sklearn.metrics import accuracy_score
|
5 |
-
import random
|
6 |
-
from transformers import pipeline, AutoConfig
|
7 |
import os
|
8 |
from concurrent.futures import ThreadPoolExecutor
|
9 |
from typing import List, Dict, Tuple
|
10 |
-
import numpy as np
|
11 |
import torch
|
|
|
12 |
|
13 |
from .utils.evaluation import TextEvaluationRequest
|
14 |
from .utils.emissions import tracker, clean_emissions_data, get_space_info
|
@@ -18,23 +17,23 @@ os.environ["TORCH_COMPILE_DISABLE"] = "1"
|
|
18 |
|
19 |
router = APIRouter()
|
20 |
|
21 |
-
DESCRIPTION = "
|
22 |
ROUTE = "/text"
|
23 |
|
24 |
class TextClassifier:
|
25 |
def __init__(self):
|
26 |
-
|
27 |
max_retries = 3
|
28 |
for attempt in range(max_retries):
|
29 |
try:
|
30 |
-
|
31 |
-
self.
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
-
|
36 |
-
batch_size=16
|
37 |
)
|
|
|
38 |
print("Model initialized successfully")
|
39 |
break
|
40 |
except Exception as e:
|
@@ -43,21 +42,37 @@ class TextClassifier:
|
|
43 |
print(f"Attempt {attempt + 1} failed, retrying...")
|
44 |
time.sleep(1)
|
45 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
46 |
def process_batch(self, batch: List[str], batch_idx: int) -> Tuple[List[int], int]:
|
47 |
"""Process a batch of texts and return their predictions"""
|
48 |
max_retries = 3
|
49 |
for attempt in range(max_retries):
|
50 |
try:
|
51 |
print(f"Processing batch {batch_idx} with {len(batch)} items (attempt {attempt + 1})")
|
52 |
-
# Process texts one by one in case of errors
|
53 |
predictions = []
|
|
|
|
|
54 |
for text in batch:
|
55 |
-
|
56 |
-
|
57 |
-
pred_label = self.label2id[pred[0]["label"]]
|
58 |
-
predictions.append(pred_label)
|
59 |
-
except Exception as e:
|
60 |
-
print(f"Error processing text in batch {batch_idx}: {str(e)}")
|
61 |
|
62 |
if not predictions:
|
63 |
raise Exception("No predictions generated for batch")
|
@@ -68,21 +83,14 @@ class TextClassifier:
|
|
68 |
except Exception as e:
|
69 |
if attempt == max_retries - 1:
|
70 |
print(f"Final error in batch {batch_idx}: {str(e)}")
|
71 |
-
return [0] * len(batch), batch_idx
|
72 |
print(f"Error in batch {batch_idx} (attempt {attempt + 1}): {str(e)}")
|
73 |
time.sleep(1)
|
74 |
|
75 |
-
|
76 |
-
@router.post(ROUTE, tags=["Text Task"],
|
77 |
-
description=DESCRIPTION)
|
78 |
async def evaluate_text(request: TextEvaluationRequest):
|
79 |
-
"""
|
80 |
-
Evaluate text classification for climate disinformation detection.
|
81 |
|
82 |
-
Current Model: Random Baseline
|
83 |
-
- Makes random predictions from the label space (0-7)
|
84 |
-
- Used as a baseline for comparison
|
85 |
-
"""
|
86 |
# Get space info
|
87 |
username, space_url = get_space_info()
|
88 |
|
@@ -100,30 +108,20 @@ async def evaluate_text(request: TextEvaluationRequest):
|
|
100 |
|
101 |
# Load and prepare the dataset
|
102 |
dataset = load_dataset(request.dataset_name)
|
103 |
-
|
104 |
-
# Convert string labels to integers
|
105 |
dataset = dataset.map(lambda x: {"label": LABEL_MAPPING[x["label"]]})
|
106 |
-
|
107 |
-
# Split dataset
|
108 |
-
train_test = dataset["train"]
|
109 |
test_dataset = dataset["test"]
|
110 |
|
111 |
# Start tracking emissions
|
112 |
tracker.start()
|
113 |
tracker.start_task("inference")
|
114 |
|
115 |
-
#--------------------------------------------------------------------------------------------
|
116 |
-
# YOUR MODEL INFERENCE CODE HERE
|
117 |
-
# 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.
|
118 |
-
#--------------------------------------------------------------------------------------------
|
119 |
-
|
120 |
true_labels = test_dataset["label"]
|
121 |
|
122 |
# Initialize the model once
|
123 |
classifier = TextClassifier()
|
124 |
|
125 |
# Prepare batches
|
126 |
-
batch_size =
|
127 |
quotes = test_dataset["quote"]
|
128 |
num_batches = len(quotes) // batch_size + (1 if len(quotes) % batch_size != 0 else 0)
|
129 |
batches = [
|
@@ -131,54 +129,44 @@ async def evaluate_text(request: TextEvaluationRequest):
|
|
131 |
for i in range(num_batches)
|
132 |
]
|
133 |
|
134 |
-
# Initialize batch_results
|
135 |
batch_results = [[] for _ in range(num_batches)]
|
136 |
|
137 |
# Process batches in parallel
|
138 |
-
max_workers = min(os.cpu_count(), 4)
|
139 |
print(f"Processing with {max_workers} workers")
|
140 |
|
141 |
with ThreadPoolExecutor(max_workers=max_workers) as executor:
|
142 |
-
# Submit all batches for processing
|
143 |
future_to_batch = {
|
144 |
-
executor.submit(
|
145 |
-
|
146 |
-
batch,
|
147 |
-
idx
|
148 |
-
): idx for idx, batch in enumerate(batches)
|
149 |
}
|
150 |
|
151 |
-
# Collect results in order
|
152 |
for future in future_to_batch:
|
153 |
batch_idx = future_to_batch[future]
|
154 |
try:
|
155 |
predictions, idx = future.result()
|
156 |
-
if predictions:
|
157 |
batch_results[idx] = predictions
|
158 |
print(f"Stored results for batch {idx} ({len(predictions)} predictions)")
|
159 |
except Exception as e:
|
160 |
print(f"Failed to get results for batch {batch_idx}: {e}")
|
161 |
-
# Use default predictions instead of empty list
|
162 |
batch_results[batch_idx] = [0] * len(batches[batch_idx])
|
163 |
|
164 |
-
# Flatten predictions
|
165 |
predictions = []
|
166 |
for batch_preds in batch_results:
|
167 |
if batch_preds is not None:
|
168 |
predictions.extend(batch_preds)
|
169 |
-
|
170 |
-
#--------------------------------------------------------------------------------------------
|
171 |
-
# YOUR MODEL INFERENCE STOPS HERE
|
172 |
-
#--------------------------------------------------------------------------------------------
|
173 |
|
174 |
# Stop tracking emissions
|
175 |
emissions_data = tracker.stop_task()
|
176 |
|
177 |
# Calculate accuracy
|
178 |
accuracy = accuracy_score(true_labels, predictions)
|
179 |
-
print("accuracy
|
180 |
|
181 |
-
# Prepare results
|
182 |
results = {
|
183 |
"username": username,
|
184 |
"space_url": space_url,
|
@@ -196,6 +184,5 @@ async def evaluate_text(request: TextEvaluationRequest):
|
|
196 |
}
|
197 |
}
|
198 |
|
199 |
-
print("results
|
200 |
-
|
201 |
return results
|
|
|
1 |
from fastapi import APIRouter
|
2 |
from datetime import datetime
|
3 |
+
import time
|
4 |
from datasets import load_dataset
|
5 |
from sklearn.metrics import accuracy_score
|
|
|
|
|
6 |
import os
|
7 |
from concurrent.futures import ThreadPoolExecutor
|
8 |
from typing import List, Dict, Tuple
|
|
|
9 |
import torch
|
10 |
+
from transformers import AutoModelForSequenceClassification, AutoTokenizer
|
11 |
|
12 |
from .utils.evaluation import TextEvaluationRequest
|
13 |
from .utils.emissions import tracker, clean_emissions_data, get_space_info
|
|
|
17 |
|
18 |
router = APIRouter()
|
19 |
|
20 |
+
DESCRIPTION = "Climate Guard Toxic Agent Classifier"
|
21 |
ROUTE = "/text"
|
22 |
|
23 |
class TextClassifier:
|
24 |
def __init__(self):
|
25 |
+
self.device = "cuda" if torch.cuda.is_available() else "cpu"
|
26 |
max_retries = 3
|
27 |
for attempt in range(max_retries):
|
28 |
try:
|
29 |
+
# Load model and tokenizer directly instead of using pipeline
|
30 |
+
self.model = AutoModelForSequenceClassification.from_pretrained(
|
31 |
+
"Tonic/climate-guard-toxic-agent"
|
32 |
+
).to(self.device)
|
33 |
+
self.tokenizer = AutoTokenizer.from_pretrained(
|
34 |
+
"Tonic/climate-guard-toxic-agent"
|
|
|
35 |
)
|
36 |
+
self.model.eval() # Set to evaluation mode
|
37 |
print("Model initialized successfully")
|
38 |
break
|
39 |
except Exception as e:
|
|
|
42 |
print(f"Attempt {attempt + 1} failed, retrying...")
|
43 |
time.sleep(1)
|
44 |
|
45 |
+
def predict_single(self, text: str) -> int:
|
46 |
+
"""Predict single text instance"""
|
47 |
+
try:
|
48 |
+
inputs = self.tokenizer(
|
49 |
+
text,
|
50 |
+
return_tensors="pt",
|
51 |
+
truncation=True,
|
52 |
+
max_length=512,
|
53 |
+
padding=True
|
54 |
+
).to(self.device)
|
55 |
+
|
56 |
+
with torch.no_grad():
|
57 |
+
outputs = self.model(**inputs)
|
58 |
+
predictions = outputs.logits.argmax(-1)
|
59 |
+
return predictions.item()
|
60 |
+
except Exception as e:
|
61 |
+
print(f"Error in single prediction: {str(e)}")
|
62 |
+
return 0 # Return default prediction on error
|
63 |
+
|
64 |
def process_batch(self, batch: List[str], batch_idx: int) -> Tuple[List[int], int]:
|
65 |
"""Process a batch of texts and return their predictions"""
|
66 |
max_retries = 3
|
67 |
for attempt in range(max_retries):
|
68 |
try:
|
69 |
print(f"Processing batch {batch_idx} with {len(batch)} items (attempt {attempt + 1})")
|
|
|
70 |
predictions = []
|
71 |
+
|
72 |
+
# Process texts one by one for better error handling
|
73 |
for text in batch:
|
74 |
+
pred = self.predict_single(text)
|
75 |
+
predictions.append(pred)
|
|
|
|
|
|
|
|
|
76 |
|
77 |
if not predictions:
|
78 |
raise Exception("No predictions generated for batch")
|
|
|
83 |
except Exception as e:
|
84 |
if attempt == max_retries - 1:
|
85 |
print(f"Final error in batch {batch_idx}: {str(e)}")
|
86 |
+
return [0] * len(batch), batch_idx
|
87 |
print(f"Error in batch {batch_idx} (attempt {attempt + 1}): {str(e)}")
|
88 |
time.sleep(1)
|
89 |
|
90 |
+
@router.post(ROUTE, tags=["Text Task"], description=DESCRIPTION)
|
|
|
|
|
91 |
async def evaluate_text(request: TextEvaluationRequest):
|
92 |
+
"""Evaluate text classification for climate disinformation detection."""
|
|
|
93 |
|
|
|
|
|
|
|
|
|
94 |
# Get space info
|
95 |
username, space_url = get_space_info()
|
96 |
|
|
|
108 |
|
109 |
# Load and prepare the dataset
|
110 |
dataset = load_dataset(request.dataset_name)
|
|
|
|
|
111 |
dataset = dataset.map(lambda x: {"label": LABEL_MAPPING[x["label"]]})
|
|
|
|
|
|
|
112 |
test_dataset = dataset["test"]
|
113 |
|
114 |
# Start tracking emissions
|
115 |
tracker.start()
|
116 |
tracker.start_task("inference")
|
117 |
|
|
|
|
|
|
|
|
|
|
|
118 |
true_labels = test_dataset["label"]
|
119 |
|
120 |
# Initialize the model once
|
121 |
classifier = TextClassifier()
|
122 |
|
123 |
# Prepare batches
|
124 |
+
batch_size = 16 # Reduced batch size for better memory management
|
125 |
quotes = test_dataset["quote"]
|
126 |
num_batches = len(quotes) // batch_size + (1 if len(quotes) % batch_size != 0 else 0)
|
127 |
batches = [
|
|
|
129 |
for i in range(num_batches)
|
130 |
]
|
131 |
|
132 |
+
# Initialize batch_results
|
133 |
batch_results = [[] for _ in range(num_batches)]
|
134 |
|
135 |
# Process batches in parallel
|
136 |
+
max_workers = min(os.cpu_count(), 4)
|
137 |
print(f"Processing with {max_workers} workers")
|
138 |
|
139 |
with ThreadPoolExecutor(max_workers=max_workers) as executor:
|
|
|
140 |
future_to_batch = {
|
141 |
+
executor.submit(classifier.process_batch, batch, idx): idx
|
142 |
+
for idx, batch in enumerate(batches)
|
|
|
|
|
|
|
143 |
}
|
144 |
|
|
|
145 |
for future in future_to_batch:
|
146 |
batch_idx = future_to_batch[future]
|
147 |
try:
|
148 |
predictions, idx = future.result()
|
149 |
+
if predictions:
|
150 |
batch_results[idx] = predictions
|
151 |
print(f"Stored results for batch {idx} ({len(predictions)} predictions)")
|
152 |
except Exception as e:
|
153 |
print(f"Failed to get results for batch {batch_idx}: {e}")
|
|
|
154 |
batch_results[batch_idx] = [0] * len(batches[batch_idx])
|
155 |
|
156 |
+
# Flatten predictions
|
157 |
predictions = []
|
158 |
for batch_preds in batch_results:
|
159 |
if batch_preds is not None:
|
160 |
predictions.extend(batch_preds)
|
|
|
|
|
|
|
|
|
161 |
|
162 |
# Stop tracking emissions
|
163 |
emissions_data = tracker.stop_task()
|
164 |
|
165 |
# Calculate accuracy
|
166 |
accuracy = accuracy_score(true_labels, predictions)
|
167 |
+
print("accuracy:", accuracy)
|
168 |
|
169 |
+
# Prepare results
|
170 |
results = {
|
171 |
"username": username,
|
172 |
"space_url": space_url,
|
|
|
184 |
}
|
185 |
}
|
186 |
|
187 |
+
print("results:", results)
|
|
|
188 |
return results
|