Spaces:
Sleeping
Sleeping
Update tasks/text.py
Browse files- tasks/text.py +9 -2
tasks/text.py
CHANGED
@@ -24,6 +24,12 @@ DESCRIPTION = "First Baseline"
|
|
24 |
ROUTE = "/text"
|
25 |
|
26 |
|
|
|
|
|
|
|
|
|
|
|
|
|
27 |
MODEL = "mlp" #mlp, ct, modern
|
28 |
|
29 |
class ConspiracyClassification(
|
@@ -125,7 +131,7 @@ async def evaluate_text(request: TextEvaluationRequest):
|
|
125 |
#--------------------------------------------------------------------------------------------
|
126 |
if MODEL =="mlp":
|
127 |
model = ConspiracyClassification.from_pretrained("ypesk/frugal-ai-mlp-baseline")
|
128 |
-
|
129 |
emb_model = SentenceTransformer("paraphrase-MiniLM-L3-v2")
|
130 |
batch_size = 6
|
131 |
|
@@ -136,6 +142,7 @@ async def evaluate_text(request: TextEvaluationRequest):
|
|
136 |
|
137 |
elif MODEL == "ct":
|
138 |
model = CovidTwitterBertClassifier.from_pretrained("ypesk/ct-baseline")
|
|
|
139 |
tokenizer = AutoTokenizer.from_pretrained('digitalepidemiologylab/covid-twitter-bert')
|
140 |
|
141 |
test_texts = [t['quote'] for t in test_dataset]
|
@@ -158,7 +165,7 @@ async def evaluate_text(request: TextEvaluationRequest):
|
|
158 |
model.eval()
|
159 |
predictions = []
|
160 |
for batch in tqdm(test_dataloader):
|
161 |
-
|
162 |
with torch.no_grad():
|
163 |
if MODEL =="mlp":
|
164 |
b_texts = batch[0]
|
|
|
24 |
ROUTE = "/text"
|
25 |
|
26 |
|
27 |
+
if torch.cuda.is_available():
|
28 |
+
device = torch.device("cuda")
|
29 |
+
else:
|
30 |
+
device = torch.device("cpu")
|
31 |
+
|
32 |
+
|
33 |
MODEL = "mlp" #mlp, ct, modern
|
34 |
|
35 |
class ConspiracyClassification(
|
|
|
131 |
#--------------------------------------------------------------------------------------------
|
132 |
if MODEL =="mlp":
|
133 |
model = ConspiracyClassification.from_pretrained("ypesk/frugal-ai-mlp-baseline")
|
134 |
+
model = model.to(device)
|
135 |
emb_model = SentenceTransformer("paraphrase-MiniLM-L3-v2")
|
136 |
batch_size = 6
|
137 |
|
|
|
142 |
|
143 |
elif MODEL == "ct":
|
144 |
model = CovidTwitterBertClassifier.from_pretrained("ypesk/ct-baseline")
|
145 |
+
model = model.to(device)
|
146 |
tokenizer = AutoTokenizer.from_pretrained('digitalepidemiologylab/covid-twitter-bert')
|
147 |
|
148 |
test_texts = [t['quote'] for t in test_dataset]
|
|
|
165 |
model.eval()
|
166 |
predictions = []
|
167 |
for batch in tqdm(test_dataloader):
|
168 |
+
batch = tuple(t.to(device) for t in batch)
|
169 |
with torch.no_grad():
|
170 |
if MODEL =="mlp":
|
171 |
b_texts = batch[0]
|