Spaces:
Sleeping
Sleeping
Varun Wadhwa
commited on
Logs
Browse files
app.py
CHANGED
@@ -151,8 +151,8 @@ def evaluate_model(model, dataloader, device):
|
|
151 |
# Process each sequence in the batch
|
152 |
for i in range(current_batch_size):
|
153 |
valid_mask = (labels[i] != -100) & (attention_mask[i] != 0)
|
154 |
-
valid_preds = preds[i][valid_mask].flatten()
|
155 |
-
valid_labels = labels[i][valid_mask].flatten()
|
156 |
print(valid_mask.dtype)
|
157 |
print(labels[i].shape)
|
158 |
print(labels[i])
|
@@ -160,11 +160,7 @@ def evaluate_model(model, dataloader, device):
|
|
160 |
print(valid_mask.shape)
|
161 |
print(valid_labels)
|
162 |
print(valid_mask)
|
163 |
-
all_preds.extend(valid_preds.tolist())
|
164 |
-
all_labels.extend(valid_labels.tolist())
|
165 |
-
|
166 |
assert not torch.any(valid_labels == -100), f"Found -100 in valid_labels for batch {i}"
|
167 |
-
|
168 |
if sample_count < num_samples:
|
169 |
print(f"Sample {sample_count + 1}:")
|
170 |
print(f"Tokens: {tokenizer.convert_ids_to_tokens(input_ids[i])}")
|
@@ -172,6 +168,8 @@ def evaluate_model(model, dataloader, device):
|
|
172 |
print(f"Predicted Labels: {[id2label[pred] for pred in valid_preds]}")
|
173 |
print("-" * 50)
|
174 |
sample_count += 1
|
|
|
|
|
175 |
|
176 |
# Calculate evaluation metrics
|
177 |
print("evaluate_model sizes")
|
|
|
151 |
# Process each sequence in the batch
|
152 |
for i in range(current_batch_size):
|
153 |
valid_mask = (labels[i] != -100) & (attention_mask[i] != 0)
|
154 |
+
valid_preds = preds[i][valid_mask[i]].flatten()
|
155 |
+
valid_labels = labels[i][valid_mask[i]].flatten()
|
156 |
print(valid_mask.dtype)
|
157 |
print(labels[i].shape)
|
158 |
print(labels[i])
|
|
|
160 |
print(valid_mask.shape)
|
161 |
print(valid_labels)
|
162 |
print(valid_mask)
|
|
|
|
|
|
|
163 |
assert not torch.any(valid_labels == -100), f"Found -100 in valid_labels for batch {i}"
|
|
|
164 |
if sample_count < num_samples:
|
165 |
print(f"Sample {sample_count + 1}:")
|
166 |
print(f"Tokens: {tokenizer.convert_ids_to_tokens(input_ids[i])}")
|
|
|
168 |
print(f"Predicted Labels: {[id2label[pred] for pred in valid_preds]}")
|
169 |
print("-" * 50)
|
170 |
sample_count += 1
|
171 |
+
all_preds.extend(valid_preds.tolist())
|
172 |
+
all_labels.extend(valid_labels.tolist())
|
173 |
|
174 |
# Calculate evaluation metrics
|
175 |
print("evaluate_model sizes")
|