Spaces:
Build error
Build error
fix the data type
Browse files- app.py +1 -2
- requirements.txt +1 -1
- sescore.py +5 -8
app.py
CHANGED
@@ -1,6 +1,5 @@
|
|
1 |
import evaluate
|
2 |
from evaluate.utils import launch_gradio_widget
|
3 |
|
4 |
-
|
5 |
module = evaluate.load("xu1998hz/sescore")
|
6 |
-
launch_gradio_widget(module)
|
|
|
1 |
import evaluate
|
2 |
from evaluate.utils import launch_gradio_widget
|
3 |
|
|
|
4 |
module = evaluate.load("xu1998hz/sescore")
|
5 |
+
launch_gradio_widget(module)
|
requirements.txt
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
git+https://github.com/huggingface/evaluate@main
|
2 |
-
gdown
|
3 |
unbabel-comet
|
|
|
|
1 |
git+https://github.com/huggingface/evaluate@main
|
|
|
2 |
unbabel-comet
|
3 |
+
torch
|
sescore.py
CHANGED
@@ -107,8 +107,8 @@ class SEScore(evaluate.Metric):
|
|
107 |
inputs_description=_KWARGS_DESCRIPTION,
|
108 |
# This defines the format of each prediction and reference
|
109 |
features=datasets.Features({
|
110 |
-
'predictions': datasets.Value(
|
111 |
-
'references': datasets.Value(
|
112 |
}),
|
113 |
# Homepage of the module for documentation
|
114 |
homepage="http://module.homepage",
|
@@ -121,23 +121,20 @@ class SEScore(evaluate.Metric):
|
|
121 |
"""download SEScore checkpoints to compute the scores"""
|
122 |
# Download SEScore checkpoint
|
123 |
from comet import load_from_checkpoint
|
124 |
-
import gdown
|
125 |
import os
|
126 |
from huggingface_hub import snapshot_download
|
127 |
# initialize roberta into str2encoder
|
128 |
comet.encoders.str2encoder['RoBERTa'] = robertaEncoder
|
129 |
-
# url = "https://drive.google.com/uc?id=1QgMP_Y4QCbvDMTeVacYt0J76OYvwWK9V&export=download&confirm=true"
|
130 |
-
# output = 'sescore_ckpt.gz'
|
131 |
-
# gdown.download(url, output, quiet=False)
|
132 |
-
# cmd = 'tar -xvf sescore_ckpt.gz'
|
133 |
-
# os.system(cmd)
|
134 |
destination = snapshot_download(repo_id="xu1998hz/sescore_english_mt", revision="main")
|
135 |
self.scorer = load_from_checkpoint(f'{destination}/checkpoint/sescore_english_mt.ckpt')
|
136 |
|
137 |
def _compute(self, predictions, references, gpus=None, progress_bar=False):
|
138 |
if gpus is None:
|
139 |
gpus = 1 if torch.cuda.is_available() else 0
|
|
|
140 |
data = {"src": references, "mt": predictions}
|
|
|
141 |
data = [dict(zip(data, t)) for t in zip(*data.values())]
|
|
|
142 |
scores, mean_score = self.scorer.predict(data, gpus=gpus, progress_bar=progress_bar)
|
143 |
return {"mean_score": mean_score, "scores": scores}
|
|
|
107 |
inputs_description=_KWARGS_DESCRIPTION,
|
108 |
# This defines the format of each prediction and reference
|
109 |
features=datasets.Features({
|
110 |
+
'predictions': datasets.Value("string", id="sequence"),
|
111 |
+
'references': datasets.Value("string", id="sequence"),
|
112 |
}),
|
113 |
# Homepage of the module for documentation
|
114 |
homepage="http://module.homepage",
|
|
|
121 |
"""download SEScore checkpoints to compute the scores"""
|
122 |
# Download SEScore checkpoint
|
123 |
from comet import load_from_checkpoint
|
|
|
124 |
import os
|
125 |
from huggingface_hub import snapshot_download
|
126 |
# initialize roberta into str2encoder
|
127 |
comet.encoders.str2encoder['RoBERTa'] = robertaEncoder
|
|
|
|
|
|
|
|
|
|
|
128 |
destination = snapshot_download(repo_id="xu1998hz/sescore_english_mt", revision="main")
|
129 |
self.scorer = load_from_checkpoint(f'{destination}/checkpoint/sescore_english_mt.ckpt')
|
130 |
|
131 |
def _compute(self, predictions, references, gpus=None, progress_bar=False):
|
132 |
if gpus is None:
|
133 |
gpus = 1 if torch.cuda.is_available() else 0
|
134 |
+
|
135 |
data = {"src": references, "mt": predictions}
|
136 |
+
print(data)
|
137 |
data = [dict(zip(data, t)) for t in zip(*data.values())]
|
138 |
+
print(data)
|
139 |
scores, mean_score = self.scorer.predict(data, gpus=gpus, progress_bar=progress_bar)
|
140 |
return {"mean_score": mean_score, "scores": scores}
|