updated readme
Browse files- .gitattributes +15 -0
- 7690148_urls.txt +7 -0
- README.md +24 -22
- dcase23-task2-enriched.py +46 -55
.gitattributes
CHANGED
@@ -58,3 +58,18 @@ data/MIT_ast-finetuned-audioset-10-10-0-4593-embeddings_dev_train.npz filter=lfs
|
|
58 |
data/MIT_ast-finetuned-audioset-10-10-0-4593-embeddings_dev_test.npz filter=lfs diff=lfs merge=lfs -text
|
59 |
data/PCA_REDUCED_dcase2023_task2_baseline_ae-embeddings_dev_test.npz filter=lfs diff=lfs merge=lfs -text
|
60 |
data/PCA_REDUCED_dcase2023_task2_baseline_ae-embeddings_dev_train.npz filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
58 |
data/MIT_ast-finetuned-audioset-10-10-0-4593-embeddings_dev_test.npz filter=lfs diff=lfs merge=lfs -text
|
59 |
data/PCA_REDUCED_dcase2023_task2_baseline_ae-embeddings_dev_test.npz filter=lfs diff=lfs merge=lfs -text
|
60 |
data/PCA_REDUCED_dcase2023_task2_baseline_ae-embeddings_dev_train.npz filter=lfs diff=lfs merge=lfs -text
|
61 |
+
data/preview_dcase_2.png filter=lfs diff=lfs merge=lfs -text
|
62 |
+
data/preview_dcase.png filter=lfs diff=lfs merge=lfs -text
|
63 |
+
data/spotlight_save_layout.png filter=lfs diff=lfs merge=lfs -text
|
64 |
+
data/config-spotlight-layout.json filter=lfs diff=lfs merge=lfs -text
|
65 |
+
data/dcase2023_task2_baseline_ae-embeddings_dev_test.npz filter=lfs diff=lfs merge=lfs -text
|
66 |
+
data/dev_test.tar.gz.lock filter=lfs diff=lfs merge=lfs -text
|
67 |
+
data/dev_train.tar.gz filter=lfs diff=lfs merge=lfs -text
|
68 |
+
data/preview_dcase_1.png filter=lfs diff=lfs merge=lfs -text
|
69 |
+
data/dcase2023_task2_baseline_ae-embeddings_dev_train.npz filter=lfs diff=lfs merge=lfs -text
|
70 |
+
data/dev_test.tar.gz filter=lfs diff=lfs merge=lfs -text
|
71 |
+
data/dev_train.tar.gz.lock filter=lfs diff=lfs merge=lfs -text
|
72 |
+
data/MIT_ast-finetuned-audioset-10-10-0.4593-embeddings_add_train.npz filter=lfs diff=lfs merge=lfs -text
|
73 |
+
data/MIT_ast-finetuned-audioset-10-10-0.4593-embeddings_eval_test.npz filter=lfs diff=lfs merge=lfs -text
|
74 |
+
data/add_train.tar.gz filter=lfs diff=lfs merge=lfs -text
|
75 |
+
data/eval_test.tar.gz filter=lfs diff=lfs merge=lfs -text
|
7690148_urls.txt
ADDED
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
https://zenodo.org/record/7690148/files/dev_bearing.zip
|
2 |
+
https://zenodo.org/record/7690148/files/dev_fan.zip
|
3 |
+
https://zenodo.org/record/7690148/files/dev_gearbox.zip
|
4 |
+
https://zenodo.org/record/7690148/files/dev_slider.zip
|
5 |
+
https://zenodo.org/record/7690148/files/dev_ToyCar.zip
|
6 |
+
https://zenodo.org/record/7690148/files/dev_ToyTrain.zip
|
7 |
+
https://zenodo.org/record/7690148/files/dev_valve.zip
|
README.md
CHANGED
@@ -160,10 +160,6 @@ a ClassLabel for the label and a ClassLabel for the class.
|
|
160 |
'embeddings_dcase2023_task2_baseline_ae': [12.602639198303223,
|
161 |
16.997364044189453, ...,
|
162 |
-0.20931333303451538]
|
163 |
-
'anomaly_score_dcase2023_task2_baseline_ae': 8.284389
|
164 |
-
'prediction_dcase2023_task2_baseline_ae': 0
|
165 |
-
'prediction_correct_dcase2023_task2_baseline_ae': 1
|
166 |
-
'anomaly_score_embedding_lof': 0.191043
|
167 |
}
|
168 |
```
|
169 |
|
@@ -183,11 +179,6 @@ The length of each audio file is 10 seconds.
|
|
183 |
- `label`: an integer whose value may be either _0_, indicating that the audio sample is _normal_, _1_, indicating that the audio sample contains an _anomaly_.
|
184 |
- `embeddings_ast-finetuned-audioset-10-10-0.4593`: an `datasets.Sequence(Value("float32"), shape=(1, 768))` representing audio embeddings that are generated with an [Audio Spectrogram Transformer](https://huggingface.co/docs/transformers/model_doc/audio-spectrogram-transformer#transformers.ASTFeatureExtractor).
|
185 |
- `embeddings_dcase2023_task2_baseline_ae`: an `datasets.Sequence(Value("float32"), shape=(1, 512))` representing audio embeddings that are generated with the [**DCASE 2023 Challenge Task 2 Baseline Auto Encoder**](https://github.com/nttcslab/dcase2023_task2_baseline_ae). **Seven individual class-specific AEs** are trained. Dimensionality Reduction is applied with **PCA** separately for each class with a fit on the respecting training set of samples.
|
186 |
-
- `anomaly_score_dcase2023_task2_baseline_ae`: a float representation of the anomaly score according to the baseline implementation
|
187 |
-
- `prediction_dcase2023_task2_baseline_ae`: an integer whose value may be either _0_, indicating that the audio sample is considered _normal_ by the baseline algorithm, _1_, indicating that the audio sample contains an _anomaly_.
|
188 |
-
- `prediction_correct_dcase2023_task2_baseline_ae`: an integer whose value may be either _0_, indicating that the baseline prediction is wrong or _1_, indicating that prediction is correct.
|
189 |
-
- `anomaly_score_embedding_lof`: a float representation of the anomaly score computed with the PyOD implementation of the Local Outlier Factor algorithm on the pre-computed embedding.
|
190 |
-
|
191 |
|
192 |
### Data Splits
|
193 |
|
@@ -198,7 +189,18 @@ The development dataset has 2 splits: _train_ and _test_.
|
|
198 |
| Train | 7000 | 6930 / 70 |
|
199 |
| Test | 1400 | 700 / 700 |
|
200 |
|
201 |
-
The
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
202 |
|
203 |
## Dataset Creation
|
204 |
|
@@ -300,22 +302,22 @@ If you use this dataset, please cite all the following papers. We will publish a
|
|
300 |
- Noboru Harada, Daisuke Niizumi, Daiki Takeuchi, Yasunori Ohishi, and Masahiro Yasuda. First-shot anomaly detection for machine condition monitoring: a domain generalization baseline. In arXiv e-prints: 2303.00455, 2023. [[URL](https://arxiv.org/abs/2303.00455.pdf)]
|
301 |
|
302 |
```
|
303 |
-
@dataset{
|
304 |
author = {Kota Dohi and
|
305 |
-
Keisuke and
|
306 |
-
Noboru and
|
307 |
-
Daisuke and
|
308 |
-
Yuma and
|
309 |
-
Tomoya and
|
310 |
-
Harsh and
|
311 |
-
Takashi and
|
312 |
-
Yohei},
|
313 |
title = {DCASE 2023 Challenge Task 2 Development Dataset},
|
314 |
month = mar,
|
315 |
year = 2023,
|
316 |
publisher = {Zenodo},
|
317 |
-
version = {
|
318 |
-
doi = {10.5281/zenodo.
|
319 |
-
url = {https://doi.org/10.5281/zenodo.
|
320 |
}
|
321 |
```
|
|
|
160 |
'embeddings_dcase2023_task2_baseline_ae': [12.602639198303223,
|
161 |
16.997364044189453, ...,
|
162 |
-0.20931333303451538]
|
|
|
|
|
|
|
|
|
163 |
}
|
164 |
```
|
165 |
|
|
|
179 |
- `label`: an integer whose value may be either _0_, indicating that the audio sample is _normal_, _1_, indicating that the audio sample contains an _anomaly_.
|
180 |
- `embeddings_ast-finetuned-audioset-10-10-0.4593`: an `datasets.Sequence(Value("float32"), shape=(1, 768))` representing audio embeddings that are generated with an [Audio Spectrogram Transformer](https://huggingface.co/docs/transformers/model_doc/audio-spectrogram-transformer#transformers.ASTFeatureExtractor).
|
181 |
- `embeddings_dcase2023_task2_baseline_ae`: an `datasets.Sequence(Value("float32"), shape=(1, 512))` representing audio embeddings that are generated with the [**DCASE 2023 Challenge Task 2 Baseline Auto Encoder**](https://github.com/nttcslab/dcase2023_task2_baseline_ae). **Seven individual class-specific AEs** are trained. Dimensionality Reduction is applied with **PCA** separately for each class with a fit on the respecting training set of samples.
|
|
|
|
|
|
|
|
|
|
|
182 |
|
183 |
### Data Splits
|
184 |
|
|
|
189 |
| Train | 7000 | 6930 / 70 |
|
190 |
| Test | 1400 | 700 / 700 |
|
191 |
|
192 |
+
The additional training dataset has 1 split: _train_.
|
193 |
+
|
194 |
+
| Dataset Split | Number of Instances in Split | Source Domain / Target Domain Samples |
|
195 |
+
| ------------- |------------------------------|---------------------------------------|
|
196 |
+
| Train | 7000 | 6930 / 70 |
|
197 |
+
|
198 |
+
The evaluation dataset has 1 split: _test_.
|
199 |
+
|
200 |
+
| Dataset Split | Number of Instances in Split | Source Domain / Target Domain Samples |
|
201 |
+
|---------------|------------------------------|---------------------------------------|
|
202 |
+
| Test | 1400 | ? |
|
203 |
+
|
204 |
|
205 |
## Dataset Creation
|
206 |
|
|
|
302 |
- Noboru Harada, Daisuke Niizumi, Daiki Takeuchi, Yasunori Ohishi, and Masahiro Yasuda. First-shot anomaly detection for machine condition monitoring: a domain generalization baseline. In arXiv e-prints: 2303.00455, 2023. [[URL](https://arxiv.org/abs/2303.00455.pdf)]
|
303 |
|
304 |
```
|
305 |
+
@dataset{kota_dohi_2023_7882613,
|
306 |
author = {Kota Dohi and
|
307 |
+
Keisuke Imoto and
|
308 |
+
Noboru Harada and
|
309 |
+
Daisuke Niizumi and
|
310 |
+
Yuma Koizumi and
|
311 |
+
Tomoya Nishida and
|
312 |
+
Harsh Purohit and
|
313 |
+
Takashi Endo and
|
314 |
+
Yohei Kawaguchi},
|
315 |
title = {DCASE 2023 Challenge Task 2 Development Dataset},
|
316 |
month = mar,
|
317 |
year = 2023,
|
318 |
publisher = {Zenodo},
|
319 |
+
version = {3.0},
|
320 |
+
doi = {10.5281/zenodo.7882613},
|
321 |
+
url = {https://doi.org/10.5281/zenodo.7882613}
|
322 |
}
|
323 |
```
|
dcase23-task2-enriched.py
CHANGED
@@ -13,23 +13,23 @@ from typing import Iterable, Dict, Optional, Union, List
|
|
13 |
|
14 |
|
15 |
_CITATION = """\
|
16 |
-
@dataset{
|
17 |
author = {Kota Dohi and
|
18 |
-
Keisuke and
|
19 |
-
Noboru and
|
20 |
-
Daisuke and
|
21 |
-
Yuma and
|
22 |
-
Tomoya and
|
23 |
-
Harsh and
|
24 |
-
Takashi and
|
25 |
-
Yohei},
|
26 |
title = {DCASE 2023 Challenge Task 2 Development Dataset},
|
27 |
month = mar,
|
28 |
year = 2023,
|
29 |
publisher = {Zenodo},
|
30 |
-
version = {
|
31 |
-
doi = {10.5281/zenodo.
|
32 |
-
url = {https://doi.org/10.5281/zenodo.
|
33 |
}
|
34 |
"""
|
35 |
_LICENSE = "Creative Commons Attribution 4.0 International Public License"
|
@@ -37,13 +37,13 @@ _LICENSE = "Creative Commons Attribution 4.0 International Public License"
|
|
37 |
_METADATA_REG = r"attributes_\d+.csv"
|
38 |
|
39 |
_NUM_TARGETS = 2
|
40 |
-
_NUM_CLASSES =
|
41 |
|
42 |
_TARGET_NAMES = ["normal", "anomaly"]
|
43 |
-
_CLASS_NAMES = ["gearbox", "fan", "bearing", "slider", "ToyCar", "ToyTrain", "valve"]
|
44 |
|
45 |
_HOMEPAGE = {
|
46 |
-
"dev": "https://zenodo.org/record/
|
47 |
"add": "",
|
48 |
"eval": "",
|
49 |
}
|
@@ -52,24 +52,23 @@ DATA_URLS = {
|
|
52 |
"dev": {
|
53 |
"train": "data/dev_train.tar.gz",
|
54 |
"test": "data/dev_test.tar.gz",
|
55 |
-
"metadata": "data/
|
56 |
},
|
57 |
"add": {
|
58 |
"train": "data/add_train.tar.gz",
|
59 |
-
"
|
60 |
-
"metadata": "data/add_metadata_extended.csv",
|
61 |
},
|
62 |
"eval": {
|
63 |
"test": "data/eval_test.tar.gz",
|
64 |
-
"metadata":
|
65 |
},
|
66 |
}
|
67 |
|
68 |
EMBEDDING_URLS = {
|
69 |
"dev": {
|
70 |
"embeddings_ast-finetuned-audioset-10-10-0.4593": {
|
71 |
-
"train": "data/MIT_ast-finetuned-audioset-10-10-0
|
72 |
-
"test": "data/MIT_ast-finetuned-audioset-10-10-0
|
73 |
"size": (1, 768),
|
74 |
"dtype": "float32",
|
75 |
},
|
@@ -83,22 +82,16 @@ EMBEDDING_URLS = {
|
|
83 |
},
|
84 |
"add": {
|
85 |
"embeddings_ast-finetuned-audioset-10-10-0.4593": {
|
86 |
-
"train": "",
|
87 |
-
"
|
88 |
-
|
89 |
-
"embeddings_dcase2023_task2_baseline_ae": {
|
90 |
-
"train": "",
|
91 |
-
"test": "",
|
92 |
},
|
93 |
},
|
94 |
"eval": {
|
95 |
"embeddings_ast-finetuned-audioset-10-10-0.4593": {
|
96 |
-
"
|
97 |
-
"
|
98 |
-
|
99 |
-
"embeddings_dcase2023_task2_baseline_ae": {
|
100 |
-
"train": "",
|
101 |
-
"test": "",
|
102 |
},
|
103 |
},
|
104 |
}
|
@@ -108,22 +101,22 @@ STATS = {
|
|
108 |
"configs": {
|
109 |
'dev': {
|
110 |
'date': "Mar 1, 2023",
|
111 |
-
'version': "
|
112 |
-
'homepage': "https://zenodo.org/record/
|
113 |
"splits": ["train", "test"],
|
114 |
},
|
115 |
-
|
116 |
-
|
117 |
-
|
118 |
-
|
119 |
-
|
120 |
-
|
121 |
-
|
122 |
-
|
123 |
-
|
124 |
-
|
125 |
-
|
126 |
-
|
127 |
}
|
128 |
}
|
129 |
|
@@ -381,10 +374,6 @@ class DCASE2023Task2Dataset(datasets.GeneratorBasedBuilder):
|
|
381 |
"d2v": datasets.Value("string"),
|
382 |
"d3p": datasets.Value("string"),
|
383 |
"d3v": datasets.Value("string"),
|
384 |
-
"anomaly_score_dcase2023_task2_baseline_ae": datasets.Value("float32"),
|
385 |
-
"prediction_dcase2023_task2_baseline_ae": datasets.Value("int64"),
|
386 |
-
"prediction_correct_dcase2023_task2_baseline_ae": datasets.Value("int64"),
|
387 |
-
"anomaly_score_embedding_lof": datasets.Value("float32"),
|
388 |
}
|
389 |
if self.config.embeddings_urls is not None:
|
390 |
features.update({
|
@@ -436,7 +425,7 @@ class DCASE2023Task2Dataset(datasets.GeneratorBasedBuilder):
|
|
436 |
"local_extracted_archive": local_extracted_archive[split],
|
437 |
"audio_files": dl_manager.iter_archive(audio_path[split]),
|
438 |
"embeddings": embeddings[split],
|
439 |
-
"metadata_file": dl_manager.download_and_extract(self.config.data_urls["metadata"]),
|
440 |
"is_streaming": dl_manager.is_streaming,
|
441 |
},
|
442 |
) for split in split_type if split in self.config.splits
|
@@ -452,20 +441,22 @@ class DCASE2023Task2Dataset(datasets.GeneratorBasedBuilder):
|
|
452 |
is_streaming: Optional[bool],
|
453 |
):
|
454 |
"""Yields examples."""
|
455 |
-
|
|
|
456 |
data_fields = list(self._info().features.keys())
|
457 |
|
458 |
id_ = 0
|
459 |
for path, f in audio_files:
|
460 |
lookup = Path(path).parent.name + "/" + Path(path).name
|
461 |
-
if lookup in metadata["path"].values:
|
462 |
path = os.path.join(local_extracted_archive, path) if local_extracted_archive else path
|
463 |
if is_streaming:
|
464 |
audio = {"path": path, "bytes": f.read()}
|
465 |
else:
|
466 |
audio = {"path": path, "bytes": None}
|
467 |
result = {field: None for field in data_fields}
|
468 |
-
|
|
|
469 |
for emb_key in embeddings.keys():
|
470 |
result[emb_key] = np.asarray(embeddings[emb_key][lookup]).squeeze().tolist()
|
471 |
result["path"] = path
|
|
|
13 |
|
14 |
|
15 |
_CITATION = """\
|
16 |
+
@dataset{kota_dohi_2023_7882613,
|
17 |
author = {Kota Dohi and
|
18 |
+
Keisuke Imoto and
|
19 |
+
Noboru Harada and
|
20 |
+
Daisuke Niizumi and
|
21 |
+
Yuma Koizumi and
|
22 |
+
Tomoya Nishida and
|
23 |
+
Harsh Purohit and
|
24 |
+
Takashi Endo and
|
25 |
+
Yohei Kawaguchi},
|
26 |
title = {DCASE 2023 Challenge Task 2 Development Dataset},
|
27 |
month = mar,
|
28 |
year = 2023,
|
29 |
publisher = {Zenodo},
|
30 |
+
version = {3.0},
|
31 |
+
doi = {10.5281/zenodo.7882613},
|
32 |
+
url = {https://doi.org/10.5281/zenodo.7882613}
|
33 |
}
|
34 |
"""
|
35 |
_LICENSE = "Creative Commons Attribution 4.0 International Public License"
|
|
|
37 |
_METADATA_REG = r"attributes_\d+.csv"
|
38 |
|
39 |
_NUM_TARGETS = 2
|
40 |
+
_NUM_CLASSES = 14
|
41 |
|
42 |
_TARGET_NAMES = ["normal", "anomaly"]
|
43 |
+
_CLASS_NAMES = ["gearbox", "fan", "bearing", "slider", "ToyCar", "ToyTrain", "valve", "bandsaw", "grinder", "shaker", "ToyDrone", "ToyNscale", "ToyTank", "Vacuum"]
|
44 |
|
45 |
_HOMEPAGE = {
|
46 |
+
"dev": "https://zenodo.org/record/7690157",
|
47 |
"add": "",
|
48 |
"eval": "",
|
49 |
}
|
|
|
52 |
"dev": {
|
53 |
"train": "data/dev_train.tar.gz",
|
54 |
"test": "data/dev_test.tar.gz",
|
55 |
+
"metadata": "data/dev_metadata.csv",
|
56 |
},
|
57 |
"add": {
|
58 |
"train": "data/add_train.tar.gz",
|
59 |
+
"metadata": "data/add_metadata.csv",
|
|
|
60 |
},
|
61 |
"eval": {
|
62 |
"test": "data/eval_test.tar.gz",
|
63 |
+
"metadata": None,
|
64 |
},
|
65 |
}
|
66 |
|
67 |
EMBEDDING_URLS = {
|
68 |
"dev": {
|
69 |
"embeddings_ast-finetuned-audioset-10-10-0.4593": {
|
70 |
+
"train": "data/MIT_ast-finetuned-audioset-10-10-0.4593-embeddings_dev_train.npz",
|
71 |
+
"test": "data/MIT_ast-finetuned-audioset-10-10-0.4593-embeddings_dev_test.npz",
|
72 |
"size": (1, 768),
|
73 |
"dtype": "float32",
|
74 |
},
|
|
|
82 |
},
|
83 |
"add": {
|
84 |
"embeddings_ast-finetuned-audioset-10-10-0.4593": {
|
85 |
+
"train": "data/MIT_ast-finetuned-audioset-10-10-0.4593-embeddings_add_train.npz",
|
86 |
+
"size": (1, 768),
|
87 |
+
"dtype": "float32",
|
|
|
|
|
|
|
88 |
},
|
89 |
},
|
90 |
"eval": {
|
91 |
"embeddings_ast-finetuned-audioset-10-10-0.4593": {
|
92 |
+
"test": "data/MIT_ast-finetuned-audioset-10-10-0.4593-embeddings_eval_test.npz",
|
93 |
+
"size": (1, 768),
|
94 |
+
"dtype": "float32",
|
|
|
|
|
|
|
95 |
},
|
96 |
},
|
97 |
}
|
|
|
101 |
"configs": {
|
102 |
'dev': {
|
103 |
'date': "Mar 1, 2023",
|
104 |
+
'version': "3.0.0",
|
105 |
+
'homepage': "https://zenodo.org/record/7882613",
|
106 |
"splits": ["train", "test"],
|
107 |
},
|
108 |
+
'add': {
|
109 |
+
'date': "Apr 15, 2023",
|
110 |
+
'version': "1.0.0",
|
111 |
+
'homepage': "https://zenodo.org/record/7830345",
|
112 |
+
"splits": ["train"],
|
113 |
+
},
|
114 |
+
'eval': {
|
115 |
+
'date': "May 1, 2023",
|
116 |
+
'version': "1.0.0",
|
117 |
+
'homepage': "https://zenodo.org/record/7860847",
|
118 |
+
"splits": ["test"],
|
119 |
+
},
|
120 |
}
|
121 |
}
|
122 |
|
|
|
374 |
"d2v": datasets.Value("string"),
|
375 |
"d3p": datasets.Value("string"),
|
376 |
"d3v": datasets.Value("string"),
|
|
|
|
|
|
|
|
|
377 |
}
|
378 |
if self.config.embeddings_urls is not None:
|
379 |
features.update({
|
|
|
425 |
"local_extracted_archive": local_extracted_archive[split],
|
426 |
"audio_files": dl_manager.iter_archive(audio_path[split]),
|
427 |
"embeddings": embeddings[split],
|
428 |
+
"metadata_file": dl_manager.download_and_extract(self.config.data_urls["metadata"]) if self.config.data_urls["metadata"] is not None else None,
|
429 |
"is_streaming": dl_manager.is_streaming,
|
430 |
},
|
431 |
) for split in split_type if split in self.config.splits
|
|
|
441 |
is_streaming: Optional[bool],
|
442 |
):
|
443 |
"""Yields examples."""
|
444 |
+
if metadata_file is not None:
|
445 |
+
metadata = pd.read_csv(metadata_file)
|
446 |
data_fields = list(self._info().features.keys())
|
447 |
|
448 |
id_ = 0
|
449 |
for path, f in audio_files:
|
450 |
lookup = Path(path).parent.name + "/" + Path(path).name
|
451 |
+
if metadata_file is None or lookup in metadata["path"].values:
|
452 |
path = os.path.join(local_extracted_archive, path) if local_extracted_archive else path
|
453 |
if is_streaming:
|
454 |
audio = {"path": path, "bytes": f.read()}
|
455 |
else:
|
456 |
audio = {"path": path, "bytes": None}
|
457 |
result = {field: None for field in data_fields}
|
458 |
+
if metadata_file is not None:
|
459 |
+
result.update(metadata[metadata["path"] == lookup].T.squeeze().to_dict())
|
460 |
for emb_key in embeddings.keys():
|
461 |
result[emb_key] = np.asarray(embeddings[emb_key][lookup]).squeeze().tolist()
|
462 |
result["path"] = path
|