Spaces:
Sleeping
Sleeping
ktsn-ud commited on
Commit ·
bd36d3a
1
Parent(s): 5544c96
インポートに関するエラー修正と、vecのダウンロードを追加
Browse files- scripts/0_download_data.py +33 -15
scripts/0_download_data.py
CHANGED
|
@@ -1,10 +1,16 @@
|
|
| 1 |
import os
|
|
|
|
|
|
|
|
|
|
| 2 |
|
| 3 |
from utils.logger import setup_logger
|
| 4 |
from utils.json import get_file_path_from_config
|
| 5 |
|
| 6 |
log = setup_logger(__name__)
|
| 7 |
|
|
|
|
|
|
|
|
|
|
| 8 |
|
| 9 |
def download_embeddings() -> None:
|
| 10 |
"""Hugging Face Datasets から Embeddings をダウンロードする"""
|
|
@@ -17,22 +23,34 @@ def download_embeddings() -> None:
|
|
| 17 |
log.info("Downloading embeddings from Hugging Face Datasets...")
|
| 18 |
|
| 19 |
# ダウンロード
|
| 20 |
-
|
| 21 |
-
|
| 22 |
-
|
| 23 |
-
|
| 24 |
-
|
| 25 |
-
|
| 26 |
-
|
| 27 |
-
|
| 28 |
-
|
| 29 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 30 |
return
|
| 31 |
|
| 32 |
|
| 33 |
if __name__ == "__main__":
|
| 34 |
-
|
| 35 |
-
if os.path.exists(fast_text_bin_path):
|
| 36 |
-
log.info("Embeddings already exist. Skipping download.")
|
| 37 |
-
else:
|
| 38 |
-
download_embeddings()
|
|
|
|
| 1 |
import os
|
| 2 |
+
import sys
|
| 3 |
+
|
| 4 |
+
sys.path.append(os.path.abspath(os.path.join(os.path.dirname(__file__), "..")))
|
| 5 |
|
| 6 |
from utils.logger import setup_logger
|
| 7 |
from utils.json import get_file_path_from_config
|
| 8 |
|
| 9 |
log = setup_logger(__name__)
|
| 10 |
|
| 11 |
+
fast_text_bin_path = get_file_path_from_config("embeddings.fasttext_bin")
|
| 12 |
+
fast_text_vec_path = get_file_path_from_config("embeddings.fasttext_vec")
|
| 13 |
+
|
| 14 |
|
| 15 |
def download_embeddings() -> None:
|
| 16 |
"""Hugging Face Datasets から Embeddings をダウンロードする"""
|
|
|
|
| 23 |
log.info("Downloading embeddings from Hugging Face Datasets...")
|
| 24 |
|
| 25 |
# ダウンロード
|
| 26 |
+
if os.path.exists(fast_text_bin_path):
|
| 27 |
+
log.info("Embeddings binary already exists. Skipping download.")
|
| 28 |
+
else:
|
| 29 |
+
path = hf_hub_download(
|
| 30 |
+
repo_id=os.getenv("HF_EMBEDDINGS_REPO_ID"),
|
| 31 |
+
repo_type="dataset",
|
| 32 |
+
filename="cc.ja.300.bin",
|
| 33 |
+
local_dir=target_dir,
|
| 34 |
+
local_dir_use_symlinks=False,
|
| 35 |
+
token=os.getenv("HF_TOKEN"),
|
| 36 |
+
)
|
| 37 |
+
log.info(f"Embeddings downloaded: {path}")
|
| 38 |
+
|
| 39 |
+
if os.path.exists(fast_text_vec_path):
|
| 40 |
+
log.info("Embeddings vector already exists. Skipping download.")
|
| 41 |
+
else:
|
| 42 |
+
path = hf_hub_download(
|
| 43 |
+
repo_id=os.getenv("HF_EMBEDDINGS_REPO_ID"),
|
| 44 |
+
repo_type="dataset",
|
| 45 |
+
filename="cc.ja.300.vec",
|
| 46 |
+
local_dir=target_dir,
|
| 47 |
+
local_dir_use_symlinks=False,
|
| 48 |
+
token=os.getenv("HF_TOKEN"),
|
| 49 |
+
)
|
| 50 |
+
log.info(f"Embeddings downloaded: {path}")
|
| 51 |
+
|
| 52 |
return
|
| 53 |
|
| 54 |
|
| 55 |
if __name__ == "__main__":
|
| 56 |
+
download_embeddings()
|
|
|
|
|
|
|
|
|
|
|
|