HW3 env, data and training (eval has problems)

HW3/data/sample_submission.jsonl

@@ -0,0 +1,10 @@
+{"title": "Anker新款真無線藍牙耳機Liberty Air 2 Pro 引進台灣市場", "id": "21710"}
+{"title": "藍染、客家美食、舊山線自行車 「苗栗一日遊」超人氣美食美景", "id": "21711"}
+{"title": "華碩打造對應軍規防護與2 in 1設計的15.6吋Chromebook", "id": "21712"}
+{"title": "產業發展變革 台灣的優勢與機會", "id": "21713"}
+{"title": "全球Windows 7裝置粗估至少還有1億台以上 市佔率穩穩卡在20％", "id": "21714"}
+{"title": "強勢台幣理財攻略", "id": "21715"}
+{"title": "「不需治療，只需到台灣！」 美國「哈台馬克杯」賣到缺貨", "id": "21716"}
+{"title": "ZenBook Duo 14、ZenBook Pro Duo 15 OLED更新 第二螢幕更方便使用了", "id": "21717"}
+{"title": "不出國更要怒吃雞加酒！「周末炸虎俱樂部」限量開賣", "id": "21718"}
+{"title": "NBA／紐約記者爆料厄文狀況 「不喜歡奈許、跟KD疏離」", "id": "21719"}

HW3/data/split.py

@@ -0,0 +1,14 @@
+import numpy as np
+np.random.seed(1114)
+
+with open("train.jsonl", 'r') as f:
+    data = f.readlines()
+
+all_indices = np.random.permutation(np.arange(len(data)))
+cut = int(len(data) * 0.2)
+with open("train_split.jsonl", 'w') as f:
+    for i in all_indices[cut:].tolist():
+        f.write(data[i])
+with open("valid_split.jsonl", 'w') as f:
+    for i in all_indices[:cut].tolist():
+        f.write(data[i])