2021/9/25 nezha-legal-fgm1.0-lsr0.1_f82.71_p76.04_r90.66

main.py

@@ -48,16 +48,25 @@ def main():
     # n_splits = 5
     # seed=42
     # --------------------------
-    version = "nezha-fgm1.0-lsr0.1"
+    # version = "nezha-fgm1.0-lsr0.1"
+    # model_type = "nezha_span"
+    # dataset_name = "cail_ner"
+    # n_splits = 5
+    # seed=42
+    # --------------------------
+    version = "nezha-legal-fgm1.0-lsr0.1"
     model_type = "nezha_span"
     dataset_name = "cail_ner"
     n_splits = 5
     seed=42
 
     test_examples = None
     test_batches = None
-    for k in range(n_splits):
-        model_path = f"./output/ner-{dataset_name}-{model_type}-{version}-fold{k}-{seed}/"
+    # for k in range(n_splits):
+    #     model_path = f"./output/ner-{dataset_name}-{model_type}-{version}-fold{k}-{seed}/"
+    model_paths = [f"./output/ner-{dataset_name}-{model_type}-{version}-fold{k}-{seed}/" for k in range(n_splits)]
+    for k in range(len(model_paths)):
+        model_path = model_paths[k]
         # 生成测试运行参数
         json_file = os.path.join(model_path, "training_args.json")
         parser = NerArgumentParser()

main_local.py

@@ -54,16 +54,25 @@ def main():
     # n_splits = 5
     # seed=42
     # --------------------------
-    version = "nezha-fgm1.0-lsr0.1"
+    # version = "nezha-fgm1.0-lsr0.1"
+    # model_type = "nezha_span"
+    # dataset_name = "cail_ner"
+    # n_splits = 5
+    # seed=42
+    # --------------------------
+    version = "nezha-legal-fgm1.0-lsr0.1"
     model_type = "nezha_span"
     dataset_name = "cail_ner"
     n_splits = 5
     seed=42
 
     test_examples = []
     test_batches = []
-    for k in range(n_splits):
-        model_path = f"./output/ner-{dataset_name}-{model_type}-{version}-fold{k}-{seed}/"
+    # for k in range(n_splits):
+    #     model_path = f"./output/ner-{dataset_name}-{model_type}-{version}-fold{k}-{seed}/"
+    model_paths = [f"./output/ner-{dataset_name}-{model_type}-{version}-fold{k}-{seed}/" for k in range(n_splits)]
+    for k in range(len(model_paths)):
+        model_path = model_paths[k]
         # 生成测试运行参数
         json_file = os.path.join(model_path, "training_args.json")
         parser = NerArgumentParser()

scripts/clear_before_submit.sh

@@ -1,5 +1,5 @@
 rm -rf args/pred*
-rm -rf tmp/*
+# rm -rf tmp/*
 rm -rf output/*/test*
 rm -rf output/*/checkpoint-999999/
 

scripts/run_span.sh

@@ -475,6 +475,62 @@ done
 # 组织机构
 # {'p': 0.8513513513513513, 'r': 0.8598014888337469, 'f': 0.8555555555555555}
 
+# Further-pretrain LSR
+for k in 0 1 2 3 4
+do
+python run_span.py \
+    --version=nezha-legal-fgm1.0-lsr0.1-fold${k} \
+    --data_dir=./data/ner-ctx0-5fold-seed42/ \
+    --train_file=train.${k}.json \
+    --dev_file=dev.${k}.json \
+    --test_file=dev.${k}.json \
+    --model_type=nezha_span \
+    --model_name_or_path=/home/louishsu/NewDisk/Code/CAIL2021/nezha-legal-cn-base-wwm/ \
+    --do_train \
+    --overwrite_output_dir \
+    --evaluate_during_training \
+    --evaluate_each_epoch \
+    --save_best_checkpoints \
+    --max_span_length=40 \
+    --width_embedding_dim=128 \
+    --train_max_seq_length=512 \
+    --eval_max_seq_length=512 \
+    --do_lower_case \
+    --per_gpu_train_batch_size=8 \
+    --per_gpu_eval_batch_size=16 \
+    --gradient_accumulation_steps=2 \
+    --learning_rate=5e-5 \
+    --other_learning_rate=1e-3 \
+    --num_train_epochs=8.0 \
+    --warmup_proportion=0.1 \
+    --do_fgm --fgm_epsilon=1.0 \
+    --loss_type=lsr --label_smooth_eps=0.1 \
+    --seed=42
+done
+# main_local
+# avg
+# {'p': 0.9032722314800787, 'r': 0.8945650950827051, 'f': 0.898897578441534}
+# 犯罪嫌疑人
+# {'p': 0.9654852190063458, 'r': 0.965186445922946, 'f': 0.9653358093469513}
+# 受害人
+# {'p': 0.9271829682196853, 'r': 0.9668597168597168, 'f': 0.9466057646873522}
+# 被盗货币
+# {'p': 0.8280590717299579, 'r': 0.8579234972677595, 'f': 0.8427267847557702}
+# 物品价值
+# {'p': 0.9745069745069745, 'r': 0.969377990430622, 'f': 0.9719357159990405}
+# 盗窃获利
+# {'p': 0.8812877263581489, 'r': 0.9106029106029107, 'f': 0.8957055214723928}
+# 被盗物品
+# {'p': 0.8194369732831271, 'r': 0.7905206711641585, 'f': 0.8047191406937841}
+# 作案工具
+# {'p': 0.7972972972972973, 'r': 0.8027210884353742, 'f': 0.7999999999999999}
+# 时间
+# {'p': 0.950575994054255, 'r': 0.9251356238698011, 'f': 0.937683284457478}
+# 地点
+# {'p': 0.8714835652946402, 'r': 0.836792721069093, 'f': 0.8537859007832898}
+# 组织机构
+# {'p': 0.8789407313997478, 'r': 0.8647642679900744, 'f': 0.8717948717948718}
+
 # TODO: 全部数据
 
 # TODO: EMA