Update time_series_library.py

time_series_library.py

@@ -22,37 +22,58 @@ class TimeSeriesLibrary(GeneratorBasedBuilder):
         return self.dataset_info_from_configs(features=Features(cols))
 
 
+    from datasets import Split, SplitGenerator
+    
     def _split_generators(self, dl_manager):
         dfiles = self.config.data_files
         files = list(dfiles) if isinstance(dfiles, (list, tuple)) else [dfiles]
     
-        def find_and_pop(keys):
-            keys = [k.lower() for k in keys]
-            for i, f in enumerate(files):
-                name = os.path.basename(str(f)).lower()
-                if any(k in name for k in keys):
-                    return files.pop(i)
-            return None
+        gens = []
     
-        train_f = find_and_pop(["train"])
-        val_f   = find_and_pop(["val", "valid", "validation"])
-        test_f  = find_and_pop(["test"])
-        label_f = find_and_pop(["test_label", "label", "labels"])
+        def split_name_for(path, used):
+            """按文件名自动判别 split；若重复则加后缀避免合并"""
+            name = os.path.basename(str(path)).lower()
+            if "train" in name:
+                base = "train"
+            elif any(k in name for k in ["val", "valid", "validation"]):
+                base = "validation"
+            elif "test_label" in name or "label" in name:
+                base = "test_label"
+            elif "test" in name:
+                base = "test"
+            else:
+                # 没有关键字就用文件名（去掉拓展名）
+                base = os.path.splitext(os.path.basename(str(path)))[0]
     
-        gens = []
-        if train_f:
-            gens.append(SplitGenerator(name=Split.TRAIN, gen_kwargs={"filepath": train_f}))
-        if val_f:
-            gens.append(SplitGenerator(name=Split.VALIDATION, gen_kwargs={"filepath": val_f}))
-        if test_f:
-            gens.append(SplitGenerator(name=Split.TEST, gen_kwargs={"filepath": test_f}))
-        if label_f:
-            gens.append(SplitGenerator(name="test_label", gen_kwargs={"filepath": label_f}))
+            # 避免同名冲突：train/train_1/train_2...
+            s = base
+            idx = 1
+            while s in used:
+                s = f"{base}_{idx}"
+                idx += 1
+            used.add(s)
+            return s
     
+        used_names = set()
+        for f in files:
+            sname = split_name_for(f, used_names)
+            # 标准 split 用内置常量；自定义用 Split(sname)
+            if sname == "train":
+                gens.append(SplitGenerator(name=Split.TRAIN,       gen_kwargs={"filepath": f}))
+            elif sname == "validation":
+                gens.append(SplitGenerator(name=Split.VALIDATION,  gen_kwargs={"filepath": f}))
+            elif sname == "test":
+                gens.append(SplitGenerator(name=Split.TEST,        gen_kwargs={"filepath": f}))
+            else:
+                gens.append(SplitGenerator(name=Split(sname),      gen_kwargs={"filepath": f}))
+    
+        # 如果列表为空（极端情况），兜底把所有文件各自做一个 split
         if not gens:
-            gens = [SplitGenerator(name=Split.TRAIN, gen_kwargs={"filepath": files})]
+            for i, f in enumerate(files):
+                gens.append(SplitGenerator(name=Split(f"file_{i}"), gen_kwargs={"filepath": f}))
     
         return gens
+
     
 
     def _generate_examples(self, filepath, row_slice=None):