fix dataset types when using random dataset

.gitignore

@@ -145,3 +145,4 @@ core.*
 llm_ckpts
 events.*
 memory_trace
+RUN*/

internlm/train/training_internlm.py

@@ -226,6 +226,7 @@ def get_train_data_loader(
     dataset_types = list(get_dataset_type_ids_map(train_folder).keys())
 
     if not train_folder:
+        dataset_types = ["en", "cn", "code"]
         train_ds = RandomDataset(num_samples=1000000, max_len=data_cfg.seq_len)
         if data_cfg.pack_sample_into_one:
             train_ds = PackedDatasetWithoutCuSeqlen(