Make gptxdata optional dependency

Also sort imports

megatron/tokenizer/tokenizer.py

@@ -4,10 +4,29 @@
 
 from abc import ABC
 from abc import abstractmethod
-from gptxdata.tokenization import HFTokenizer, SPTokenizer, PretrainedHFTokenizer
+
+try:
+    from gptxdata.tokenization import (
+        HFTokenizer,
+        PretrainedHFTokenizer,
+        SPTokenizer,
+    )
+except ImportError:
+    HFTokenizer = None
+    PretrainedHFTokenizer = None
+    SPTokenizer = None
+
 from .bert_tokenization import FullTokenizer as FullBertTokenizer
 from .gpt2_tokenization import GPT2Tokenizer
 
+
+def _assert_gptx_tokenizer_available(tokenizer_name, tokenizer_cls):
+    assert tokenizer_cls is not None, (
+        f'Please install `gptxdata` to use {tokenizer_name}, e.g., with '
+        f'`pip install git+https://github.com/OpenGPTX/opengptx_data.git`'
+    )
+
+
 def build_tokenizer(args):
     """Initialize tokenizer."""
     if args.rank == 0:
@@ -39,10 +58,14 @@ def build_tokenizer(args):
         assert args.vocab_size is not None
         tokenizer = _NullTokenizer(args.vocab_size)
     elif args.tokenizer_type == "OpenGPTX-HFTokenizer":
+        _assert_gptx_tokenizer_available(args.tokenizer_type, HFTokenizer)
         tokenizer = HFTokenizer.instantiate_from_file_or_name(model_file_or_name=args.tokenizer_model)
     elif args.tokenizer_type == "OpenGPTX-PretrainedHFTokenizer":
+        _assert_gptx_tokenizer_available(
+            args.tokenizer_type, PretrainedHFTokenizer)
         tokenizer = PretrainedHFTokenizer.instantiate_from_file_or_name(model_file_or_name=args.tokenizer_model)
     elif args.tokenizer_type == "OpenGPTX-SPTokenizer":
+        _assert_gptx_tokenizer_available(args.tokenizer_type, SPTokenizer)
         tokenizer = SPTokenizer.instantiate_from_file_or_name(model_file_or_name=args.tokenizer_model)
     else:
         raise NotImplementedError('{} tokenizer is not '