Upload tokenizer

added_tokens.json

@@ -0,0 +1,5 @@
+{
+  "<|im_end|>": 100279,
+  "<|im_start|>": 100278,
+  "<|pad|>": 100277
+}

special_tokens_map.json

@@ -3,8 +3,26 @@
     "<|im_start|>",
     "<|im_end|>"
   ],
-  "bos_token": "<|endoftext|>",
-  "eos_token": "<|endoftext|>",
+  "bos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
   "pad_token": "<|pad|>",
-  "unk_token": "<|endoftext|>"
+  "unk_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
 }

tiktoken.py

@@ -1,11 +1,14 @@
 # Copyright 2022 MosaicML LLM Foundry authors
 # SPDX-License-Identifier: Apache-2.0
 
+import warnings
 from typing import Any, Dict, List, Optional, Tuple, Union
 
 import torch
 from transformers import PreTrainedTokenizer
 
+DEFAULT_SYSTEM_PROMPT = """You are a helpful, respectful and honest assistant. Always answer as helpfully as possible."""
+
 
 class TiktokenTokenizerWrapper(PreTrainedTokenizer):
     """A thin wrapper around tiktoken to make it compatible with Hugging Face.
@@ -22,11 +25,12 @@ class TiktokenTokenizerWrapper(PreTrainedTokenizer):
                  encoding_name: Optional[str] = None,
                  add_bos_token: bool = False,
                  add_eos_token: bool = False,
+                 use_default_system_prompt: bool = False,
                  unk_token: Optional[str] = '<|endoftext|>',
                  eos_token: Optional[str] = '<|endoftext|>',
                  bos_token: Optional[str] = '<|endoftext|>',
                  pad_token: Optional[str] = None,
-                 **kwargs: Dict[str, Any]):
+                 **kwargs: Any):
         """Constructor creates a tiktoken tokenizer to use as the underlying.
 
         tokenizer.
@@ -38,6 +42,7 @@ class TiktokenTokenizerWrapper(PreTrainedTokenizer):
                 Either model_name or encoding_name must be set, but not both.
             add_bos_token (bool, optional): Whether to add bos tokens. Defaults to False.
             add_eos_token (bool, optional): Whether to add eos tokens. Defaults to False.
+            use_default_system_prompt (bool, optional): Use the default system prompt or not. Defaults to False.
             unk_token (Optional[str], optional): The unk token. Defaults to '<|endoftext|>'.
             eos_token (Optional[str], optional): The eos token. Defaults to '<|endoftext|>'.
             bos_token (Optional[str], optional): The bos token. Defaults to '<|endoftext|>'.
@@ -49,6 +54,23 @@ class TiktokenTokenizerWrapper(PreTrainedTokenizer):
             raise ImportError(
                 'You need to install tiktoken to use TiktokenTokenizerWrapper.')
 
+        # Workaround to make tiktokenizer picklable.
+        # https://github.com/huggingface/datasets/issues/5536#issuecomment-1682309347
+        # There is an open PR from HF to add this to tiktoken: https://github.com/openai/tiktoken/pull/181
+        import copyreg
+        import functools
+
+        from tiktoken import Encoding  # type: ignore (thirdParty)
+
+        def pickle_Encoding(enc: Encoding):
+            return (functools.partial(Encoding,
+                                      enc.name,
+                                      pat_str=enc._pat_str,
+                                      mergeable_ranks=enc._mergeable_ranks,
+                                      special_tokens=enc._special_tokens), ())
+
+        copyreg.pickle(Encoding, pickle_Encoding)
+
         if model_name is not None and encoding_name is not None:
             raise ValueError(
                 'You need to specify either model_name or encoding_name, not both.'
@@ -69,11 +91,13 @@ class TiktokenTokenizerWrapper(PreTrainedTokenizer):
 
         self.add_bos_token = add_bos_token
         self.add_eos_token = add_eos_token
+        self.use_default_system_prompt = use_default_system_prompt
 
         super().__init__(model_name=model_name,
                          encoding_name=encoding_name,
                          add_bos_token=add_bos_token,
                          add_eos_token=add_eos_token,
+                         use_default_system_prompt=use_default_system_prompt,
                          unk_token=unk_token,
                          eos_token=eos_token,
                          bos_token=bos_token,
@@ -89,8 +113,39 @@ class TiktokenTokenizerWrapper(PreTrainedTokenizer):
     def is_fast(self) -> bool:
         return False
 
+    @property
+    def default_chat_template(self):
+        """Chat ML Template for User/Assistant.
+
+        Pinning default Chat ML template in case defaults change.
+        """
+        template = (
+            "{% set system_message = '' %}"
+            '{% if USE_DEFAULT_PROMPT == true %}'
+            "{{'<|im_start|>system\n' + 'DEFAULT_SYSTEM_PROMPT'}}"
+            '{% endif %}'
+            '{% for message in messages %}'
+            "{{'<|im_start|>' + message['role'] + '\n' + message['content'] + '<|im_end|>' + '\n'}}"
+            '{% endfor %}')
+        template = template.replace(
+            'USE_DEFAULT_PROMPT',
+            'true' if self.use_default_system_prompt else 'false')
+        template = template.replace('DEFAULT_SYSTEM_PROMPT',
+                                    DEFAULT_SYSTEM_PROMPT)
+        return template
+
     def get_vocab(self) -> Dict[str, int]:
-        """Returns vocab as a dict."""
+        """Returns vocab as a dict.
+
+        Note: This function does not work properly due to difference in assumptions between tiktoken and Hugging Face tokenizers.
+        Most uses do not need to use get_vocab, so this is not a priority to fix.
+        """
+        warnings.warn(
+            'get_vocab does not work properly with TiktokenTokenizerWrapper. Please do not rely on it being perfectly correct.'
+            +
+            ' It will be called once init just to get the size of the vocab inside the base class.'
+        )
+
         vocab = {}
         for i in range(self.vocab_size):
             try:
@@ -101,6 +156,24 @@ class TiktokenTokenizerWrapper(PreTrainedTokenizer):
             except KeyError:
                 pass
 
+        # As far as I can tell, we don't require get_vocab to completely work,
+        # but when using additional_special_tokens, Hugging Face determines the next
+        # token index to add with len(self.get_vocab()) so we need the _size_ of this dictionary to be correct.
+        extra_id_index = 0
+        candidate_extra_id = f'<extra_id_{extra_id_index}>'
+        indices_to_fill_in = {i for i in range(self.vocab_size)} - set(
+            vocab.values())
+
+        # Add enough indices to make get_vocab() the right length
+        for index_to_add in indices_to_fill_in:
+            # Make sure we don't overwrite a token that already exists
+            while candidate_extra_id in vocab:
+                extra_id_index += 1
+                candidate_extra_id = f'<extra_id_{extra_id_index}>'
+
+            # Get an index to add and add the item
+            vocab[candidate_extra_id] = index_to_add
+
         return vocab
 
     def _tokenize(self, text: str) -> List[int]:
@@ -155,7 +228,7 @@ class TiktokenTokenizerWrapper(PreTrainedTokenizer):
         """
         if isinstance(ids, int):
             if ids in self.added_tokens_decoder:
-                return self.added_tokens_decoder[ids]
+                return str(self.added_tokens_decoder[ids])
 
             return self._convert_id_to_token(ids)
 
@@ -171,7 +244,7 @@ class TiktokenTokenizerWrapper(PreTrainedTokenizer):
             if index in self.added_tokens_decoder:
                 tokens.append(self.encoding.decode(current_stream))
                 current_stream = []
-                tokens.append(self.added_tokens_decoder[index])
+                tokens.append(str(self.added_tokens_decoder[index]))
             else:
                 current_stream.append(index)
 

tokenizer_config.json

@@ -2,6 +2,72 @@
   "add_bos_token": false,
   "add_eos_token": false,
   "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "100257": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "100258": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "100259": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "100260": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "100276": {
+      "content": "<|endofprompt|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "100277": {
+      "content": "<|pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "100278": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "100279": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
   "additional_special_tokens": [
     "<|im_start|>",
     "<|im_end|>"
@@ -20,5 +86,6 @@
   "model_name": "gpt-4",
   "pad_token": "<|pad|>",
   "tokenizer_class": "TiktokenTokenizerWrapper",
-  "unk_token": "<|endoftext|>"
+  "unk_token": "<|endoftext|>",
+  "use_default_system_prompt": false
 }