New model World run error, is it support new vocab 'rwkv_vocab_v20230424.txt'?

[Helloworld2345567]

When I finish lora training and run 'python chat.py',
Traceback (most recent call last):
File "/root/data1/ym/RWKV-LM/RWKV-LM-LoRA/RWKV-v4neo/chat.py", line 28, in
tokenizer = TOKENIZER(WORD_NAME, UNKNOWN_CHAR=UNKNOWN_CHAR)
File "/root/data1/ym/RWKV-LM/RWKV-LM-LoRA/RWKV-v4neo/src/utils.py", line 22, in init
self.tokenizer = PreTrainedTokenizerFast(tokenizer_file=WORD_NAME[0])
File "/usr/local/lib/python3.10/dist-packages/transformers/tokenization_utils_fast.py", line 111, in init
fast_tokenizer = TokenizerFast.from_file(fast_tokenizer_file)
Exception: invalid type: integer 1, expected struct Tokenizer at line 1 column 1

[Helloworld2345567]

Title: Update TOKENIZER class and load vocabulary from ChatRWKV

Body:

To resolve the issue, make the following changes in the code:

1. Replace the TOKENIZER class in RWKV-LM-LoRA/RWKV-v4neo/src/utils.py with the TOKENIZER class from ChatRWKV/src/utils.py.

2. Update the vocabulary loading code in ChatRWKV/v2/chat.py with the following:

   from rwkv.model import RWKV
   from rwkv.utils import PIPELINE
   
   if 'world/' in args.MODEL_NAME or '-World-' in args.MODEL_NAME:
       pipeline = PIPELINE(model, "rwkv_vocab_v20230424")
       END_OF_TEXT = 0
       END_OF_LINE = 11

3. Replace any instances of tokenizer.tokenizer. method with pipeline. method in the code.

Or you can merge lora weights with original weights. Then you don't catch this issues.