basujindal
commited on
Commit
•
b7feb49
1
Parent(s):
56d6481
Update tokenizer_config.json
Browse filesConfiguring bos, eos, pad and unk tokens helps load the tokenizer successfully, otherwise getting a recursion error "RecursionError: maximum recursion depth exceeded while getting the str of an object." As done here: https://huggingface.co/medalpaca/medalpaca-13b/discussions/7
- tokenizer_config.json +4 -3
tokenizer_config.json
CHANGED
@@ -1,9 +1,10 @@
|
|
1 |
{
|
2 |
-
"bos_token": "",
|
3 |
-
"eos_token": "",
|
4 |
"model_max_length": 512,
|
5 |
"padding_side": "right",
|
6 |
"special_tokens_map_file": "../llama-7b-hf/special_tokens_map.json",
|
7 |
"tokenizer_class": "LlamaTokenizer",
|
8 |
-
"unk_token": ""
|
|
|
9 |
}
|
|
|
1 |
{
|
2 |
+
"bos_token": "</s>",
|
3 |
+
"eos_token": "</s>",
|
4 |
"model_max_length": 512,
|
5 |
"padding_side": "right",
|
6 |
"special_tokens_map_file": "../llama-7b-hf/special_tokens_map.json",
|
7 |
"tokenizer_class": "LlamaTokenizer",
|
8 |
+
"unk_token": "</s>",
|
9 |
+
"pad_token": "[PAD]"
|
10 |
}
|