fix: Update tokenizer_config.json with correct 3B ORPO byte-fallback tokenizer (vocab=64256)
Browse files- tokenizer_config.json +7 -4
tokenizer_config.json
CHANGED
|
@@ -1,11 +1,14 @@
|
|
| 1 |
{
|
| 2 |
-
"
|
| 3 |
-
"tokenizer_class": "PreTrainedTokenizerFast",
|
| 4 |
"bos_token": "<s>",
|
|
|
|
| 5 |
"eos_token": "</s>",
|
| 6 |
-
"
|
|
|
|
|
|
|
| 7 |
"pad_token": "<pad>",
|
| 8 |
-
"
|
|
|
|
| 9 |
"added_tokens_decoder": {
|
| 10 |
"64000": {
|
| 11 |
"content": "<0x00>",
|
|
|
|
| 1 |
{
|
| 2 |
+
"backend": "tokenizers",
|
|
|
|
| 3 |
"bos_token": "<s>",
|
| 4 |
+
"clean_up_tokenization_spaces": false,
|
| 5 |
"eos_token": "</s>",
|
| 6 |
+
"is_local": true,
|
| 7 |
+
"model_max_length": 1000000000000000019884624838656,
|
| 8 |
+
"model_type": "llama",
|
| 9 |
"pad_token": "<pad>",
|
| 10 |
+
"tokenizer_class": "TokenizersBackend",
|
| 11 |
+
"unk_token": "<unk>",
|
| 12 |
"added_tokens_decoder": {
|
| 13 |
"64000": {
|
| 14 |
"content": "<0x00>",
|