Hezar: Upload tokenizer and config
Browse files
preprocessor/tokenizer_config.yaml
ADDED
@@ -0,0 +1,37 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
name: bpe_tokenizer
|
2 |
+
config_type: preprocessor
|
3 |
+
max_length: 512
|
4 |
+
truncation_strategy: longest_first
|
5 |
+
truncation_direction: right
|
6 |
+
stride: 0
|
7 |
+
padding_strategy: longest
|
8 |
+
padding_direction: right
|
9 |
+
pad_to_multiple_of: 0
|
10 |
+
pad_token_type_id: 0
|
11 |
+
bos_token: <s>
|
12 |
+
eos_token: </s>
|
13 |
+
unk_token: <unk>
|
14 |
+
sep_token: <sep>
|
15 |
+
pad_token: <pad>
|
16 |
+
cls_token: <cls>
|
17 |
+
mask_token: <mask>
|
18 |
+
special_tokens:
|
19 |
+
- <s>
|
20 |
+
- <pad>
|
21 |
+
- </s>
|
22 |
+
- <unk>
|
23 |
+
- <mask>
|
24 |
+
- <|endoftext|>
|
25 |
+
- <|startoftext|>
|
26 |
+
- <nl>
|
27 |
+
- <hs>
|
28 |
+
- <sep>
|
29 |
+
- <cls>
|
30 |
+
continuing_subword_prefix: ''
|
31 |
+
end_of_word_suffix: ''
|
32 |
+
fuse_unk: false
|
33 |
+
vocab_size: 42000
|
34 |
+
min_frequency: 2
|
35 |
+
limit_alphabet: 1000
|
36 |
+
initial_alphabet: []
|
37 |
+
show_progress: true
|