Update README.md
Browse files
README.md
CHANGED
@@ -9,28 +9,14 @@ pipeline_tag: token-classification
|
|
9 |
---
|
10 |
|
11 |
```python
|
12 |
-
from peft import PeftModel,
|
13 |
-
from transformers import
|
14 |
-
import torch
|
15 |
|
16 |
-
|
|
|
|
|
17 |
|
18 |
-
|
19 |
-
load_in_4bit=True,
|
20 |
-
bnb_4bit_quant_type="nf4",
|
21 |
-
bnb_4bit_use_double_quant=True,
|
22 |
-
bnb_4bit_compute_dtype=torch.bfloat16,
|
23 |
-
)
|
24 |
-
|
25 |
-
model = T5ForTokenClassification.from_pretrained(model_id,
|
26 |
-
num_labels=2,
|
27 |
-
torch_dtype=torch.bfloat16,
|
28 |
-
quantization_config=bnb_config,
|
29 |
-
device_map="auto",)
|
30 |
-
model = prepare_model_for_kbit_training(model)
|
31 |
-
model = PeftModel.from_pretrained(model, 'bite-the-byte/byt5-small-deASCIIfy-TR')
|
32 |
-
|
33 |
-
def test_mask(data):
|
34 |
"""
|
35 |
Masks the padded tokens in the input.
|
36 |
Args:
|
@@ -39,21 +25,16 @@ def test_mask(data):
|
|
39 |
dataset (list): List of dictionaries.
|
40 |
"""
|
41 |
|
42 |
-
|
43 |
-
for sample in data:
|
44 |
-
new_sample = dict()
|
45 |
|
46 |
-
|
47 |
-
|
48 |
-
|
49 |
-
|
50 |
-
|
51 |
-
|
52 |
-
|
53 |
-
|
54 |
-
dataset.append(new_sample)
|
55 |
-
|
56 |
-
return dataset
|
57 |
|
58 |
def rewrite(model, data):
|
59 |
"""
|
@@ -66,24 +47,23 @@ def rewrite(model, data):
|
|
66 |
"""
|
67 |
|
68 |
with torch.no_grad():
|
69 |
-
|
70 |
-
|
71 |
-
|
72 |
output = list() # save the indices of the characters as list of integers
|
73 |
|
74 |
# Conversion table for Turkish characters {100: [300, 350], ...}
|
75 |
en2tr = {en: tr for tr, en in zip(list(map(list, map(str.encode, list('ÜİĞŞÇÖüığşçö')))), list(map(ord, list('UIGSCOuigsco'))))}
|
76 |
|
77 |
-
for inp, lab in zip((data['input_ids'] - 3)
|
78 |
if lab and inp in en2tr:
|
79 |
# if the model predicts a diacritic, replace it with the corresponding Turkish character
|
80 |
output.extend(en2tr[inp])
|
81 |
elif inp >= 0: output.append(inp)
|
82 |
return bytes(output).decode()
|
83 |
|
84 |
-
def try_it(text, model)
|
85 |
-
sample = test_mask(
|
86 |
-
return rewrite(model, sample
|
87 |
|
88 |
try_it('Cekoslovakyalilastiramadiklarimizdan misiniz?', model)
|
89 |
```
|
|
|
9 |
---
|
10 |
|
11 |
```python
|
12 |
+
from peft import PeftModel, PeftConfig
|
13 |
+
from transformers import AutoModelForTokenClassification
|
|
|
14 |
|
15 |
+
config = PeftConfig.from_pretrained("bite-the-byte/byt5-small-deASCIIfy-TR")
|
16 |
+
model = AutoModelForTokenClassification.from_pretrained("google/byt5-small")
|
17 |
+
model = PeftModel.from_pretrained(model, "bite-the-byte/byt5-small-deASCIIfy-TR")
|
18 |
|
19 |
+
def test_mask(device, sample):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
20 |
"""
|
21 |
Masks the padded tokens in the input.
|
22 |
Args:
|
|
|
25 |
dataset (list): List of dictionaries.
|
26 |
"""
|
27 |
|
28 |
+
tokens = dict()
|
|
|
|
|
29 |
|
30 |
+
input_tokens = [i + 3 for i in sample.encode('utf-8')]
|
31 |
+
input_tokens.append(0) # eos token
|
32 |
+
tokens['input_ids'] = torch.tensor([input_tokens], dtype=torch.int64, device=device)
|
33 |
+
|
34 |
+
# Create attention mask
|
35 |
+
tokens['attention_mask'] = torch.ones_like(tokens['input_ids'], dtype=torch.int64, device=device)
|
36 |
+
|
37 |
+
return tokens
|
|
|
|
|
|
|
38 |
|
39 |
def rewrite(model, data):
|
40 |
"""
|
|
|
47 |
"""
|
48 |
|
49 |
with torch.no_grad():
|
50 |
+
pred = torch.argmax(model(**data).logits, dim=2).squeeze(0)
|
51 |
+
|
|
|
52 |
output = list() # save the indices of the characters as list of integers
|
53 |
|
54 |
# Conversion table for Turkish characters {100: [300, 350], ...}
|
55 |
en2tr = {en: tr for tr, en in zip(list(map(list, map(str.encode, list('ÜİĞŞÇÖüığşçö')))), list(map(ord, list('UIGSCOuigsco'))))}
|
56 |
|
57 |
+
for inp, lab in zip((data['input_ids'].squeeze(0) - 3).tolist(), pred.tolist()):
|
58 |
if lab and inp in en2tr:
|
59 |
# if the model predicts a diacritic, replace it with the corresponding Turkish character
|
60 |
output.extend(en2tr[inp])
|
61 |
elif inp >= 0: output.append(inp)
|
62 |
return bytes(output).decode()
|
63 |
|
64 |
+
def try_it(text, model):
|
65 |
+
sample = test_mask(model.device, text)
|
66 |
+
return rewrite(model, sample)
|
67 |
|
68 |
try_it('Cekoslovakyalilastiramadiklarimizdan misiniz?', model)
|
69 |
```
|