Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -1,4 +1,3 @@
|
|
1 |
-
import os
|
2 |
import numpy as np
|
3 |
import onnxruntime as ort
|
4 |
from transformers import AutoTokenizer
|
@@ -15,22 +14,42 @@ def translate_text(input_text):
|
|
15 |
input_text, return_tensors="np", padding=True, truncation=True, max_length=512
|
16 |
)
|
17 |
|
|
|
18 |
input_ids = tokenized_input["input_ids"].astype(np.int64)
|
19 |
attention_mask = tokenized_input["attention_mask"].astype(np.int64)
|
20 |
|
21 |
-
#
|
22 |
-
|
23 |
-
|
24 |
-
{
|
25 |
-
"input_ids": input_ids,
|
26 |
-
"attention_mask": attention_mask,
|
27 |
-
}
|
28 |
-
)
|
29 |
|
30 |
-
#
|
31 |
-
translated_tokens =
|
32 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
33 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
34 |
return translated_text
|
35 |
|
36 |
# Create a Gradio interface
|
|
|
|
|
1 |
import numpy as np
|
2 |
import onnxruntime as ort
|
3 |
from transformers import AutoTokenizer
|
|
|
14 |
input_text, return_tensors="np", padding=True, truncation=True, max_length=512
|
15 |
)
|
16 |
|
17 |
+
# Prepare encoder inputs
|
18 |
input_ids = tokenized_input["input_ids"].astype(np.int64)
|
19 |
attention_mask = tokenized_input["attention_mask"].astype(np.int64)
|
20 |
|
21 |
+
# Prepare decoder inputs (start with the start token)
|
22 |
+
decoder_start_token_id = translation_tokenizer.cls_token_id or translation_tokenizer.pad_token_id
|
23 |
+
decoder_input_ids = np.array([[decoder_start_token_id]], dtype=np.int64)
|
|
|
|
|
|
|
|
|
|
|
24 |
|
25 |
+
# Iteratively generate output tokens
|
26 |
+
translated_tokens = []
|
27 |
+
for _ in range(512): # Max length of output
|
28 |
+
# Run inference with the ONNX model
|
29 |
+
outputs = translation_session.run(
|
30 |
+
None,
|
31 |
+
{
|
32 |
+
"input_ids": input_ids,
|
33 |
+
"attention_mask": attention_mask,
|
34 |
+
"decoder_input_ids": decoder_input_ids,
|
35 |
+
}
|
36 |
+
)
|
37 |
+
|
38 |
+
# Get the next token ID
|
39 |
+
next_token_id = np.argmax(outputs[0][0, -1, :], axis=-1)
|
40 |
+
translated_tokens.append(next_token_id)
|
41 |
|
42 |
+
# Stop if the end-of-sequence token is generated
|
43 |
+
if next_token_id == translation_tokenizer.eos_token_id:
|
44 |
+
break
|
45 |
+
|
46 |
+
# Update decoder_input_ids for the next iteration
|
47 |
+
decoder_input_ids = np.concatenate(
|
48 |
+
[decoder_input_ids, np.array([[next_token_id]], dtype=np.int64)], axis=1
|
49 |
+
)
|
50 |
+
|
51 |
+
# Decode the output tokens
|
52 |
+
translated_text = translation_tokenizer.decode(translated_tokens, skip_special_tokens=True)
|
53 |
return translated_text
|
54 |
|
55 |
# Create a Gradio interface
|