Views
No views yet
1import torch
2from transformers import (
3 AutoModelForSeq2SeqLM,
4 AutoModelForCausalLM,
5 AutoTokenizer
6)
7
8model_name = "baidu/ernie-code-560m"
9
10model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
11tokenizer = AutoTokenizer.from_pretrained(model_name)
12# note that you can use aforementioned `clean_up_code_spaces` to proprocess the code
13
14
15def format_code_with_spm_compatablity(line: str):
16 format_dict = {
17 " " : "<|space|>"
18 }
19 tokens = list(line)
20 i = 0
21 while i < len(tokens):
22 if line[i] == "\n":
23 while i+1 < len(tokens) and tokens[i+1] == " ":
24 tokens[i+1] = format_dict.get(" ")
25 i += 1
26 i += 1
27 formatted_line = ''.join(tokens)
28 return formatted_line
29
30"""
31TYPE="code" # define input type in ("code", "text")
32input="arr.sort()"
33prompt="translate python to java: \n%s" % (input) # your prompt here
34"""
35
36TYPE="text" # define input type in ("code", "text")
37input="quick sort"
38prompt="translate English to Japanese: \n%s" % (input) # your prompt here
39
40assert TYPE in ("code", "text")
41
42# preprocess for code input
43if TYPE=="code":
44 prompt = format_code_with_spm_compatablity(prompt)
45
46model_inputs = tokenizer(prompt, max_length=512, padding=False, truncation=True, return_tensors="pt")
47
48model = model.cuda() # by default
49input_ids = model_inputs.input_ids.cuda() # by default
50attention_mask = model_inputs.attention_mask.cuda() # by default
51
52output = model.generate(input_ids=input_ids, attention_mask=attention_mask,
53 num_beams=5, max_length=20) # change to your needs
54
55# Ensure to customize the post-processing of `clean_up_code_spaces` output according to specific requirements.
56output = tokenizer.decode(output.flatten(), skip_special_tokens=True)
57
58
59# post-process the code generation
60def clean_up_code_spaces(s: str):
61 # post process
62 # ===========================
63 new_tokens = ["<pad>", "</s>", "<unk>", "\n", "\t", "<|space|>"*4, "<|space|>"*2, "<|space|>"]
64 for tok in new_tokens:
65 s = s.replace(f"{tok} ", tok)
66
67 cleaned_tokens = ["<pad>", "</s>", "<unk>"]
68 for tok in cleaned_tokens:
69 s = s.replace(tok, "")
70 s = s.replace("<|space|>", " ")
71 return s
72output = [clean_up_code_spaces(pred) for pred in output]
73@inproceedings{chai-etal-2023-ernie,
title = "{ERNIE}-Code: Beyond {E}nglish-Centric Cross-lingual Pretraining for Programming Languages",
author = "Chai, Yekun and
Wang, Shuohuan and
Pang, Chao and
Sun, Yu and
Tian, Hao and
Wu, Hua",
booktitle = "Findings of the Association for Computational Linguistics: ACL 2023",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-acl.676",
pages = "10628--10650",
abstract = "Software engineers working with the same programming language (PL) may speak different natural languages (NLs) and vice versa, erecting huge barriers to communication and working efficiency. Recent studies have demonstrated the effectiveness of generative pre-training in computer programs, yet they are always English-centric. In this work, we step towards bridging the gap between multilingual NLs and multilingual PLs for large language models (LLMs). We release ERNIE-Code, a unified pre-trained language model for 116 NLs and 6 PLs. We employ two methods for universal cross-lingual pre-training: span-corruption language modeling that learns patterns from monolingual NL or PL; and pivot-based translation language modeling that relies on parallel data of many NLs and PLs. Extensive results show that ERNIE-Code outperforms previous multilingual LLMs for PL or NL across a wide range of end tasks of code intelligence, including multilingual code-to-text, text-to-code, code-to-code, and text-to-text generation. We further show its advantage of zero-shot prompting on multilingual code summarization and text-to-text translation. We release our code and pre-trained checkpoints.",
}