Datasets:

Modalities:
Text
Formats:
parquet
Languages:
English
Libraries:
Datasets
pandas
License:
arjunguha commited on
Commit
7b6337d
1 Parent(s): fcdb075

Added an example

Browse files
Files changed (1) hide show
  1. README.md +42 -1
README.md CHANGED
@@ -37,5 +37,46 @@ generation that supports 18 programming languages. It takes the OpenAI
37
  "HumanEval" Python benchmarks and uses little compilers to translate them
38
  to other languages. It is easy to add support for new languages and benchmarks.
39
 
40
- [More Information Needed]
41
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
37
  "HumanEval" Python benchmarks and uses little compilers to translate them
38
  to other languages. It is easy to add support for new languages and benchmarks.
39
 
40
+ ## Example
41
 
42
+ The following script uses the Salesforce/codegen model to generate Lua
43
+ and MultiPL-E to produce a script with unit tests for luaunit.
44
+
45
+ ```python
46
+ import datasets
47
+ from transformers import AutoTokenizer, AutoModelForCausalLM
48
+
49
+ LANG = "lua"
50
+ MODEL_NAME = "Salesforce/codegen-350M-multi"
51
+ tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
52
+ model = AutoModelForCausalLM.from_pretrained(MODEL_NAME).half().cuda()
53
+ problems = datasets.load_dataset("nuprl/MultiPL-E", LANG)
54
+
55
+ def stop_at_stop_token(decoded_string, problem):
56
+ """
57
+ Truncates the output at stop tokens, taking care to skip the prompt
58
+ which may have stop tokens.
59
+ """
60
+ min_stop_index = len(decoded_string)
61
+ for stop_token in problem["stop_tokens"]:
62
+ stop_index = decoded_string.find(stop_token)
63
+ if stop_index != -1 and stop_index > len(problem["prompt"]) and stop_index < min_stop_index:
64
+ min_stop_index = stop_index
65
+ return decoded_string[:min_stop_index]
66
+
67
+ for problem in problems["test"]:
68
+ input_ids = tokenizer(
69
+ problem["prompt"],
70
+ return_tensors="pt",
71
+ ).input_ids.cuda()
72
+ generated_ids = model.generate(
73
+ input_ids, max_length=256, pad_token_id=tokenizer.eos_token_id + 2
74
+ )
75
+ truncated_string = stop_at_stop_token(tokenizer.decode(generated_ids[0]), problem)
76
+ filename = problem["name"] + "." + LANG
77
+ with open(filename, "w") as f:
78
+ print(f"Created {filename}")
79
+ f.write(truncated_string)
80
+ f.write("\n")
81
+ f.write(problem["tests"])
82
+ ```