| { |
| "sha": "4993b99b6a97e5e19cfa7d4cd75e12545580ad27", |
| "node_id": "C_kwDOJKNTJdoAKDQ5OTNiOTliNmE5N2U1ZTE5Y2ZhN2Q0Y2Q3NWUxMjU0NTU4MGFkMjc", |
| "commit": { |
| "author": { |
| "name": "Mohammad Abdullah Matin Khan", |
| "email": "zarziskhan@gmail.com", |
| "date": "2024-09-17T21:16:04Z" |
| }, |
| "committer": { |
| "name": "GitHub", |
| "email": "noreply@github.com", |
| "date": "2024-09-17T21:16:04Z" |
| }, |
| "message": "Merge pull request #8 from ntunlp/evaluation_code\n\nEvaluation code", |
| "tree": { |
| "sha": "b543a448c95ef034731b6321bdfba04623cd6a5f", |
| "url": "https://api.github.com/repos/ntunlp/xCodeEval/git/trees/b543a448c95ef034731b6321bdfba04623cd6a5f" |
| }, |
| "url": "https://api.github.com/repos/ntunlp/xCodeEval/git/commits/4993b99b6a97e5e19cfa7d4cd75e12545580ad27", |
| "comment_count": 0, |
| "verification": { |
| "verified": true, |
| "reason": "valid", |
| "signature": "-----BEGIN PGP SIGNATURE-----\n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\nPJmqegKXTjpOdYxh95tM\n=YsX+\n-----END PGP SIGNATURE-----\n", |
| "payload": "tree b543a448c95ef034731b6321bdfba04623cd6a5f\nparent 93a4200949b0d7558d98510113d95f3bb39ce72d\nparent 392f05ef229c228a3712020dcd532eb97139ff40\nauthor Mohammad Abdullah Matin Khan <zarziskhan@gmail.com> 1726607764 +0800\ncommitter GitHub <noreply@github.com> 1726607764 +0800\n\nMerge pull request #8 from ntunlp/evaluation_code\n\nEvaluation code", |
| "verified_at": "2024-11-10T06:19:36Z" |
| } |
| }, |
| "url": "https://api.github.com/repos/ntunlp/xCodeEval/commits/4993b99b6a97e5e19cfa7d4cd75e12545580ad27", |
| "html_url": "https://github.com/ntunlp/xCodeEval/commit/4993b99b6a97e5e19cfa7d4cd75e12545580ad27", |
| "comments_url": "https://api.github.com/repos/ntunlp/xCodeEval/commits/4993b99b6a97e5e19cfa7d4cd75e12545580ad27/comments", |
| "author": { |
| "login": "Jackal1586", |
| "id": 7633619, |
| "node_id": "MDQ6VXNlcjc2MzM2MTk=", |
| "avatar_url": "https://avatars.githubusercontent.com/u/7633619?v=4", |
| "gravatar_id": "", |
| "url": "https://api.github.com/users/Jackal1586", |
| "html_url": "https://github.com/Jackal1586", |
| "followers_url": "https://api.github.com/users/Jackal1586/followers", |
| "following_url": "https://api.github.com/users/Jackal1586/following{/other_user}", |
| "gists_url": "https://api.github.com/users/Jackal1586/gists{/gist_id}", |
| "starred_url": "https://api.github.com/users/Jackal1586/starred{/owner}{/repo}", |
| "subscriptions_url": "https://api.github.com/users/Jackal1586/subscriptions", |
| "organizations_url": "https://api.github.com/users/Jackal1586/orgs", |
| "repos_url": "https://api.github.com/users/Jackal1586/repos", |
| "events_url": "https://api.github.com/users/Jackal1586/events{/privacy}", |
| "received_events_url": "https://api.github.com/users/Jackal1586/received_events", |
| "type": "User", |
| "user_view_type": "public", |
| "site_admin": false |
| }, |
| "committer": { |
| "login": "web-flow", |
| "id": 19864447, |
| "node_id": "MDQ6VXNlcjE5ODY0NDQ3", |
| "avatar_url": "https://avatars.githubusercontent.com/u/19864447?v=4", |
| "gravatar_id": "", |
| "url": "https://api.github.com/users/web-flow", |
| "html_url": "https://github.com/web-flow", |
| "followers_url": "https://api.github.com/users/web-flow/followers", |
| "following_url": "https://api.github.com/users/web-flow/following{/other_user}", |
| "gists_url": "https://api.github.com/users/web-flow/gists{/gist_id}", |
| "starred_url": "https://api.github.com/users/web-flow/starred{/owner}{/repo}", |
| "subscriptions_url": "https://api.github.com/users/web-flow/subscriptions", |
| "organizations_url": "https://api.github.com/users/web-flow/orgs", |
| "repos_url": "https://api.github.com/users/web-flow/repos", |
| "events_url": "https://api.github.com/users/web-flow/events{/privacy}", |
| "received_events_url": "https://api.github.com/users/web-flow/received_events", |
| "type": "User", |
| "user_view_type": "public", |
| "site_admin": false |
| }, |
| "parents": [ |
| { |
| "sha": "93a4200949b0d7558d98510113d95f3bb39ce72d", |
| "url": "https://api.github.com/repos/ntunlp/xCodeEval/commits/93a4200949b0d7558d98510113d95f3bb39ce72d", |
| "html_url": "https://github.com/ntunlp/xCodeEval/commit/93a4200949b0d7558d98510113d95f3bb39ce72d" |
| }, |
| { |
| "sha": "392f05ef229c228a3712020dcd532eb97139ff40", |
| "url": "https://api.github.com/repos/ntunlp/xCodeEval/commits/392f05ef229c228a3712020dcd532eb97139ff40", |
| "html_url": "https://github.com/ntunlp/xCodeEval/commit/392f05ef229c228a3712020dcd532eb97139ff40" |
| } |
| ], |
| "stats": { |
| "total": 1855, |
| "additions": 1855, |
| "deletions": 0 |
| }, |
| "files": [ |
| { |
| "sha": "228d9126794775db58e2646add312190aa56b8c8", |
| "filename": "README.md", |
| "status": "modified", |
| "additions": 5, |
| "deletions": 0, |
| "changes": 5, |
| "blob_url": "https://github.com/ntunlp/xCodeEval/blob/4993b99b6a97e5e19cfa7d4cd75e12545580ad27/README.md", |
| "raw_url": "https://github.com/ntunlp/xCodeEval/raw/4993b99b6a97e5e19cfa7d4cd75e12545580ad27/README.md", |
| "contents_url": "https://api.github.com/repos/ntunlp/xCodeEval/contents/README.md?ref=4993b99b6a97e5e19cfa7d4cd75e12545580ad27", |
| "patch": "@@ -2,6 +2,8 @@\n [xCodeEval: A Large Scale Multilingual Multitask Benchmark for Code Understanding, Generation, Translation and Retrieval](https://arxiv.org/abs/2303.03004)\n \n # Update:\n+\n+- September 18, 2024: Evaluation code for generative tasks released. [Follow it here](#evaluation)\n - Nov 7, 2023: A sample eval script is [here](https://github.com/ntunlp/xCodeEval/pull/8).\n - July 13, 2023: StarEncode retrieval model released. [Follow it here](#additional-resources)\n - Jul 6, 2023: [ExecEval](https://github.com/ntunlp/ExecEval) has been updated with changes for java, kotlin, go. Please `git pull`, `docker build`, `docker run` for latest updates.\n@@ -78,6 +80,9 @@ We propose 7 Tasks.\n 6. [Code-Code Retrieval](./retrieval.md)\n 7. [NL-Code Retrieval](./retrieval.md)\n \n+# Evaluation\n+For details on evaluation please follow instructions from [evaluation/README.md](./evaluation/README.md).\n+\n # Common Data for different tasks\n \n " |
| }, |
| { |
| "sha": "86b45771e83747fb2b25a88c421953041f3fa86c", |
| "filename": "evaluation/README.md", |
| "status": "added", |
| "additions": 53, |
| "deletions": 0, |
| "changes": 53, |
| "blob_url": "https://github.com/ntunlp/xCodeEval/blob/4993b99b6a97e5e19cfa7d4cd75e12545580ad27/evaluation%2FREADME.md", |
| "raw_url": "https://github.com/ntunlp/xCodeEval/raw/4993b99b6a97e5e19cfa7d4cd75e12545580ad27/evaluation%2FREADME.md", |
| "contents_url": "https://api.github.com/repos/ntunlp/xCodeEval/contents/evaluation%2FREADME.md?ref=4993b99b6a97e5e19cfa7d4cd75e12545580ad27", |
| "patch": "@@ -0,0 +1,53 @@\n+# How to perform evaluation using ExecEval\n+\n+## Configure ExecEval\n+\n+Follow the instrauction [here](https://github.com/ntunlp/execeval).\n+\n+> :warning: ❌ Do not run ExecEval without Docker image\n+\n+`Java` and `Kotlin 1.5` has a lots of Memory related issue. If you don't have large amount of memory please reduce the multiple worker. If you are running in a laptop or Desktop with limited amount of RAM (~ 32 GB), please use num `NUM_WORKERS`=1. Performance holds well upto 1/3rd of the CPU available at max.\n+\n+## Setup Environment\n+Install python packages in your own environment.\n+```\n+pip install -r requirement.txt\n+```\n+\n+Install ExecEval.\n+```\n+git clone https://github.com/ntunlp/ExecEval\n+cd ExecEval\n+docker build . -t exec-eval:1.0\n+docker run -it -p 5000:5000 -e NUM_WORKERS=37 exec-eval:1.0\n+```\n+\n+## Generate samples\n+\n+Generate samples using OpenAI api.\n+\n+```\n+python evaluation/program_synthesis/gen_program_synthesis.py\n+python evaluation/code_translation/gen_code_translation.py\n+python evaluation/apr/gen_apr.py\n+```\n+\n+## Eval Samples using ExecEval\n+\n+Keep ExecEval server/endpoint running and then run the following code, \n+\n+```\n+python evaluation/program_synthesis/eval_program_synthesis.py\n+python evaluation/code_translation/eval_code_translation.py\n+python evaluation/apr/eval_apr.py\n+```\n+\n+## Calculate pass@k\n+\n+Calculate pass@k by the following script,\n+\n+```\n+python evaluation/program_synthesis/get_result.py\n+python evaluation/code_translation/get_result.py\n+python evaluation/apr/get_result.py\n+```\n\\ No newline at end of file" |
| }, |
| { |
| "sha": "54cf58ee29b2efa705f28d49c1e992ee8b138340", |
| "filename": "evaluation/apr/eval_apr.py", |
| "status": "added", |
| "additions": 259, |
| "deletions": 0, |
| "changes": 259, |
| "blob_url": "https://github.com/ntunlp/xCodeEval/blob/4993b99b6a97e5e19cfa7d4cd75e12545580ad27/evaluation%2Fapr%2Feval_apr.py", |
| "raw_url": "https://github.com/ntunlp/xCodeEval/raw/4993b99b6a97e5e19cfa7d4cd75e12545580ad27/evaluation%2Fapr%2Feval_apr.py", |
| "contents_url": "https://api.github.com/repos/ntunlp/xCodeEval/contents/evaluation%2Fapr%2Feval_apr.py?ref=4993b99b6a97e5e19cfa7d4cd75e12545580ad27", |
| "patch": "@@ -0,0 +1,259 @@\n+import os\n+import json\n+import tqdm\n+import jsonlines\n+import datasets\n+import concurrent.futures\n+from dataclasses import dataclass, field\n+import itertools\n+\n+import requests\n+from typing import List, Optional, Union, Tuple\n+from enum import Enum\n+from multiprocessing import Pool\n+\n+\n+class ExecOutcome(Enum):\n+ PASSED = \"PASSED\" # code executes and output matches expected output\n+ WRONG_ANSWER = (\n+ \"WRONG_ANSWER\" # code executes and output does NOT matches expected output\n+ )\n+ TIME_LIMIT_EXCEEDED = \"TIME_LIMIT_EXCEEDED\" # code executes and didn't exit in time, output is ignored in this case\n+ RUNTIME_ERROR = \"RUNTIME_ERROR\" # code failed to execute (crashed)\n+ COMPILATION_ERROR = \"COMPILATION_ERROR\" # code failed to compile\n+ MEMORY_LIMIT_EXCEEDED = (\n+ \"MEMORY_LIMIT_EXCEEDED\" # code exceeded memory limit during execution\n+ )\n+\n+\n+@dataclass\n+class ExtendedUnittest:\n+ input: str\n+ output: List[str] = field(default_factory=list)\n+ result: Optional[str] = None\n+ exec_outcome: Optional[ExecOutcome] = None\n+\n+ def json(self):\n+ _json = self.__dict__\n+ if self.exec_outcome is not None:\n+ _json[\"exec_outcome\"] = self.exec_outcome.name\n+\n+ return _json\n+\n+ @classmethod\n+ def from_json(cls, _json):\n+ return cls(\n+ input=_json.get(\"input\", \"\"),\n+ output=_json.get(\"output\", list()),\n+ result=_json.get(\"result\", None),\n+ exec_outcome=_json.get(\"exec_outcome\", None),\n+ )\n+\n+\n+class EmptyValueError(Exception):\n+ def __init__(self, *args, **kwargs):\n+ super().__init__(*args, **kwargs)\n+\n+\n+class EmptyUnittestError(EmptyValueError):\n+ pass\n+\n+\n+class EmptyLanguageError(EmptyValueError):\n+ pass\n+\n+\n+class EmptySourceCodeError(EmptyValueError):\n+ pass\n+\n+\n+class APICommunication:\n+ _session: requests.Session\n+\n+ def __init__(self, server_url: str = \"http://localhost:5000\"):\n+ self._session = requests.Session()\n+ self.execute_code_url = f\"{server_url}/api/execute_code\"\n+ self.get_runtimes_url = f\"{server_url}/api/all_runtimes\"\n+\n+ def __enter__(self):\n+ return self\n+\n+ def __exit__(self, *args):\n+ self._session.close()\n+\n+ def get_runtimes(self):\n+ return self._session.get(self.get_runtimes_url).json()\n+\n+ def execute_code(\n+ self,\n+ language: str,\n+ source_code: str,\n+ unittests: List[dict],\n+ limits: Optional[dict] = None,\n+ block_network: bool = True,\n+ stop_on_first_fail: bool = True,\n+ use_sanitizer: bool = False,\n+ compiler_program_name: Optional[str] = None,\n+ compiler_flags: Optional[str] = None,\n+ interpreter_cmd: Optional[str] = None,\n+ interpreter_flags: Optional[str] = None,\n+ sample_id: Optional[int] = None,\n+ task_id: Union[str, int, None] = None,\n+ ) -> Tuple[List[ExtendedUnittest], Optional[int], Union[str, int, None]]:\n+ if language is None:\n+ raise EmptyLanguageError\n+\n+ if source_code is None:\n+ raise EmptySourceCodeError\n+\n+ if unittests is None or len(unittests) == 0:\n+ raise EmptyUnittestError\n+\n+ request_body = dict(\n+ language=language,\n+ source_code=source_code,\n+ unittests=unittests,\n+ limits=limits if isinstance(limits, dict) else None,\n+ compile_cmd=compiler_program_name,\n+ compile_flags=compiler_flags,\n+ execute_cmd=interpreter_cmd,\n+ execute_flags=interpreter_flags,\n+ block_network=block_network,\n+ stop_on_first_fail=stop_on_first_fail,\n+ use_sanitizer=use_sanitizer,\n+ )\n+ json_response = self._session.post(\n+ self.execute_code_url,\n+ json=request_body,\n+ headers={\"Content-Type\": \"application/json\"},\n+ ).json()\n+\n+ if \"data\" not in json_response:\n+ return json_response, sample_id, task_id\n+\n+ return (\n+ json_response[\"data\"],\n+ sample_id,\n+ task_id,\n+ )\n+\n+\n+def get_idx(file_name):\n+ return int(file_name.split(\".json\")[0].split(\"_\")[0])\n+\n+\n+def sanitize_code(code):\n+ FLAG = True\n+ while FLAG == True:\n+ FLAG = False\n+ if code.startswith(\"```\"):\n+ FLAG = True\n+ code = code.replace(\"```\", \"\", 1)\n+ last_index = code.rfind(\"```\")\n+ if last_index != -1:\n+ FLAG = True\n+ code = code[:last_index] + \"\" + code[last_index + len(\"```\") :]\n+ if code.startswith(\"cpp\"):\n+ FLAG = True\n+ code = code.replace(\"cpp\", \"\", 1)\n+ return code\n+\n+\n+def fix_uts(uts):\n+ uts_fx = []\n+ for ut in uts:\n+ uts_fx.append(\n+ {\n+ \"input\": ut[\"input\"],\n+ \"output\": ut[\"output\"],\n+ }\n+ )\n+ return uts_fx\n+\n+\n+def process(args):\n+ sample, execeval = args\n+ src_uid = sample[\"source_data\"][\"src_uid\"]\n+ unit_tests = json.loads(sample[\"source_data\"][\"hidden_unit_tests\"])\n+ compiler = LANG_CLUSTER_TO_LANG_COMPILER[sample[\"source_data\"][\"lang_cluster\"]]\n+ sample[\"unit_test_results\"] = []\n+ for choice in sample[\"oai_response\"][\"choices\"]:\n+ code = choice[\"message\"][\"content\"]\n+ code = sanitize_code(code)\n+ unit_test_results, _, _ = execeval.execute_code(\n+ compiler,\n+ code,\n+ fix_uts(unit_tests),\n+ task_id=src_uid,\n+ # stop_on_first_fail=False\n+ )\n+ # print(unit_test_results)\n+ # print(file, code, [e['exec_outcome'] for e in unit_test_results])\n+ sample[\"unit_test_results\"].append(unit_test_results)\n+ return sample\n+\n+\n+LANG_CLUSTER_TO_LANG_COMPILER = {\n+ \"C\": \"GNU C11\",\n+ \"C#\": \"Mono C#\",\n+ \"C++\": \"GNU C++17\",\n+ \"Go\": \"Go\",\n+ \"Java\": \"Java 17\",\n+ \"Javascript\": \"Node.js\",\n+ \"Kotlin\": \"Kotlin 1.4\",\n+ \"PHP\": \"PHP\",\n+ \"Python\": \"PyPy 3\",\n+ \"Ruby\": \"Ruby 3\",\n+ \"Rust\": \"Rust 2018\",\n+}\n+\n+\n+def main():\n+ path = f'{os.environ[\"DUMP_FOLDER\"]}/oai/apr_n_sample_20/'\n+ for k, debug_compiler in LANG_CLUSTER_TO_LANG_COMPILER.items():\n+ output_path = os.path.join(path, \"eval_apr_val_execeval\")\n+ os.makedirs(output_path, exist_ok=True)\n+ output_file = os.path.join(output_path, f\"{debug_compiler}.jsonl\")\n+ with jsonlines.open(output_file, \"w\") as jwp:\n+ with concurrent.futures.ThreadPoolExecutor(\n+ max_workers=129\n+ ) as thread_executor:\n+ files = sorted(os.listdir(path))\n+ with APICommunication(server_url=\"http://localhost:5000\") as execeval:\n+ all_samples = []\n+ for file in files:\n+ full_path = os.path.join(path, file)\n+ if os.path.isdir(full_path):\n+ continue\n+ sample = json.load(open(full_path))\n+ if (\n+ sample[\"source_data\"][\"lang_cluster\"]\n+ not in LANG_CLUSTER_TO_LANG_COMPILER\n+ ):\n+ continue\n+ compiler = LANG_CLUSTER_TO_LANG_COMPILER[\n+ sample[\"source_data\"][\"lang_cluster\"]\n+ ]\n+ if compiler != debug_compiler:\n+ continue\n+ all_samples.append(sample)\n+ future_to_val_results = {\n+ thread_executor.submit(process, args)\n+ for args in itertools.product(all_samples, [execeval])\n+ }\n+\n+ for _out in tqdm.tqdm(\n+ concurrent.futures.as_completed(future_to_val_results),\n+ total=len(all_samples),\n+ desc=f\"{debug_compiler}\",\n+ ):\n+ try:\n+ __out = _out.result()\n+ jwp.write(__out)\n+ except Exception as emsg:\n+ print(\"Exception msg: {}\".format(emsg))\n+ pass\n+\n+\n+if __name__ == \"__main__\":\n+ main()" |
| }, |
| { |
| "sha": "8a83911385cd8b38fc2f655f0a459e1a5be0debc", |
| "filename": "evaluation/apr/gen_apr.py", |
| "status": "added", |
| "additions": 182, |
| "deletions": 0, |
| "changes": 182, |
| "blob_url": "https://github.com/ntunlp/xCodeEval/blob/4993b99b6a97e5e19cfa7d4cd75e12545580ad27/evaluation%2Fapr%2Fgen_apr.py", |
| "raw_url": "https://github.com/ntunlp/xCodeEval/raw/4993b99b6a97e5e19cfa7d4cd75e12545580ad27/evaluation%2Fapr%2Fgen_apr.py", |
| "contents_url": "https://api.github.com/repos/ntunlp/xCodeEval/contents/evaluation%2Fapr%2Fgen_apr.py?ref=4993b99b6a97e5e19cfa7d4cd75e12545580ad27", |
| "patch": "@@ -0,0 +1,182 @@\n+import os\n+import time\n+import tqdm\n+import json\n+import openai\n+import argparse\n+import datasets\n+import concurrent\n+import numpy as np\n+from promptsource.templates import Template\n+\n+SHORT_LANG_MAP = {\n+ \"GNU C++\": \"C++\",\n+ \"GNU C++17\": \"C++\",\n+ \"MS C++ 2017\": \"C++\",\n+ \"MS C++\": \"C++\",\n+ \"Java 8\": \"Java\",\n+ \"Java 6\": \"Java\",\n+ \"GNU C++11\": \"C++\",\n+ \"Java 11\": \"Java\",\n+ \"GNU C++14\": \"C++\",\n+ \"Mono C#\": \"C#\",\n+ \"GNU C\": \"C\",\n+ \"Python 3\": \"Python\",\n+ \"PyPy 3\": \"Python\",\n+ \"GNU C11\": \"C\",\n+ \"Go\": \"Go\",\n+ \"Rust\": \"Rust\",\n+ \"PyPy 2\": \"Python\",\n+ \"Python 2\": \"Python\",\n+ \"MS C#\": \"C#\",\n+ \"Kotlin\": \"Kotlin\",\n+ \"GNU C++0x\": \"C++\",\n+ \"Java 7\": \"Java\",\n+ \"Node.js\": \"Javascript\",\n+ \".NET Core C#\": \"C#\",\n+ \"PHP\": \"PHP\",\n+ \"GNU C++17 Diagnostics\": \"C++\",\n+ \"Clang++17 Diagnostics\": \"C++\",\n+ \"JavaScript\": \"Javascript\",\n+ \"Ruby\": \"Ruby\",\n+ \"C# 10\": \"C#\",\n+ \"C# 8\": \"C#\",\n+ \"Clang++20 Diagnostics\": \"C++\",\n+ \"GNU C++17 (64)\": \"C++\",\n+ \"GNU C++20 (64)\": \"C++\",\n+ \"Java 17\": \"Java\",\n+ \"Kotlin 1.4\": \"Kotlin\",\n+ \"Kotlin 1.5\": \"Kotlin\",\n+ \"Kotlin 1.6\": \"Kotlin\",\n+ \"Kotlin 1.7\": \"Kotlin\",\n+ \"PyPy 3-64\": \"Python\",\n+ \"Python 3 + libs\": \"Python\",\n+ \"Ruby 3\": \"Ruby\",\n+ \"Rust 2021\": \"Rust\",\n+}\n+\n+LANGS = sorted(set([v for k, v in SHORT_LANG_MAP.items()]))\n+\n+\n+openai.api_key = os.environ[\"OPENAI_API_KEY\"]\n+\n+\n+def gen(prompt, temperature, nsample):\n+ cnt = 0\n+ while True:\n+ if cnt == 999:\n+ return None\n+ try:\n+ c = openai.ChatCompletion.create(\n+ model=\"gpt-3.5-turbo\",\n+ messages=[\n+ {\"role\": \"user\", \"content\": f\"{prompt}\"},\n+ ],\n+ temperature=temperature,\n+ top_p=1,\n+ n=nsample,\n+ frequency_penalty=0.0,\n+ presence_penalty=0.0,\n+ )\n+ break\n+ except Exception as e:\n+ cnt += 1\n+ time.sleep(5)\n+ print(f\"{e}\")\n+ c[\"prompt\"] = prompt\n+ return c\n+\n+\n+xcodeeval_prompt_template = {\n+ \"apr\": [\n+ \"Fix a buggy program written in {{lang_cluster}} language to solve the following programming problem:\\nDescription: {{prob_desc_description}}\\nInput Specification: {{prob_desc_input_spec}}\\nOutput Specification: {{prob_desc_output_spec}}\\n{% for input, output in zip(prob_desc_sample_inputs, prob_desc_sample_outputs) %}\\nSample Input:\\n{{input}}\\nSample Output:\\n{{output}}\\n{% endfor %}\\nNotes: {{prob_desc_notes}}\\nTake input from {{prob_desc_input_from}} and output to {{prob_desc_output_to}}\\n\\nHere is the code with a bug of {{bug_exec_outcome}}:\\n\\n{{bug_source_code}}\\n\\nProvide the fixed {{lang_cluster}} code without any description or extra tokens.\\n\\nFixed source code:\\n ||END-of-SRC|| \"\n+ ]\n+}\n+\n+\n+def process_prompt(dt, temperature, template, nsample, output_dir, index, dry_run=0):\n+ language = dt[\"lang_cluster\"]\n+ file_path = os.path.join(output_dir, f\"{index}_{temperature}_{language}.json\")\n+ if not os.path.exists(file_path):\n+ dt[\"prob_desc_sample_inputs\"] = json.loads(dt[\"prob_desc_sample_inputs\"])\n+ dt[\"prob_desc_sample_outputs\"] = json.loads(dt[\"prob_desc_sample_outputs\"])\n+ lm_io = template.apply(dt)\n+ assert len(lm_io) == 2, f\"{json.dumps(lm_io, indent=4)}\"\n+ if dry_run:\n+ open(file_path, \"w\").write(f\"{json.dumps(lm_io[0], indent=4)}\")\n+ else:\n+ out = gen(lm_io[0], temperature, nsample)\n+ export_data = {\"oai_response\": out, \"source_data\": dt}\n+ open(file_path, \"w\").write(f\"{json.dumps(export_data, indent=4)}\")\n+\n+\n+def main():\n+ parser = argparse.ArgumentParser()\n+ parser.add_argument(\n+ \"--output-dir\",\n+ default=\"dumped/oai/apr_n_sample_20\",\n+ help=\"Output Folder to save the API request.\",\n+ )\n+ parser.add_argument(\n+ \"--num-proc\",\n+ default=1,\n+ help=\"Number of parallel API request.\",\n+ )\n+ parser.add_argument(\n+ \"--dry-run\",\n+ default=0,\n+ help=\"Number of parallel API request.\",\n+ )\n+ parser.add_argument(\n+ \"--nsample\",\n+ default=20,\n+ type=int,\n+ help=\"Number of parallel API request.\",\n+ )\n+ args = parser.parse_args()\n+ if not os.path.exists(args.output_dir):\n+ os.makedirs(args.output_dir, exist_ok=True)\n+ templates = [\n+ Template(f\"apr_{idx}\", template, \"xCodeEval\", delimeter=\"||END-of-SRC||\")\n+ for idx, template in enumerate(xcodeeval_prompt_template[\"apr\"])\n+ ]\n+ template = templates[0]\n+\n+ apr_dataset = datasets.load_dataset(\"NTU-NLP-sg/xCodeEval\", \"apr\", num_proc=16, trust_remote_code=True)[\"compact\"]\n+ # temperature_list = np.linspace(0, 2, args.nsample)\n+ temperature_list = [0.3157894736842105]\n+ with concurrent.futures.ProcessPoolExecutor(\n+ max_workers=int(args.num_proc)\n+ ) as executor:\n+ futures = []\n+ for idx, dt in tqdm.tqdm(\n+ enumerate(apr_dataset),\n+ total=len(apr_dataset),\n+ desc=f\"Preparing samples lang\",\n+ ):\n+ for temperature in temperature_list:\n+ future = executor.submit(\n+ process_prompt,\n+ dt,\n+ temperature,\n+ template,\n+ args.nsample,\n+ args.output_dir,\n+ idx,\n+ args.dry_run,\n+ )\n+ futures.append(future)\n+\n+ for future in tqdm.tqdm(\n+ concurrent.futures.as_completed(futures),\n+ total=len(futures),\n+ desc=f\"Calling OpenAI API\",\n+ ):\n+ try:\n+ future.result()\n+ except Exception as e:\n+ print(f\"Error occurred: {e}\")\n+\n+\n+if __name__ == \"__main__\":\n+ main()" |
| }, |
| { |
| "sha": "1331c96fa652a673ad1d3e0e8007a7754dc6a88e", |
| "filename": "evaluation/apr/get_result.py", |
| "status": "added", |
| "additions": 105, |
| "deletions": 0, |
| "changes": 105, |
| "blob_url": "https://github.com/ntunlp/xCodeEval/blob/4993b99b6a97e5e19cfa7d4cd75e12545580ad27/evaluation%2Fapr%2Fget_result.py", |
| "raw_url": "https://github.com/ntunlp/xCodeEval/raw/4993b99b6a97e5e19cfa7d4cd75e12545580ad27/evaluation%2Fapr%2Fget_result.py", |
| "contents_url": "https://api.github.com/repos/ntunlp/xCodeEval/contents/evaluation%2Fapr%2Fget_result.py?ref=4993b99b6a97e5e19cfa7d4cd75e12545580ad27", |
| "patch": "@@ -0,0 +1,105 @@\n+import os\n+from collections import defaultdict\n+import tqdm\n+import jsonlines\n+from typing import List, Union\n+import itertools\n+import numpy as np\n+\n+LANG_CLUSTER_TO_LANG_COMPILER = {\n+ \"C\": \"GNU C11\",\n+ \"C#\": \"Mono C#\",\n+ \"C++\": \"GNU C++17\",\n+ \"Go\": \"Go\",\n+ \"Java\": \"Java 17\",\n+ \"Javascript\": \"Node.js\",\n+ \"Kotlin\": \"Kotlin 1.4\",\n+ \"PHP\": \"PHP\",\n+ \"Python\": \"PyPy 3\",\n+ \"Ruby\": \"Ruby 3\",\n+ \"Rust\": \"Rust 2018\",\n+}\n+\n+path = f'{os.environ[\"DUMP_FOLDER\"]}/oai/apr_n_sample_20/'\n+output_path = os.path.join(\n+ path, \"eval_apr_val_execeval\"\n+) \n+ks = range(1, 21)\n+\n+\n+def estimate_pass_at_k(\n+ num_samples: Union[int, List[int], np.ndarray],\n+ num_correct: Union[List[int], np.ndarray],\n+ k: int,\n+) -> np.ndarray:\n+ \"\"\"\n+ Estimates pass@k of each problem and returns them in an array.\n+ \"\"\"\n+\n+ def estimator(n: int, c: int, k: int):\n+ \"\"\"\n+ Calculates 1 - comb(n - c, k) / comb(n, k).\n+ \"\"\"\n+ if n - c < k:\n+ return 1.0\n+ return 1.0 - np.prod(1.0 - k / np.arange(n - c + 1, n + 1))\n+\n+ if isinstance(num_samples, int):\n+ num_samples_it = itertools.repeat(num_samples, len(num_correct))\n+ else:\n+ assert len(num_samples) == len(num_correct)\n+ num_samples_it = iter(num_samples)\n+\n+ return np.array(\n+ [estimator(int(n), int(c), k) for n, c in zip(num_samples_it, num_correct)]\n+ )\n+\n+\n+def get_execeval_out_file_name(compiler):\n+ return os.path.join(output_path, f\"{compiler}.jsonl\")\n+\n+\n+# construct result as {[task_id]: [unit_test_results]}\n+# task_id will be src_uid_lang\n+\n+pass_at_k = defaultdict(dict)\n+\n+for lang, compiler in tqdm.tqdm(LANG_CLUSTER_TO_LANG_COMPILER.items()):\n+ execeval_out_file = get_execeval_out_file_name(compiler)\n+ results = defaultdict(list)\n+ with jsonlines.open(execeval_out_file) as jrp:\n+ for sample in jrp:\n+ src_uid = sample[\"source_data\"][\"src_uid\"]\n+ task_id = f\"{src_uid}|||{lang}\"\n+ for ut_res in sample[\"unit_test_results\"]:\n+ if \"error\" in ut_res:\n+ continue\n+ results[task_id].append(ut_res)\n+\n+ total, correct = [], []\n+ for result in results.values():\n+ passed = [\n+ all(x[\"exec_outcome\"] == \"PASSED\" for x in ut_res) for ut_res in result\n+ ]\n+ total.append(len(passed))\n+ correct.append(sum(passed))\n+ total = np.array(total)\n+ correct = np.array(correct)\n+\n+ pass_at_k[lang] = {\n+ f\"pass@{k}\": estimate_pass_at_k(total, correct, k).mean()\n+ for k in ks\n+ if (total >= k).all()\n+ }\n+\n+\n+langs = sorted(list(pass_at_k.keys()))\n+for lang in langs:\n+ print(f\" & {lang}\", end=\"\")\n+print()\n+avg = 0\n+for lang in langs:\n+ print(f\" & {round(pass_at_k[lang]['pass@5']*100, 2)}\", end=\"\")\n+ avg += pass_at_k[lang][\"pass@5\"] * 100\n+avg /= len(langs)\n+print(f\" & {round(avg, 2)}\")" |
| }, |
| { |
| "sha": "eb7facba706e758d2482c056f16af6264d7eddd9", |
| "filename": "evaluation/code_translation/eval_code_translation.py", |
| "status": "added", |
| "additions": 264, |
| "deletions": 0, |
| "changes": 264, |
| "blob_url": "https://github.com/ntunlp/xCodeEval/blob/4993b99b6a97e5e19cfa7d4cd75e12545580ad27/evaluation%2Fcode_translation%2Feval_code_translation.py", |
| "raw_url": "https://github.com/ntunlp/xCodeEval/raw/4993b99b6a97e5e19cfa7d4cd75e12545580ad27/evaluation%2Fcode_translation%2Feval_code_translation.py", |
| "contents_url": "https://api.github.com/repos/ntunlp/xCodeEval/contents/evaluation%2Fcode_translation%2Feval_code_translation.py?ref=4993b99b6a97e5e19cfa7d4cd75e12545580ad27", |
| "patch": "@@ -0,0 +1,264 @@\n+import os\n+import json\n+import tqdm\n+import jsonlines\n+import datasets\n+import concurrent.futures\n+from dataclasses import dataclass, field\n+import itertools\n+from collections import defaultdict\n+\n+import requests\n+from typing import List, Optional, Union, Tuple\n+from enum import Enum\n+from multiprocessing import Pool\n+\n+\n+class ExecOutcome(Enum):\n+ PASSED = \"PASSED\" # code executes and output matches expected output\n+ WRONG_ANSWER = (\n+ \"WRONG_ANSWER\" # code executes and output does NOT matches expected output\n+ )\n+ TIME_LIMIT_EXCEEDED = \"TIME_LIMIT_EXCEEDED\" # code executes and didn't exit in time, output is ignored in this case\n+ RUNTIME_ERROR = \"RUNTIME_ERROR\" # code failed to execute (crashed)\n+ COMPILATION_ERROR = \"COMPILATION_ERROR\" # code failed to compile\n+ MEMORY_LIMIT_EXCEEDED = (\n+ \"MEMORY_LIMIT_EXCEEDED\" # code exceeded memory limit during execution\n+ )\n+\n+\n+@dataclass\n+class ExtendedUnittest:\n+ input: str\n+ output: List[str] = field(default_factory=list)\n+ result: Optional[str] = None\n+ exec_outcome: Optional[ExecOutcome] = None\n+\n+ def json(self):\n+ _json = self.__dict__\n+ if self.exec_outcome is not None:\n+ _json[\"exec_outcome\"] = self.exec_outcome.name\n+\n+ return _json\n+\n+ @classmethod\n+ def from_json(cls, _json):\n+ return cls(\n+ input=_json.get(\"input\", \"\"),\n+ output=_json.get(\"output\", list()),\n+ result=_json.get(\"result\", None),\n+ exec_outcome=_json.get(\"exec_outcome\", None),\n+ )\n+\n+\n+class EmptyValueError(Exception):\n+ def __init__(self, *args, **kwargs):\n+ super().__init__(*args, **kwargs)\n+\n+\n+class EmptyUnittestError(EmptyValueError):\n+ pass\n+\n+\n+class EmptyLanguageError(EmptyValueError):\n+ pass\n+\n+\n+class EmptySourceCodeError(EmptyValueError):\n+ pass\n+\n+\n+class APICommunication:\n+ _session: requests.Session\n+\n+ def __init__(self, server_url: str = \"http://localhost:5000\"):\n+ self._session = requests.Session()\n+ self.execute_code_url = f\"{server_url}/api/execute_code\"\n+ self.get_runtimes_url = f\"{server_url}/api/all_runtimes\"\n+\n+ def __enter__(self):\n+ return self\n+\n+ def __exit__(self, *args):\n+ self._session.close()\n+\n+ def get_runtimes(self):\n+ return self._session.get(self.get_runtimes_url).json()\n+\n+ def execute_code(\n+ self,\n+ language: str,\n+ source_code: str,\n+ unittests: List[dict],\n+ limits: Optional[dict] = None,\n+ block_network: bool = True,\n+ stop_on_first_fail: bool = True,\n+ use_sanitizer: bool = False,\n+ compiler_program_name: Optional[str] = None,\n+ compiler_flags: Optional[str] = None,\n+ interpreter_cmd: Optional[str] = None,\n+ interpreter_flags: Optional[str] = None,\n+ sample_id: Optional[int] = None,\n+ task_id: Union[str, int, None] = None,\n+ ) -> Tuple[List[ExtendedUnittest], Optional[int], Union[str, int, None]]:\n+ if language is None:\n+ raise EmptyLanguageError\n+\n+ if source_code is None:\n+ raise EmptySourceCodeError\n+\n+ if unittests is None or len(unittests) == 0:\n+ raise EmptyUnittestError\n+\n+ request_body = dict(\n+ language=language,\n+ source_code=source_code,\n+ unittests=unittests,\n+ limits=limits if isinstance(limits, dict) else None,\n+ compile_cmd=compiler_program_name,\n+ compile_flags=compiler_flags,\n+ execute_cmd=interpreter_cmd,\n+ execute_flags=interpreter_flags,\n+ block_network=block_network,\n+ stop_on_first_fail=stop_on_first_fail,\n+ use_sanitizer=use_sanitizer,\n+ )\n+ json_response = self._session.post(\n+ self.execute_code_url,\n+ json=request_body,\n+ headers={\"Content-Type\": \"application/json\"},\n+ ).json()\n+\n+ if \"data\" not in json_response:\n+ return json_response, sample_id, task_id\n+\n+ return (\n+ json_response[\"data\"],\n+ sample_id,\n+ task_id,\n+ )\n+\n+\n+def get_idx(file_name):\n+ return int(file_name.split(\".json\")[0].split(\"_\")[0])\n+\n+\n+def sanitize_code(code):\n+ FLAG = True\n+ while FLAG == True:\n+ FLAG = False\n+ if code.startswith(\"```\"):\n+ FLAG = True\n+ code = code.replace(\"```\", \"\", 1)\n+ last_index = code.rfind(\"```\")\n+ if last_index != -1:\n+ FLAG = True\n+ code = code[:last_index] + \"\" + code[last_index + len(\"```\") :]\n+ if code.startswith(\"cpp\"):\n+ FLAG = True\n+ code = code.replace(\"cpp\", \"\", 1)\n+ return code\n+\n+\n+def fix_uts(uts):\n+ uts_fx = []\n+ for ut in uts:\n+ uts_fx.append(\n+ {\n+ \"input\": ut[\"input\"],\n+ \"output\": ut[\"output\"],\n+ }\n+ )\n+ return uts_fx\n+\n+\n+def process(args):\n+ sample, execeval = args\n+ src_uid = sample[\"source_data\"][\"src_uid\"]\n+ unit_tests = json.loads(sample[\"source_data\"][\"hidden_unit_tests\"])\n+ compiler = LANG_CLUSTER_TO_LANG_COMPILER[sample[\"source_data\"][\"target_lang\"]]\n+ sample[\"unit_test_results\"] = list()\n+ for choice in sample[\"oai_response\"][\"choices\"]:\n+ code = choice[\"message\"][\"content\"]\n+ code = sanitize_code(code)\n+ unit_test_results, _, _ = execeval.execute_code(\n+ compiler,\n+ code,\n+ fix_uts(unit_tests),\n+ task_id=src_uid, # stop_on_first_fail=False\n+ )\n+ # print(unit_test_results)\n+ # print(file, code, [e['exec_outcome'] for e in unit_test_results])\n+ sample[\"unit_test_results\"].append(unit_test_results)\n+ return sample\n+\n+\n+LANG_CLUSTER_TO_LANG_COMPILER = {\n+ \"C\": \"GNU C11\",\n+ \"C#\": \"Mono C#\",\n+ \"C++\": \"GNU C++17\",\n+ \"Go\": \"Go\",\n+ \"Java\": \"Java 17\",\n+ \"Javascript\": \"Node.js\",\n+ \"Kotlin\": \"Kotlin 1.4\",\n+ \"PHP\": \"PHP\",\n+ \"Python\": \"PyPy 3\",\n+ \"Ruby\": \"Ruby 3\",\n+ \"Rust\": \"Rust 2018\",\n+}\n+\n+\n+def main():\n+ parent_path = (\n+ f'{os.environ[\"DUMP_FOLDER\"]}/oai/code_translation_n_sample_20/'\n+ )\n+ for split in (\"compact\", \"compact_small\"):\n+ path = os.path.join(parent_path, split)\n+ for k, debug_compiler in LANG_CLUSTER_TO_LANG_COMPILER.items():\n+ output_path = os.path.join(path, f\"eval_code_translation_{split}_execeval\")\n+ os.makedirs(output_path, exist_ok=True)\n+ output_file = os.path.join(output_path, f\"{debug_compiler}.jsonl\")\n+ with jsonlines.open(output_file, \"w\") as jwp:\n+ with concurrent.futures.ThreadPoolExecutor(\n+ max_workers=129\n+ ) as thread_executor:\n+ files = sorted(os.listdir(path))\n+ with APICommunication(\n+ server_url=\"http://localhost:5000\"\n+ ) as execeval:\n+ all_samples = []\n+ for file in files:\n+ full_path = os.path.join(path, file)\n+ if os.path.isdir(full_path):\n+ continue\n+ sample = json.load(open(full_path))\n+ if (\n+ sample[\"source_data\"][\"target_lang\"]\n+ not in LANG_CLUSTER_TO_LANG_COMPILER\n+ ):\n+ continue\n+ compiler = LANG_CLUSTER_TO_LANG_COMPILER[\n+ sample[\"source_data\"][\"target_lang\"]\n+ ]\n+ if compiler != debug_compiler:\n+ continue\n+ all_samples.append(sample)\n+ future_to_val_results = {\n+ thread_executor.submit(process, args)\n+ for args in itertools.product(all_samples, [execeval])\n+ }\n+ for _out in tqdm.tqdm(\n+ concurrent.futures.as_completed(future_to_val_results),\n+ total=len(all_samples),\n+ desc=f\"{debug_compiler}\",\n+ ):\n+ # try:\n+ __out = _out.result()\n+ jwp.write(__out)\n+ # except Exception as emsg:\n+ # print(\"Exception msg: {}\".format(emsg))\n+ # pass\n+\n+\n+if __name__ == \"__main__\":\n+ main()\n\\ No newline at end of file" |
| }, |
| { |
| "sha": "7d68faa2d2b6fcd07ef678e955ea012e70c0cb7b", |
| "filename": "evaluation/code_translation/gen_code_translation.py", |
| "status": "added", |
| "additions": 243, |
| "deletions": 0, |
| "changes": 243, |
| "blob_url": "https://github.com/ntunlp/xCodeEval/blob/4993b99b6a97e5e19cfa7d4cd75e12545580ad27/evaluation%2Fcode_translation%2Fgen_code_translation.py", |
| "raw_url": "https://github.com/ntunlp/xCodeEval/raw/4993b99b6a97e5e19cfa7d4cd75e12545580ad27/evaluation%2Fcode_translation%2Fgen_code_translation.py", |
| "contents_url": "https://api.github.com/repos/ntunlp/xCodeEval/contents/evaluation%2Fcode_translation%2Fgen_code_translation.py?ref=4993b99b6a97e5e19cfa7d4cd75e12545580ad27", |
| "patch": "@@ -0,0 +1,243 @@\n+import os\n+import time\n+import tqdm\n+import json\n+import openai\n+import argparse\n+import datasets\n+import concurrent\n+import numpy as np\n+from promptsource.templates import Template\n+\n+SHORT_LANG_MAP = {\n+ \"GNU C++\": \"C++\",\n+ \"GNU C++17\": \"C++\",\n+ \"MS C++ 2017\": \"C++\",\n+ \"MS C++\": \"C++\",\n+ \"Java 8\": \"Java\",\n+ \"Java 6\": \"Java\",\n+ \"GNU C++11\": \"C++\",\n+ \"Java 11\": \"Java\",\n+ \"GNU C++14\": \"C++\",\n+ \"Mono C#\": \"C#\",\n+ \"GNU C\": \"C\",\n+ \"Python 3\": \"Python\",\n+ \"PyPy 3\": \"Python\",\n+ \"GNU C11\": \"C\",\n+ \"Go\": \"Go\",\n+ \"Rust\": \"Rust\",\n+ \"PyPy 2\": \"Python\",\n+ \"Python 2\": \"Python\",\n+ \"MS C#\": \"C#\",\n+ \"Kotlin\": \"Kotlin\",\n+ \"GNU C++0x\": \"C++\",\n+ \"Java 7\": \"Java\",\n+ \"Node.js\": \"Javascript\",\n+ \".NET Core C#\": \"C#\",\n+ \"PHP\": \"PHP\",\n+ \"GNU C++17 Diagnostics\": \"C++\",\n+ \"Clang++17 Diagnostics\": \"C++\",\n+ \"JavaScript\": \"Javascript\",\n+ \"Ruby\": \"Ruby\",\n+ \"C# 10\": \"C#\",\n+ \"C# 8\": \"C#\",\n+ \"Clang++20 Diagnostics\": \"C++\",\n+ \"GNU C++17 (64)\": \"C++\",\n+ \"GNU C++20 (64)\": \"C++\",\n+ \"Java 17\": \"Java\",\n+ \"Kotlin 1.4\": \"Kotlin\",\n+ \"Kotlin 1.5\": \"Kotlin\",\n+ \"Kotlin 1.6\": \"Kotlin\",\n+ \"Kotlin 1.7\": \"Kotlin\",\n+ \"PyPy 3-64\": \"Python\",\n+ \"Python 3 + libs\": \"Python\",\n+ \"Ruby 3\": \"Ruby\",\n+ \"Rust 2021\": \"Rust\",\n+}\n+\n+LANGS = sorted(set([v for k, v in SHORT_LANG_MAP.items()]))\n+\n+\n+openai.api_key = os.environ[\"OPENAI_API_KEY\"]\n+\n+\n+def gen(prompt, temperature, nsample):\n+ cnt = 0\n+ while True:\n+ if cnt == 999:\n+ return None\n+ try:\n+ c = openai.ChatCompletion.create(\n+ model=\"gpt-3.5-turbo\",\n+ messages=[\n+ {\"role\": \"user\", \"content\": f\"{prompt}\"},\n+ ],\n+ temperature=temperature,\n+ top_p=1,\n+ n=nsample,\n+ frequency_penalty=0.0,\n+ presence_penalty=0.0,\n+ )\n+ break\n+ except Exception as e:\n+ cnt += 1\n+ time.sleep(5)\n+ print(f\"{e}\")\n+ c[\"prompt\"] = prompt\n+ return c\n+\n+\n+xcodeeval_prompt_template = {\n+ \"code_translation\": [\n+ \"Here is code in {{source_lang}} programming lanaguge. Translate the following code from {{source_lang}} to {{target_lang}} programming lanaguge. Do not output any extra description or tokens other than the translated code. \\n\\n{{source_code}}||END-of-SRC|| \"\n+ ]\n+}\n+\n+\n+def process_prompt(\n+ dt, temperature, template, language, nsample, output_dir, index, dry_run=0\n+):\n+ dt[\"source_lang\"] = dt[\"lang\"]\n+ dt[\"target_lang\"] = language\n+ language = f\"{dt['source_lang']}--{dt['target_lang']}\"\n+ file_path = os.path.join(output_dir, f\"{index}_{temperature}_{language}.json\")\n+ if not os.path.exists(file_path):\n+ dt[\"prob_desc_sample_inputs\"] = json.loads(dt[\"prob_desc_sample_inputs\"])\n+ dt[\"prob_desc_sample_outputs\"] = json.loads(dt[\"prob_desc_sample_outputs\"])\n+ lm_io = template.apply(dt)\n+ assert len(lm_io) == 2, f\"{json.dumps(lm_io, indent=4)}\"\n+ if dry_run:\n+ open(file_path, \"w\").write(f\"{json.dumps(lm_io[0], indent=4)}\")\n+ else:\n+ out = gen(lm_io[0], temperature, nsample)\n+ export_data = {\"oai_response\": out, \"source_data\": dt}\n+ open(file_path, \"w\").write(f\"{json.dumps(export_data, indent=4)}\")\n+\n+\n+def main():\n+ parser = argparse.ArgumentParser()\n+ parser.add_argument(\n+ \"--output-dir\",\n+ default=\"dumped/oai/code_translation_n_sample_20\",\n+ help=\"Output Folder to save the API request.\",\n+ )\n+ parser.add_argument(\n+ \"--num-proc\",\n+ default=1,\n+ help=\"Number of parallel API request.\",\n+ )\n+ parser.add_argument(\n+ \"--dry-run\",\n+ default=0,\n+ help=\"Number of parallel API request.\",\n+ )\n+ parser.add_argument(\n+ \"--nsample\",\n+ default=20,\n+ type=int,\n+ help=\"Number of parallel API request.\",\n+ )\n+ args = parser.parse_args()\n+ if not os.path.exists(args.output_dir):\n+ os.makedirs(args.output_dir, exist_ok=True)\n+ templates = [\n+ Template(\n+ f\"code_translation_{idx}\", template, \"xCodeEval\", delimeter=\"||END-of-SRC||\"\n+ )\n+ for idx, template in enumerate(xcodeeval_prompt_template[\"code_translation\"])\n+ ]\n+ template = templates[0]\n+\n+ code_translation_dataset_small = datasets.load_dataset(\n+ \"NTU-NLP-sg/xCodeEval\", \"code_translation\", num_proc=16, trust_remote_code=True\n+ )[\n+ \"compact_small\"\n+ ]\n+ code_translation_dataset = datasets.load_dataset(\n+ \"NTU-NLP-sg/xCodeEval\", \"code_translation\", num_proc=16\n+ )[\n+ \"compact\"\n+ ]\n+ temperature_list = [0.3157894736842105]\n+\n+ out_dir = args.output_dir + \"/compact_small\"\n+ if not os.path.exists(out_dir):\n+ os.makedirs(out_dir, exist_ok=True)\n+ with concurrent.futures.ProcessPoolExecutor(\n+ max_workers=int(args.num_proc)\n+ ) as executor:\n+ futures = []\n+ for idx, dt in tqdm.tqdm(\n+ enumerate(code_translation_dataset_small),\n+ total=len(code_translation_dataset_small),\n+ desc=f\"Preparing samples\",\n+ ):\n+ for language in LANGS:\n+ if SHORT_LANG_MAP[dt[\"lang\"]] == language:\n+ continue\n+ for temperature in temperature_list:\n+ future = executor.submit(\n+ process_prompt,\n+ dt,\n+ temperature,\n+ template,\n+ language,\n+ args.nsample,\n+ out_dir,\n+ idx,\n+ args.dry_run,\n+ )\n+ futures.append(future)\n+\n+ for future in tqdm.tqdm(\n+ concurrent.futures.as_completed(futures),\n+ total=len(futures),\n+ desc=f\"Calling OpenAI API\",\n+ ):\n+ try:\n+ future.result()\n+ except Exception as e:\n+ print(f\"Error occurred: {e}\")\n+\n+ out_dir = args.output_dir + \"/compact\"\n+ if not os.path.exists(out_dir):\n+ os.makedirs(out_dir, exist_ok=True)\n+ with concurrent.futures.ProcessPoolExecutor(\n+ max_workers=int(args.num_proc)\n+ ) as executor:\n+ futures = []\n+ for idx, dt in tqdm.tqdm(\n+ enumerate(code_translation_dataset),\n+ total=len(code_translation_dataset),\n+ desc=f\"Preparing samples\",\n+ ):\n+ for language in [\"Python\"]:\n+ if SHORT_LANG_MAP[dt[\"lang\"]] == language:\n+ continue\n+ for temperature in temperature_list:\n+ future = executor.submit(\n+ process_prompt,\n+ dt,\n+ temperature,\n+ template,\n+ language,\n+ args.nsample,\n+ out_dir,\n+ idx,\n+ args.dry_run,\n+ )\n+ futures.append(future)\n+\n+ for future in tqdm.tqdm(\n+ concurrent.futures.as_completed(futures),\n+ total=len(futures),\n+ desc=f\"Calling OpenAI API\",\n+ ):\n+ try:\n+ future.result()\n+ except Exception as e:\n+ print(f\"Error occurred: {e}\")\n+\n+\n+if __name__ == \"__main__\":\n+ main()" |
| }, |
| { |
| "sha": "423c13ffad689804f45917eecd0a718241470525", |
| "filename": "evaluation/code_translation/get_result.py", |
| "status": "added", |
| "additions": 104, |
| "deletions": 0, |
| "changes": 104, |
| "blob_url": "https://github.com/ntunlp/xCodeEval/blob/4993b99b6a97e5e19cfa7d4cd75e12545580ad27/evaluation%2Fcode_translation%2Fget_result.py", |
| "raw_url": "https://github.com/ntunlp/xCodeEval/raw/4993b99b6a97e5e19cfa7d4cd75e12545580ad27/evaluation%2Fcode_translation%2Fget_result.py", |
| "contents_url": "https://api.github.com/repos/ntunlp/xCodeEval/contents/evaluation%2Fcode_translation%2Fget_result.py?ref=4993b99b6a97e5e19cfa7d4cd75e12545580ad27", |
| "patch": "@@ -0,0 +1,104 @@\n+import os\n+from collections import defaultdict\n+import tqdm\n+import jsonlines\n+from typing import List, Union\n+import itertools\n+import numpy as np\n+\n+LANG_CLUSTER_TO_LANG_COMPILER = {\n+ \"C\": \"GNU C11\",\n+ \"C#\": \"Mono C#\",\n+ \"C++\": \"GNU C++17\",\n+ \"Go\": \"Go\",\n+ \"Java\": \"Java 17\",\n+ \"Javascript\": \"Node.js\",\n+ \"Kotlin\": \"Kotlin 1.4\",\n+ \"PHP\": \"PHP\",\n+ \"Python\": \"PyPy 3\",\n+ \"Ruby\": \"Ruby 3\",\n+ \"Rust\": \"Rust 2018\",\n+}\n+\n+path = f'{os.environ[\"DUMP_FOLDER\"]}/oai/code_translation_n_sample_20/'\n+ks = range(1, 21)\n+\n+\n+def estimate_pass_at_k(\n+ num_samples: Union[int, List[int], np.ndarray],\n+ num_correct: Union[List[int], np.ndarray],\n+ k: int,\n+) -> np.ndarray:\n+ \"\"\"\n+ Estimates pass@k of each problem and returns them in an array.\n+ \"\"\"\n+\n+ def estimator(n: int, c: int, k: int):\n+ \"\"\"\n+ Calculates 1 - comb(n - c, k) / comb(n, k).\n+ \"\"\"\n+ if n - c < k:\n+ return 1.0\n+ return 1.0 - np.prod(1.0 - k / np.arange(n - c + 1, n + 1))\n+\n+ if isinstance(num_samples, int):\n+ num_samples_it = itertools.repeat(num_samples, len(num_correct))\n+ else:\n+ assert len(num_samples) == len(num_correct)\n+ num_samples_it = iter(num_samples)\n+\n+ return np.array(\n+ [estimator(int(n), int(c), k) for n, c in zip(num_samples_it, num_correct)]\n+ )\n+\n+\n+def get_execeval_out_file_name(split_name, compiler):\n+ return os.path.join(path, split_name, \"eval_code_translation_compact_small_execeval\", f\"{compiler}.jsonl\")\n+\n+\n+# construct result as {[task_id]: [unit_test_results]}\n+# task_id will be src_uid_lang\n+\n+pass_at_k = defaultdict(dict)\n+for split_name in (\"compact_small\", \"compact\"):\n+ for lang, compiler in tqdm.tqdm(LANG_CLUSTER_TO_LANG_COMPILER.items()):\n+ execeval_out_file = get_execeval_out_file_name(split_name, compiler)\n+ results = defaultdict(list)\n+ with jsonlines.open(execeval_out_file) as jrp:\n+ for sample in jrp:\n+ src_uid = sample[\"source_data\"][\"src_uid\"]\n+ task_id = f\"{src_uid}|||{lang}\"\n+ for ut_res in sample[\"unit_test_results\"]:\n+ if \"error\" in ut_res:\n+ continue\n+ results[task_id].append(ut_res)\n+\n+ total, correct = [], []\n+ for result in results.values():\n+ passed = [\n+ all(x[\"exec_outcome\"] == \"PASSED\" for x in ut_res) for ut_res in result\n+ ]\n+ total.append(len(passed))\n+ correct.append(sum(passed))\n+ total = np.array(total)\n+ correct = np.array(correct)\n+\n+ pass_at_k[lang] = {\n+ f\"pass@{k}\": estimate_pass_at_k(total, correct, k).mean()\n+ for k in ks\n+ if (total >= k).all()\n+ }\n+\n+ print(\"-\"*10)\n+ print(split_name)\n+ print(\"-\"*10)\n+ langs = sorted(list(pass_at_k.keys()))\n+ for lang in langs:\n+ print(f\" & {lang}\", end=\"\")\n+ print()\n+ avg = 0\n+ for lang in langs:\n+ print(f\" & {round(pass_at_k[lang]['pass@5']*100, 2)}\", end=\"\")\n+ avg += pass_at_k[lang][\"pass@5\"] * 100\n+ avg /= len(langs)\n+ print(f\" & {round(avg, 2)}\")" |
| }, |
| { |
| "sha": "a51dc6363c38b23dd3423a63410fd99b651907d1", |
| "filename": "evaluation/program_synthesis/eval_program_synthesis.py", |
| "status": "added", |
| "additions": 258, |
| "deletions": 0, |
| "changes": 258, |
| "blob_url": "https://github.com/ntunlp/xCodeEval/blob/4993b99b6a97e5e19cfa7d4cd75e12545580ad27/evaluation%2Fprogram_synthesis%2Feval_program_synthesis.py", |
| "raw_url": "https://github.com/ntunlp/xCodeEval/raw/4993b99b6a97e5e19cfa7d4cd75e12545580ad27/evaluation%2Fprogram_synthesis%2Feval_program_synthesis.py", |
| "contents_url": "https://api.github.com/repos/ntunlp/xCodeEval/contents/evaluation%2Fprogram_synthesis%2Feval_program_synthesis.py?ref=4993b99b6a97e5e19cfa7d4cd75e12545580ad27", |
| "patch": "@@ -0,0 +1,258 @@\n+import os\n+import json\n+import tqdm\n+import jsonlines\n+import datasets\n+import concurrent.futures\n+from dataclasses import dataclass, field\n+import itertools\n+from collections import defaultdict\n+\n+import requests\n+from typing import List, Optional, Union, Tuple\n+from enum import Enum\n+from multiprocessing import Pool\n+\n+\n+class ExecOutcome(Enum):\n+ PASSED = \"PASSED\" # code executes and output matches expected output\n+ WRONG_ANSWER = (\n+ \"WRONG_ANSWER\" # code executes and output does NOT matches expected output\n+ )\n+ TIME_LIMIT_EXCEEDED = \"TIME_LIMIT_EXCEEDED\" # code executes and didn't exit in time, output is ignored in this case\n+ RUNTIME_ERROR = \"RUNTIME_ERROR\" # code failed to execute (crashed)\n+ COMPILATION_ERROR = \"COMPILATION_ERROR\" # code failed to compile\n+ MEMORY_LIMIT_EXCEEDED = (\n+ \"MEMORY_LIMIT_EXCEEDED\" # code exceeded memory limit during execution\n+ )\n+\n+\n+@dataclass\n+class ExtendedUnittest:\n+ input: str\n+ output: List[str] = field(default_factory=list)\n+ result: Optional[str] = None\n+ exec_outcome: Optional[ExecOutcome] = None\n+\n+ def json(self):\n+ _json = self.__dict__\n+ if self.exec_outcome is not None:\n+ _json[\"exec_outcome\"] = self.exec_outcome.name\n+\n+ return _json\n+\n+ @classmethod\n+ def from_json(cls, _json):\n+ return cls(\n+ input=_json.get(\"input\", \"\"),\n+ output=_json.get(\"output\", list()),\n+ result=_json.get(\"result\", None),\n+ exec_outcome=_json.get(\"exec_outcome\", None),\n+ )\n+\n+\n+class EmptyValueError(Exception):\n+ def __init__(self, *args, **kwargs):\n+ super().__init__(*args, **kwargs)\n+\n+\n+class EmptyUnittestError(EmptyValueError):\n+ pass\n+\n+\n+class EmptyLanguageError(EmptyValueError):\n+ pass\n+\n+\n+class EmptySourceCodeError(EmptyValueError):\n+ pass\n+\n+\n+class APICommunication:\n+ _session: requests.Session\n+\n+ def __init__(self, server_url: str = \"http://localhost:5000\"):\n+ self._session = requests.Session()\n+ self.execute_code_url = f\"{server_url}/api/execute_code\"\n+ self.get_runtimes_url = f\"{server_url}/api/all_runtimes\"\n+\n+ def __enter__(self):\n+ return self\n+\n+ def __exit__(self, *args):\n+ self._session.close()\n+\n+ def get_runtimes(self):\n+ return self._session.get(self.get_runtimes_url).json()\n+\n+ def execute_code(\n+ self,\n+ language: str,\n+ source_code: str,\n+ unittests: List[dict],\n+ limits: Optional[dict] = None,\n+ block_network: bool = True,\n+ stop_on_first_fail: bool = True,\n+ use_sanitizer: bool = False,\n+ compiler_program_name: Optional[str] = None,\n+ compiler_flags: Optional[str] = None,\n+ interpreter_cmd: Optional[str] = None,\n+ interpreter_flags: Optional[str] = None,\n+ sample_id: Optional[int] = None,\n+ task_id: Union[str, int, None] = None,\n+ ) -> Tuple[List[ExtendedUnittest], Optional[int], Union[str, int, None]]:\n+ if language is None:\n+ raise EmptyLanguageError\n+\n+ if source_code is None:\n+ raise EmptySourceCodeError\n+\n+ if unittests is None or len(unittests) == 0:\n+ raise EmptyUnittestError\n+\n+ request_body = dict(\n+ language=language,\n+ source_code=source_code,\n+ unittests=unittests,\n+ limits=limits if isinstance(limits, dict) else None,\n+ compile_cmd=compiler_program_name,\n+ compile_flags=compiler_flags,\n+ execute_cmd=interpreter_cmd,\n+ execute_flags=interpreter_flags,\n+ block_network=block_network,\n+ stop_on_first_fail=stop_on_first_fail,\n+ use_sanitizer=use_sanitizer,\n+ )\n+ json_response = self._session.post(\n+ self.execute_code_url,\n+ json=request_body,\n+ headers={\"Content-Type\": \"application/json\"},\n+ ).json()\n+\n+ if \"data\" not in json_response:\n+ return json_response, sample_id, task_id\n+\n+ return (\n+ json_response[\"data\"],\n+ sample_id,\n+ task_id,\n+ )\n+\n+\n+def get_idx(file_name):\n+ return int(file_name.split(\".json\")[0].split(\"_\")[0])\n+\n+\n+def sanitize_code(code):\n+ FLAG = True\n+ while FLAG == True:\n+ FLAG = False\n+ if code.startswith(\"```\"):\n+ FLAG = True\n+ code = code.replace(\"```\", \"\", 1)\n+ last_index = code.rfind(\"```\")\n+ if last_index != -1:\n+ FLAG = True\n+ code = code[:last_index] + \"\" + code[last_index + len(\"```\") :]\n+ if code.startswith(\"cpp\"):\n+ FLAG = True\n+ code = code.replace(\"cpp\", \"\", 1)\n+ return code\n+\n+\n+def fix_uts(uts):\n+ uts_fx = []\n+ for ut in uts:\n+ uts_fx.append(\n+ {\n+ \"input\": ut[\"input\"],\n+ \"output\": ut[\"output\"],\n+ }\n+ )\n+ return uts_fx\n+\n+\n+def process(args):\n+ sample, execeval = args\n+ src_uid = sample[\"source_data\"][\"src_uid\"]\n+ unit_tests = json.loads(sample[\"source_data\"][\"hidden_unit_tests\"])\n+ compiler = LANG_CLUSTER_TO_LANG_COMPILER[sample[\"source_data\"][\"lang_cluster\"]]\n+ sample[\"unit_test_results\"] = []\n+ for choice in sample[\"oai_response\"][\"choices\"]:\n+ code = choice[\"message\"][\"content\"]\n+ code = sanitize_code(code)\n+ unit_test_results, _, _ = execeval.execute_code(\n+ compiler,\n+ code,\n+ fix_uts(unit_tests),\n+ task_id=src_uid,\n+ # stop_on_first_fail=False\n+ )\n+ # print(unit_test_results)\n+ # print(file, code, [e['exec_outcome'] for e in unit_test_results])\n+ sample[\"unit_test_results\"].append(unit_test_results)\n+ return sample\n+\n+\n+LANG_CLUSTER_TO_LANG_COMPILER = {\n+ \"C\": \"GNU C11\",\n+ \"C#\": \"Mono C#\",\n+ \"C++\": \"GNU C++17\",\n+ \"Go\": \"Go\",\n+ \"Java\": \"Java 17\",\n+ \"Javascript\": \"Node.js\",\n+ \"Kotlin\": \"Kotlin 1.4\",\n+ \"PHP\": \"PHP\",\n+ \"Python\": \"PyPy 3\",\n+ \"Ruby\": \"Ruby 3\",\n+ \"Rust\": \"Rust 2018\",\n+}\n+\n+\n+def main():\n+ path = f'{os.environ[\"DUMP_FOLDER\"]}/oai/prog_synthesis_n_sample_20/'\n+ for k, debug_compiler in LANG_CLUSTER_TO_LANG_COMPILER.items():\n+ output_path = os.path.join(path, \"reproduce_1\")\n+ os.makedirs(output_path, exist_ok=True)\n+ output_file = os.path.join(output_path, f\"{debug_compiler}.jsonl\")\n+ with concurrent.futures.ThreadPoolExecutor(max_workers=129) as thread_executor:\n+ with jsonlines.open(output_file, \"w\") as jwp:\n+ files = sorted(os.listdir(path))\n+ with APICommunication(server_url=\"http://localhost:5000\") as execeval:\n+ all_samples = []\n+ for file in files:\n+ full_path = os.path.join(path, file)\n+ if os.path.isdir(full_path):\n+ continue\n+ sample = json.load(open(full_path))\n+ if (\n+ sample[\"source_data\"][\"lang_cluster\"]\n+ not in LANG_CLUSTER_TO_LANG_COMPILER\n+ ):\n+ continue\n+ compiler = LANG_CLUSTER_TO_LANG_COMPILER[\n+ sample[\"source_data\"][\"lang_cluster\"]\n+ ]\n+ if compiler != debug_compiler:\n+ continue\n+ all_samples.append(sample)\n+ future_to_val_results = {\n+ thread_executor.submit(process, args)\n+ for args in itertools.product(all_samples, [execeval])\n+ }\n+\n+ for _out in tqdm.tqdm(\n+ concurrent.futures.as_completed(future_to_val_results),\n+ total=len(all_samples),\n+ desc=f\"{debug_compiler}\",\n+ ):\n+ try:\n+ __out = _out.result()\n+ jwp.write(__out)\n+ except Exception as emsg:\n+ print(\"Exception msg: {}\".format(emsg))\n+ pass\n+\n+\n+if __name__ == \"__main__\":\n+ main()" |
| }, |
| { |
| "sha": "5202d50c1f4dd379972d452b5bb6c1f59b27d895", |
| "filename": "evaluation/program_synthesis/gen_program_synthesis.py", |
| "status": "added", |
| "additions": 188, |
| "deletions": 0, |
| "changes": 188, |
| "blob_url": "https://github.com/ntunlp/xCodeEval/blob/4993b99b6a97e5e19cfa7d4cd75e12545580ad27/evaluation%2Fprogram_synthesis%2Fgen_program_synthesis.py", |
| "raw_url": "https://github.com/ntunlp/xCodeEval/raw/4993b99b6a97e5e19cfa7d4cd75e12545580ad27/evaluation%2Fprogram_synthesis%2Fgen_program_synthesis.py", |
| "contents_url": "https://api.github.com/repos/ntunlp/xCodeEval/contents/evaluation%2Fprogram_synthesis%2Fgen_program_synthesis.py?ref=4993b99b6a97e5e19cfa7d4cd75e12545580ad27", |
| "patch": "@@ -0,0 +1,188 @@\n+import os\n+import time\n+import tqdm\n+import json\n+import openai\n+import argparse\n+import datasets\n+import concurrent\n+import numpy as np\n+from promptsource.templates import Template\n+\n+SHORT_LANG_MAP = {\n+ \"GNU C++\": \"C++\",\n+ \"GNU C++17\": \"C++\",\n+ \"MS C++ 2017\": \"C++\",\n+ \"MS C++\": \"C++\",\n+ \"Java 8\": \"Java\",\n+ \"Java 6\": \"Java\",\n+ \"GNU C++11\": \"C++\",\n+ \"Java 11\": \"Java\",\n+ \"GNU C++14\": \"C++\",\n+ \"Mono C#\": \"C#\",\n+ \"GNU C\": \"C\",\n+ \"Python 3\": \"Python\",\n+ \"PyPy 3\": \"Python\",\n+ \"GNU C11\": \"C\",\n+ \"Go\": \"Go\",\n+ \"Rust\": \"Rust\",\n+ \"PyPy 2\": \"Python\",\n+ \"Python 2\": \"Python\",\n+ \"MS C#\": \"C#\",\n+ \"Kotlin\": \"Kotlin\",\n+ \"GNU C++0x\": \"C++\",\n+ \"Java 7\": \"Java\",\n+ \"Node.js\": \"Javascript\",\n+ \".NET Core C#\": \"C#\",\n+ \"PHP\": \"PHP\",\n+ \"GNU C++17 Diagnostics\": \"C++\",\n+ \"Clang++17 Diagnostics\": \"C++\",\n+ \"JavaScript\": \"Javascript\",\n+ \"Ruby\": \"Ruby\",\n+ \"C# 10\": \"C#\",\n+ \"C# 8\": \"C#\",\n+ \"Clang++20 Diagnostics\": \"C++\",\n+ \"GNU C++17 (64)\": \"C++\",\n+ \"GNU C++20 (64)\": \"C++\",\n+ \"Java 17\": \"Java\",\n+ \"Kotlin 1.4\": \"Kotlin\",\n+ \"Kotlin 1.5\": \"Kotlin\",\n+ \"Kotlin 1.6\": \"Kotlin\",\n+ \"Kotlin 1.7\": \"Kotlin\",\n+ \"PyPy 3-64\": \"Python\",\n+ \"Python 3 + libs\": \"Python\",\n+ \"Ruby 3\": \"Ruby\",\n+ \"Rust 2021\": \"Rust\",\n+}\n+\n+LANGS = sorted(set([v for k, v in SHORT_LANG_MAP.items()]))\n+\n+\n+openai.api_key = os.environ[\"OPENAI_API_KEY\"]\n+\n+\n+def gen(prompt, temperature, nsample):\n+ cnt = 0\n+ while True:\n+ if cnt == 999:\n+ return None\n+ try:\n+ c = openai.ChatCompletion.create(\n+ model=\"gpt-3.5-turbo\",\n+ messages=[\n+ {\"role\": \"user\", \"content\": f\"{prompt}\"},\n+ ],\n+ temperature=temperature,\n+ top_p=1,\n+ n=nsample,\n+ frequency_penalty=0.0,\n+ presence_penalty=0.0,\n+ )\n+ break\n+ except Exception as e:\n+ cnt += 1\n+ time.sleep(5)\n+ print(f\"{e}\")\n+ c[\"prompt\"] = prompt\n+ return c\n+\n+\n+xcodeeval_prompt_template = {\n+ \"program_synthesis\": [\n+ \"Write a program in {{lang_cluster}} to solve this programming problem:\\nDescription: {{prob_desc_description}}\\nInput Specification: {{prob_desc_input_spec}}\\nOutput Specification: {{prob_desc_output_spec}}\\n{% for input, output in zip(prob_desc_sample_inputs, prob_desc_sample_outputs) %}\\nSample Input:\\n{{input}}\\nSample Output:\\n{{output}}\\n{% endfor %}\\nNotes: {{prob_desc_notes}}\\nTake input from {{prob_desc_input_from}} and output to {{prob_desc_output_to}}\\nProvide the {{lang_cluster}} code without any extra description or tokens. Target code: ||END-of-SRC|| \",\n+ ]\n+}\n+\n+\n+def process_prompt(\n+ dt, temperature, nsample, language, template, output_dir, index, dry_run=0\n+):\n+ file_path = os.path.join(output_dir, f\"{index}_{temperature}_{language}.json\")\n+ if not os.path.exists(file_path):\n+ dt[\"lang_cluster\"] = language\n+ dt[\"prob_desc_sample_inputs\"] = json.loads(dt[\"prob_desc_sample_inputs\"])\n+ dt[\"prob_desc_sample_outputs\"] = json.loads(dt[\"prob_desc_sample_outputs\"])\n+ lm_io = template.apply(dt)\n+ assert len(lm_io) == 2, f\"{json.dumps(lm_io, indent=4)}\"\n+ if dry_run:\n+ open(file_path, \"w\").write(f\"{json.dumps(lm_io[0], indent=4)}\")\n+ else:\n+ out = gen(lm_io[0], temperature, nsample)\n+ export_data = {\"oai_response\": out, \"source_data\": dt}\n+ open(file_path, \"w\").write(f\"{json.dumps(export_data, indent=4)}\")\n+\n+\n+def main():\n+ parser = argparse.ArgumentParser()\n+ parser.add_argument(\n+ \"--output-dir\",\n+ default=\"dumped/oai/program_synthesis_n_sample_20\",\n+ help=\"Output Folder to save the API request.\",\n+ )\n+ parser.add_argument(\n+ \"--num-proc\",\n+ default=1,\n+ help=\"Number of parallel API request.\",\n+ )\n+ parser.add_argument(\n+ \"--dry-run\",\n+ default=0,\n+ help=\"Number of parallel API request.\",\n+ )\n+ parser.add_argument(\n+ \"--nsample\",\n+ default=20,\n+ type=int,\n+ help=\"Number of parallel API request.\",\n+ )\n+ args = parser.parse_args()\n+ if not os.path.exists(args.output_dir):\n+ os.makedirs(args.output_dir, exist_ok=True)\n+ templates = [\n+ Template(f\"prog_syn_{idx}\", template, \"xCodeEval\", delimeter=\"||END-of-SRC||\")\n+ for idx, template in enumerate(xcodeeval_prompt_template[\"program_synthesis\"])\n+ ]\n+ template = templates[0]\n+\n+ prog_synthesis_dataset = datasets.load_dataset(\n+ \"NTU-NLP-sg/xCodeEval\", \"program_synthesis\", num_proc=16, trust_remote_code=True\n+ )[\"compact\"]\n+ # temperature_list = np.linspace(0, 2, args.nsample)\n+ temperature_list = [0.3157894736842105]\n+ for language in LANGS:\n+ with concurrent.futures.ProcessPoolExecutor(\n+ max_workers=int(args.num_proc)\n+ ) as executor:\n+ futures = []\n+ for idx, dt in tqdm.tqdm(\n+ enumerate(prog_synthesis_dataset),\n+ total=len(prog_synthesis_dataset),\n+ desc=f\"Preparing samples {language} lang\",\n+ ):\n+ for temperature in temperature_list:\n+ future = executor.submit(\n+ process_prompt,\n+ dt,\n+ temperature,\n+ args.nsample,\n+ language,\n+ template,\n+ args.output_dir,\n+ idx,\n+ args.dry_run,\n+ )\n+ futures.append(future)\n+\n+ for future in tqdm.tqdm(\n+ concurrent.futures.as_completed(futures),\n+ total=len(futures),\n+ desc=f\"Calling OpenAI API for {language} lang\",\n+ ):\n+ try:\n+ future.result()\n+ except Exception as e:\n+ print(f\"Error occurred: {e}\")\n+\n+\n+if __name__ == \"__main__\":\n+ main()" |
| }, |
| { |
| "sha": "092734b8263800290b6bb11605d5f858e28db3a3", |
| "filename": "evaluation/program_synthesis/get_result.py", |
| "status": "added", |
| "additions": 105, |
| "deletions": 0, |
| "changes": 105, |
| "blob_url": "https://github.com/ntunlp/xCodeEval/blob/4993b99b6a97e5e19cfa7d4cd75e12545580ad27/evaluation%2Fprogram_synthesis%2Fget_result.py", |
| "raw_url": "https://github.com/ntunlp/xCodeEval/raw/4993b99b6a97e5e19cfa7d4cd75e12545580ad27/evaluation%2Fprogram_synthesis%2Fget_result.py", |
| "contents_url": "https://api.github.com/repos/ntunlp/xCodeEval/contents/evaluation%2Fprogram_synthesis%2Fget_result.py?ref=4993b99b6a97e5e19cfa7d4cd75e12545580ad27", |
| "patch": "@@ -0,0 +1,105 @@\n+import os\n+from collections import defaultdict\n+import tqdm\n+import jsonlines\n+from typing import List, Union\n+import itertools\n+import numpy as np\n+\n+LANG_CLUSTER_TO_LANG_COMPILER = {\n+ \"C\": \"GNU C11\",\n+ \"C#\": \"Mono C#\",\n+ \"C++\": \"GNU C++17\",\n+ \"Go\": \"Go\",\n+ \"Java\": \"Java 17\",\n+ \"Javascript\": \"Node.js\",\n+ \"Kotlin\": \"Kotlin 1.4\",\n+ \"PHP\": \"PHP\",\n+ \"Python\": \"PyPy 3\",\n+ \"Ruby\": \"Ruby 3\",\n+ \"Rust\": \"Rust 2018\",\n+}\n+\n+path = f'{os.environ[\"DUMP_FOLDER\"]}/oai/prog_synthesis_n_sample_20/'\n+output_path = os.path.join(\n+ path, \"reproduce_1\"\n+) # \"eval_program_synthesis_val_execeval_fixtemp_nsampling_20_stop_at_first_fail_true\",\n+ks = range(1, 21)\n+\n+\n+def estimate_pass_at_k(\n+ num_samples: Union[int, List[int], np.ndarray],\n+ num_correct: Union[List[int], np.ndarray],\n+ k: int,\n+) -> np.ndarray:\n+ \"\"\"\n+ Estimates pass@k of each problem and returns them in an array.\n+ \"\"\"\n+\n+ def estimator(n: int, c: int, k: int):\n+ \"\"\"\n+ Calculates 1 - comb(n - c, k) / comb(n, k).\n+ \"\"\"\n+ if n - c < k:\n+ return 1.0\n+ return 1.0 - np.prod(1.0 - k / np.arange(n - c + 1, n + 1))\n+\n+ if isinstance(num_samples, int):\n+ num_samples_it = itertools.repeat(num_samples, len(num_correct))\n+ else:\n+ assert len(num_samples) == len(num_correct)\n+ num_samples_it = iter(num_samples)\n+\n+ return np.array(\n+ [estimator(int(n), int(c), k) for n, c in zip(num_samples_it, num_correct)]\n+ )\n+\n+\n+def get_execeval_out_file_name(compiler):\n+ return os.path.join(output_path, f\"{compiler}.jsonl\")\n+\n+\n+# construct result as {[task_id]: [unit_test_results]}\n+# task_id will be src_uid_lang\n+\n+pass_at_k = defaultdict(dict)\n+\n+for lang, compiler in tqdm.tqdm(LANG_CLUSTER_TO_LANG_COMPILER.items()):\n+ execeval_out_file = get_execeval_out_file_name(compiler)\n+ results = defaultdict(list)\n+ with jsonlines.open(execeval_out_file) as jrp:\n+ for sample in jrp:\n+ src_uid = sample[\"source_data\"][\"src_uid\"]\n+ task_id = f\"{src_uid}|||{lang}\"\n+ for ut_res in sample[\"unit_test_results\"]:\n+ if \"error\" in ut_res:\n+ continue\n+ results[task_id].append(ut_res)\n+\n+ total, correct = [], []\n+ for result in results.values():\n+ passed = [\n+ all(x[\"exec_outcome\"] == \"PASSED\" for x in ut_res) for ut_res in result\n+ ]\n+ total.append(len(passed))\n+ correct.append(sum(passed))\n+ total = np.array(total)\n+ correct = np.array(correct)\n+\n+ pass_at_k[lang] = {\n+ f\"pass@{k}\": estimate_pass_at_k(total, correct, k).mean()\n+ for k in ks\n+ if (total >= k).all()\n+ }\n+\n+\n+langs = sorted(list(pass_at_k.keys()))\n+for lang in langs:\n+ print(f\" & {lang}\", end=\"\")\n+print()\n+avg = 0\n+for lang in langs:\n+ print(f\" & {round(pass_at_k[lang]['pass@5']*100, 2)}\", end=\"\")\n+ avg += pass_at_k[lang][\"pass@5\"] * 100\n+avg /= len(langs)\n+print(f\" & {round(avg, 2)}\")" |
| }, |
| { |
| "sha": "54e05d499213b06f8b203a0530968b4218beca54", |
| "filename": "requirement.txt", |
| "status": "added", |
| "additions": 89, |
| "deletions": 0, |
| "changes": 89, |
| "blob_url": "https://github.com/ntunlp/xCodeEval/blob/4993b99b6a97e5e19cfa7d4cd75e12545580ad27/requirement.txt", |
| "raw_url": "https://github.com/ntunlp/xCodeEval/raw/4993b99b6a97e5e19cfa7d4cd75e12545580ad27/requirement.txt", |
| "contents_url": "https://api.github.com/repos/ntunlp/xCodeEval/contents/requirement.txt?ref=4993b99b6a97e5e19cfa7d4cd75e12545580ad27", |
| "patch": "@@ -0,0 +1,89 @@\n+aiohappyeyeballs==2.4.0\n+aiohttp==3.10.5\n+aiosignal==1.3.1\n+altair==5.4.1\n+astor==0.8.1\n+async-timeout==4.0.3\n+attrs==24.2.0\n+backports.zoneinfo==0.2.1\n+base58==2.1.1\n+black==21.12b0\n+blinker==1.8.2\n+Brotli==1.1.0\n+cachetools==5.5.0\n+certifi==2024.8.30\n+charset-normalizer==3.3.2\n+click==7.1.2\n+datasets==2.16.1\n+dill==0.3.8\n+exceptiongroup==1.2.2\n+filelock==3.16.0\n+flake8==7.1.1\n+frozenlist==1.4.1\n+fsspec==2024.6.1\n+gitdb==4.0.11\n+GitPython==3.1.43\n+huggingface-hub==0.25.0\n+idna==3.10\n+importlib_resources==6.4.5\n+inflate64==1.0.0\n+iniconfig==2.0.0\n+isort==5.8.0\n+Jinja2==3.1.4\n+jsonlines==4.0.0\n+jsonschema==4.23.0\n+jsonschema-specifications==2023.12.1\n+MarkupSafe==2.1.5\n+mccabe==0.7.0\n+multidict==6.1.0\n+multiprocess==0.70.16\n+multivolumefile==0.2.3\n+mypy-extensions==1.0.0\n+narwhals==1.8.1\n+numpy==1.24.4\n+openai==0.28.0\n+packaging==24.1\n+pandas==2.0.3\n+pathspec==0.12.1\n+pillow==10.4.0\n+pkgutil_resolve_name==1.3.10\n+platformdirs==4.3.3\n+plotly==5.24.1\n+pluggy==1.5.0\n+promptsource @ git+https://github.com/sbmaruf/promptsource@70dc08cf37b6483765382de7f75db5906ca0d742\n+protobuf==5.28.1\n+psutil==6.0.0\n+py7zr==0.22.0\n+pyarrow==17.0.0\n+pybcj==1.0.2\n+pycodestyle==2.12.1\n+pycryptodomex==3.20.0\n+pydeck==0.9.1\n+pyflakes==3.2.0\n+pyppmd==1.1.0\n+pytest==8.3.3\n+python-dateutil==2.9.0.post0\n+pytz==2024.2\n+PyYAML==6.0.2\n+pyzstd==0.16.1\n+referencing==0.35.1\n+requests==2.32.3\n+rpds-py==0.20.0\n+six==1.16.0\n+smmap==5.0.1\n+streamlit==0.82.0\n+tenacity==9.0.0\n+texttable==1.7.0\n+toml==0.10.2\n+tomli==1.2.3\n+tornado==6.4.1\n+tqdm==4.66.5\n+typing_extensions==4.12.2\n+tzdata==2024.1\n+tzlocal==5.2\n+urllib3==2.2.3\n+validators==0.34.0\n+watchdog==4.0.2\n+xxhash==3.5.0\n+yarl==1.11.1\n+zipp==3.20.2" |
| } |
| ] |
| } |
|
|
