apunkt/MetaGPT
1
0
Fork 0
mirror of https://github.com/FoundationAgents/MetaGPT.git synced 2026-05-01 03:46:23 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

update ensemble example

Browse source
This commit is contained in:
didi 2024-07-01 16:56:47 +08:00
parent f1ce1330d7
commit 4d376649cc
6 changed files with 53 additions and 230 deletions
Download patch file Download diff file Expand all files Collapse all files

23
he_test.py
View file

@ -6,9 +6,13 @@ from examples.ags.w_action_node.graph import HumanEvalGraph
from examples.ags.w_action_node.operator import GenerateCode
generate_code = GenerateCode(llm=LLM())
case = get_human_eval_plus()['HumanEval/10']
solver = HumanEvalGraph(name="solver", llm=LLM(), criteria='correctness, efficiency, readability')
async def sample_generate(case):
solution_result = await solver(case['prompt'])
sample_dict = dict(task_id=case['task_id'], solution=solution_result['final_solution'])
print(sample_dict)
async def samples_generate_sequence():
sample_list = []
@ -35,18 +39,6 @@ async def samples_generate_ags():
write_jsonl("samples.jsonl", sample_list)
# humanevalgraph result (review & revise -> ensemble)
# humaneval (base tests)
# pass@1: 0.823
# humaneval+ (base + extra tests)
# pass@1: 0.774
# deepseek result
# humaneval (base tests)
# pass@1: 0.841
# humaneval+ (base + extra tests)
# pass@1: 0.780
async def samples_generate_llm():
sample_list = []
cases = list(get_human_eval_plus().values())
@ -64,5 +56,8 @@ async def samples_generate_llm():
write_jsonl("samples.jsonl", sample_list)
asyncio.run(samples_generate_llm())
# asyncio.run(sample_generate(case))
# asyncio.run(samples_generate_llm())
asyncio.run(samples_generate_ags())