mirror of
https://github.com/xtekky/gpt4free.git
synced 2024-11-19 03:25:32 +00:00
68 lines
1.6 KiB
Python
68 lines
1.6 KiB
Python
import asyncio
|
|
import sys
|
|
from pathlib import Path
|
|
sys.path.append(str(Path(__file__).parent.parent.parent))
|
|
|
|
import g4f
|
|
|
|
|
|
async def test(model: g4f.Model):
|
|
try:
|
|
try:
|
|
for response in g4f.ChatCompletion.create(
|
|
model=model,
|
|
messages=[{"role": "user", "content": "write a poem about a tree"}],
|
|
temperature=0.1,
|
|
stream=True
|
|
):
|
|
print(response, end="")
|
|
|
|
print()
|
|
except:
|
|
for response in await g4f.ChatCompletion.create_async(
|
|
model=model,
|
|
messages=[{"role": "user", "content": "write a poem about a tree"}],
|
|
temperature=0.1,
|
|
stream=True
|
|
):
|
|
print(response, end="")
|
|
|
|
print()
|
|
|
|
return True
|
|
except Exception as e:
|
|
print(model.name, "not working:", e)
|
|
print(e.__traceback__.tb_next)
|
|
return False
|
|
|
|
|
|
async def start_test():
|
|
models_to_test = [
|
|
# GPT-3.5 4K Context
|
|
g4f.models.gpt_35_turbo,
|
|
g4f.models.gpt_35_turbo_0613,
|
|
|
|
# GPT-3.5 16K Context
|
|
g4f.models.gpt_35_turbo_16k,
|
|
g4f.models.gpt_35_turbo_16k_0613,
|
|
|
|
# GPT-4 8K Context
|
|
g4f.models.gpt_4,
|
|
g4f.models.gpt_4_0613,
|
|
|
|
# GPT-4 32K Context
|
|
g4f.models.gpt_4_32k,
|
|
g4f.models.gpt_4_32k_0613,
|
|
]
|
|
|
|
models_working = []
|
|
|
|
for model in models_to_test:
|
|
if await test(model):
|
|
models_working.append(model.name)
|
|
|
|
print("working models:", models_working)
|
|
|
|
|
|
asyncio.run(start_test())
|