phoenix/litellm
1
0
Fork 0
forked from phoenix/litellm-mirror
Code Pull requests Activity

test(test_keys.py): use correct model name for token counting

Browse source
This commit is contained in:
Krrish Dholakia 2024-01-23 17:46:14 -08:00
parent 4ca4913468
commit d6844f43c8
3 changed files with 24 additions and 11 deletions
Download patch file Download diff file Expand all files Collapse all files

2
litellm/proxy/utils.py
View file

@ -556,7 +556,7 @@ class PrismaClient:
where={"token": token}, # type: ignore
data={**db_data}, # type: ignore
)
print_verbose(
verbose_proxy_logger.debug(
"\033[91m"
+ f"DB Token Table update succeeded {response}"
+ "\033[0m"

22
litellm/utils.py
View file

@ -2938,17 +2938,25 @@ def cost_per_token(
)
return prompt_tokens_cost_usd_dollar, completion_tokens_cost_usd_dollar
elif model_with_provider in model_cost_ref:
print_verbose(f"Looking up model={model_with_provider} in model_cost_map")
verbose_logger.debug(
f"Looking up model={model_with_provider} in model_cost_map"
)
verbose_logger.debug(
f"applying cost={model_cost_ref[model_with_provider]['input_cost_per_token']} for prompt_tokens={prompt_tokens}"
)
prompt_tokens_cost_usd_dollar = (
model_cost_ref[model_with_provider]["input_cost_per_token"] * prompt_tokens
)
verbose_logger.debug(
f"applying cost={model_cost_ref[model_with_provider]['output_cost_per_token']} for completion_tokens={completion_tokens}"
)
completion_tokens_cost_usd_dollar = (
model_cost_ref[model_with_provider]["output_cost_per_token"]
* completion_tokens
)
return prompt_tokens_cost_usd_dollar, completion_tokens_cost_usd_dollar
elif "ft:gpt-3.5-turbo" in model:
print_verbose(f"Cost Tracking: {model} is an OpenAI FinteTuned LLM")
verbose_logger.debug(f"Cost Tracking: {model} is an OpenAI FinteTuned LLM")
# fuzzy match ft:gpt-3.5-turbo:abcd-id-cool-litellm
prompt_tokens_cost_usd_dollar = (
model_cost_ref["ft:gpt-3.5-turbo"]["input_cost_per_token"] * prompt_tokens
@ -2959,17 +2967,23 @@ def cost_per_token(
)
return prompt_tokens_cost_usd_dollar, completion_tokens_cost_usd_dollar
elif model in litellm.azure_llms:
print_verbose(f"Cost Tracking: {model} is an Azure LLM")
verbose_logger.debug(f"Cost Tracking: {model} is an Azure LLM")
model = litellm.azure_llms[model]
verbose_logger.debug(
f"applying cost={model_cost_ref[model]['input_cost_per_token']} for prompt_tokens={prompt_tokens}"
)
prompt_tokens_cost_usd_dollar = (
model_cost_ref[model]["input_cost_per_token"] * prompt_tokens
)
verbose_logger.debug(
f"applying cost={model_cost_ref[model]['output_cost_per_token']} for completion_tokens={completion_tokens}"
)
completion_tokens_cost_usd_dollar = (
model_cost_ref[model]["output_cost_per_token"] * completion_tokens
)
return prompt_tokens_cost_usd_dollar, completion_tokens_cost_usd_dollar
elif model in litellm.azure_embedding_models:
print_verbose(f"Cost Tracking: {model} is an Azure Embedding Model")
verbose_logger.debug(f"Cost Tracking: {model} is an Azure Embedding Model")
model = litellm.azure_embedding_models[model]
prompt_tokens_cost_usd_dollar = (
model_cost_ref[model]["input_cost_per_token"] * prompt_tokens

11
tests/test_keys.py
View file

@ -2,7 +2,7 @@
## Tests /key endpoints.
import pytest
import asyncio
import asyncio, time
import aiohttp
from openai import AsyncOpenAI
import sys, os
@ -95,11 +95,10 @@ async def chat_completion(session, key, model="gpt-4"):
async def chat_completion_streaming(session, key, model="gpt-4"):
client = AsyncOpenAI(api_key=key, base_url="http://0.0.0.0:4000")
messages = [
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": "Hello!"},
{"role": "system", "content": "You are a helpful assistant"},
{"role": "user", "content": f"Hello! {time.time()}"},
]
prompt_tokens = litellm.token_counter(model="gpt-35-turbo", messages=messages)
assert prompt_tokens == 19
data = {
"model": model,
"messages": messages,
@ -114,7 +113,7 @@ async def chat_completion_streaming(session, key, model="gpt-4"):
print(f"content: {content}")
completion_tokens = litellm.token_counter(
model="azure/gpt-35-turbo", text=content, count_response_tokens=True
model="gpt-35-turbo", text=content, count_response_tokens=True
)
return prompt_tokens, completion_tokens
@ -251,7 +250,7 @@ async def test_key_info_spend_values():
)
print(f"prompt_tokens: {prompt_tokens}, completion_tokens: {completion_tokens}")
prompt_cost, completion_cost = litellm.cost_per_token(
model="gpt-35-turbo",
model="azure/gpt-35-turbo",
prompt_tokens=prompt_tokens,
completion_tokens=completion_tokens,
)