File size: 5,115 Bytes
7db0ae4 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 |
import sys, os
import traceback
sys.path.insert(
0, os.path.abspath("../..")
) # Adds the parent directory to the system path
import time
import litellm
from litellm import get_max_tokens, model_cost, open_ai_chat_completion_models
import pytest
def test_get_gpt3_tokens():
max_tokens = get_max_tokens("gpt-3.5-turbo")
print(max_tokens)
assert max_tokens == 4097
# print(results)
# test_get_gpt3_tokens()
def test_get_palm_tokens():
# # 🦄🦄🦄🦄🦄🦄🦄🦄
max_tokens = get_max_tokens("palm/chat-bison")
assert max_tokens == 4096
print(max_tokens)
# test_get_palm_tokens()
def test_zephyr_hf_tokens():
max_tokens = get_max_tokens("huggingface/HuggingFaceH4/zephyr-7b-beta")
print(max_tokens)
assert max_tokens == 32768
# test_zephyr_hf_tokens()
def test_cost_ft_gpt_35():
try:
# this tests if litellm.completion_cost can calculate cost for ft:gpt-3.5-turbo:my-org:custom_suffix:id
# it needs to lookup ft:gpt-3.5-turbo in the litellm model_cost map to get the correct cost
from litellm import ModelResponse, Choices, Message
from litellm.utils import Usage
resp = ModelResponse(
id="chatcmpl-e41836bb-bb8b-4df2-8e70-8f3e160155ac",
choices=[
Choices(
finish_reason=None,
index=0,
message=Message(
content=" Sure! Here is a short poem about the sky:\n\nA canvas of blue, a",
role="assistant",
),
)
],
created=1700775391,
model="ft:gpt-3.5-turbo:my-org:custom_suffix:id",
object="chat.completion",
system_fingerprint=None,
usage=Usage(prompt_tokens=21, completion_tokens=17, total_tokens=38),
)
cost = litellm.completion_cost(completion_response=resp)
print("\n Calculated Cost for ft:gpt-3.5", cost)
input_cost = model_cost["ft:gpt-3.5-turbo"]["input_cost_per_token"]
output_cost = model_cost["ft:gpt-3.5-turbo"]["output_cost_per_token"]
print(input_cost, output_cost)
expected_cost = (input_cost * resp.usage.prompt_tokens) + (
output_cost * resp.usage.completion_tokens
)
print("\n Excpected cost", expected_cost)
assert cost == expected_cost
except Exception as e:
pytest.fail(
f"Cost Calc failed for ft:gpt-3.5. Expected {expected_cost}, Calculated cost {cost}"
)
# test_cost_ft_gpt_35()
def test_cost_azure_gpt_35():
try:
# this tests if litellm.completion_cost can calculate cost for azure/chatgpt-deployment-2 which maps to azure/gpt-3.5-turbo
# for this test we check if passing `model` to completion_cost overrides the completion cost
from litellm import ModelResponse, Choices, Message
from litellm.utils import Usage
resp = ModelResponse(
id="chatcmpl-e41836bb-bb8b-4df2-8e70-8f3e160155ac",
choices=[
Choices(
finish_reason=None,
index=0,
message=Message(
content=" Sure! Here is a short poem about the sky:\n\nA canvas of blue, a",
role="assistant",
),
)
],
model="gpt-35-turbo", # azure always has model written like this
usage=Usage(prompt_tokens=21, completion_tokens=17, total_tokens=38),
)
cost = litellm.completion_cost(
completion_response=resp, model="azure/gpt-35-turbo"
)
print("\n Calculated Cost for azure/gpt-3.5-turbo", cost)
input_cost = model_cost["azure/gpt-35-turbo"]["input_cost_per_token"]
output_cost = model_cost["azure/gpt-35-turbo"]["output_cost_per_token"]
expected_cost = (input_cost * resp.usage.prompt_tokens) + (
output_cost * resp.usage.completion_tokens
)
print("\n Excpected cost", expected_cost)
assert cost == expected_cost
except Exception as e:
pytest.fail(
f"Cost Calc failed for azure/gpt-3.5-turbo. Expected {expected_cost}, Calculated cost {cost}"
)
test_cost_azure_gpt_35()
def test_cost_azure_embedding():
try:
import asyncio
litellm.set_verbose = True
async def _test():
response = await litellm.aembedding(
model="azure/azure-embedding-model",
input=["good morning from litellm", "gm"],
)
print(response)
return response
response = asyncio.run(_test())
cost = litellm.completion_cost(completion_response=response)
print("Cost", cost)
expected_cost = float("7e-07")
assert cost == expected_cost
except Exception as e:
pytest.fail(
f"Cost Calc failed for azure/gpt-3.5-turbo. Expected {expected_cost}, Calculated cost {cost}"
)
# test_cost_azure_embedding()
|