scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-10-09 03:18:44 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 109

Update test_llm_cost_calc_utils.py

Browse source
This commit is contained in:
Praveen11558 2026-03-23 20:52:10 +05:30 • committed by GitHub
parent f6fc7ee75a
commit d3de061d6f
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 17 additions and 8 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

21
tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py
View file

@ -1441,6 +1441,11 @@ def test_character_count_billing_does_not_fill_prompt_token_gap():
os.environ["LITELLM_LOCAL_MODEL_COST_MAP"] = "True" os.environ["LITELLM_LOCAL_MODEL_COST_MAP"] = "True"
litellm.model_cost = litellm.get_model_cost_map(url="") litellm.model_cost = litellm.get_model_cost_map(url="")
# multimodalembedding@001 has non-zero input_cost_per_character (2e-07),
# input_cost_per_token (8e-07), and output_cost_per_token (0) in the cost map,
# making the character_count billing assertion non-trivial.
model = "multimodalembedding@001"
usage = Usage( usage = Usage(
prompt_tokens=200, prompt_tokens=200,
completion_tokens=20, completion_tokens=20,
@ -1455,17 +1460,17 @@ def test_character_count_billing_does_not_fill_prompt_token_gap():
) )
prompt_cost, completion_cost = generic_cost_per_token( prompt_cost, completion_cost = generic_cost_per_token(
model="gemini-2.0-flash-001", model=model,
usage=usage, usage=usage,
custom_llm_provider="vertex_ai", custom_llm_provider="vertex_ai",
) )
model_info = litellm.model_cost["gemini-2.0-flash-001"] model_info = litellm.model_cost[model]
expected_prompt_cost = ( expected_prompt_cost = (
100 * model_info["input_cost_per_token"] 100 * model_info.get("input_cost_per_token", 0)
+ 1000 * model_info.get("input_cost_per_character", 0) + 1000 * model_info.get("input_cost_per_character", 0)
) )
expected_completion_cost = 20 * model_info["output_cost_per_token"] expected_completion_cost = 20 * model_info.get("output_cost_per_token", 0)
assert prompt_cost == pytest.approx(expected_prompt_cost), ( assert prompt_cost == pytest.approx(expected_prompt_cost), (
f"Expected prompt_cost={expected_prompt_cost}, got {prompt_cost}. " f"Expected prompt_cost={expected_prompt_cost}, got {prompt_cost}. "
@ -1483,6 +1488,10 @@ def test_video_length_billing_does_not_fill_prompt_token_gap():
os.environ["LITELLM_LOCAL_MODEL_COST_MAP"] = "True" os.environ["LITELLM_LOCAL_MODEL_COST_MAP"] = "True"
litellm.model_cost = litellm.get_model_cost_map(url="") litellm.model_cost = litellm.get_model_cost_map(url="")
# multimodalembedding@001 has non-zero input_cost_per_video_per_second (0.0005)
# and input_cost_per_token (8e-07), making the video billing assertion non-trivial.
model = "multimodalembedding@001"
usage = Usage( usage = Usage(
prompt_tokens=150, prompt_tokens=150,
completion_tokens=10, completion_tokens=10,
@ -1497,12 +1506,12 @@ def test_video_length_billing_does_not_fill_prompt_token_gap():
) )
prompt_cost, completion_cost = generic_cost_per_token( prompt_cost, completion_cost = generic_cost_per_token(
model="gemini-2.0-flash-001", model=model,
usage=usage, usage=usage,
custom_llm_provider="vertex_ai", custom_llm_provider="vertex_ai",
) )
model_info = litellm.model_cost["gemini-2.0-flash-001"] model_info = litellm.model_cost[model]
expected_prompt_cost = ( expected_prompt_cost = (
50 * model_info.get("input_cost_per_token", 0) 50 * model_info.get("input_cost_per_token", 0)
+ 12.0 * model_info.get("input_cost_per_video_per_second", 0) + 12.0 * model_info.get("input_cost_per_video_per_second", 0)