forked from phoenix/litellm-mirror
(fix) azure_base_model cost calc
This commit is contained in:
parent
6969b25946
commit
cc7a690c9b
1 changed files with 7 additions and 6 deletions
|
@ -269,14 +269,15 @@ class CompletionCustomHandler(
|
||||||
assert isinstance(kwargs["model"], str)
|
assert isinstance(kwargs["model"], str)
|
||||||
|
|
||||||
# checking we use base_model for azure cost calculation
|
# checking we use base_model for azure cost calculation
|
||||||
base_model = (
|
base_model = litellm.utils._get_base_model_from_metadata(
|
||||||
kwargs.get("litellm_params", {})
|
model_call_details=kwargs
|
||||||
.get("metadata", {})
|
|
||||||
.get("model_info", {})
|
|
||||||
.get("base_model", None)
|
|
||||||
)
|
)
|
||||||
|
|
||||||
if kwargs["model"] == "chatgpt-v-2" and base_model is not None:
|
if (
|
||||||
|
kwargs["model"] == "chatgpt-v-2"
|
||||||
|
and base_model is not None
|
||||||
|
and kwargs["stream"] != True
|
||||||
|
):
|
||||||
# when base_model is set for azure, we should use pricing for the base_model
|
# when base_model is set for azure, we should use pricing for the base_model
|
||||||
# this checks response_cost == litellm.cost_per_token(model=base_model)
|
# this checks response_cost == litellm.cost_per_token(model=base_model)
|
||||||
assert isinstance(kwargs["response_cost"], float)
|
assert isinstance(kwargs["response_cost"], float)
|
||||||
|
|
Loading…
Add table
Add a link
Reference in a new issue