[Bugfix] Account for truncate_prompt_tokens when computing max_tokens (#41800)
Signed-off-by: Viktor Pus <viktorpus@tenstorrent.com>
This commit is contained in:
@@ -164,6 +164,7 @@ class OpenAIServingRender:
|
||||
input_length,
|
||||
self.default_sampling_params,
|
||||
self.override_max_tokens,
|
||||
truncate_prompt_tokens=request.truncate_prompt_tokens,
|
||||
)
|
||||
params = request.to_sampling_params(max_tokens, self.default_sampling_params)
|
||||
|
||||
@@ -298,6 +299,7 @@ class OpenAIServingRender:
|
||||
input_length,
|
||||
self.default_sampling_params,
|
||||
self.override_max_tokens,
|
||||
truncate_prompt_tokens=request.truncate_prompt_tokens,
|
||||
)
|
||||
params = request.to_sampling_params(
|
||||
max_tokens, self.default_sampling_params
|
||||
|
||||
Reference in New Issue
Block a user