We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
There was an error while loading. Please reload this page.
1 parent 9d3551c commit e4a4573Copy full SHA for e4a4573
1 file changed
fastdeploy/model_executor/pre_and_post_process.py
@@ -22,10 +22,6 @@
22
23
from fastdeploy import envs
24
from fastdeploy.config import SpeculativeConfig
25
-from fastdeploy.model_executor.ops.gpu import (
26
- mtp_save_first_token,
27
- mtp_save_first_token_with_topk,
28
-)
29
from fastdeploy.platforms import current_platform
30
from fastdeploy.worker.input_batch import (
31
InputBatch,
@@ -105,6 +101,8 @@
105
101
limit_thinking_content_length,
106
102
speculate_limit_thinking_content_length,
107
103
custom_numpy_to_tensor,
104
+ mtp_save_first_token,
+ mtp_save_first_token_with_topk,
108
)
109
110
from fastdeploy.model_executor.entropy_utils import (
0 commit comments