mirror of
https://github.com/PaddlePaddle/FastDeploy.git
synced 2025-10-04 08:16:42 +08:00
fix topp default value (#2814)
Some checks failed
Deploy GitHub Pages / deploy (push) Has been cancelled
Some checks failed
Deploy GitHub Pages / deploy (push) Has been cancelled
This commit is contained in:
@@ -23,6 +23,7 @@ import paddle.nn as nn
|
||||
|
||||
from fastdeploy.config import FDConfig
|
||||
from fastdeploy.engine.request import Request
|
||||
from fastdeploy.model_executor.forward_meta import ForwardMeta, XPUForwardMeta
|
||||
from fastdeploy.model_executor.layers.attention import get_attention_backend
|
||||
from fastdeploy.model_executor.layers.attention.base_attention_backend import \
|
||||
AttentionBackend
|
||||
@@ -31,7 +32,6 @@ from fastdeploy.model_executor.layers.sample.meta_data import SamplingMetadata
|
||||
from fastdeploy.model_executor.layers.sample.sampler import Sampler
|
||||
from fastdeploy.model_executor.model_loader import get_model_from_loader
|
||||
from fastdeploy.utils import get_logger
|
||||
from fastdeploy.model_executor.forward_meta import ForwardMeta, XPUForwardMeta
|
||||
from fastdeploy.worker.model_runner_base import ModelRunnerBase
|
||||
from fastdeploy.worker.output import ModelOutputData, ModelRunnerOutput
|
||||
|
||||
@@ -295,7 +295,7 @@ class XPUModelRunner(ModelRunnerBase):
|
||||
self.share_inputs["eos_token_id"][:] = np.array(
|
||||
request.eos_token_ids, dtype="int64").reshape(-1, 1)
|
||||
self.share_inputs["pre_ids"][idx:idx + 1] = -1
|
||||
self.share_inputs["top_p"][idx:idx + 1] = request.get("top_p", 1.0)
|
||||
self.share_inputs["top_p"][idx:idx + 1] = request.get("top_p", 0.7)
|
||||
self.share_inputs["top_k"][idx:idx + 1] = request.get("top_k", 0)
|
||||
self.share_inputs["temperature"][idx:idx + 1] = request.get(
|
||||
"temperature", 0.95)
|
||||
|
Reference in New Issue
Block a user