diff --git a/training/kunlunxin/docker_image/pytorch/pytorch_install.sh b/training/kunlunxin/docker_image/pytorch/pytorch_install.sh
index 850a304b4..2a96fe267 100644
--- a/training/kunlunxin/docker_image/pytorch/pytorch_install.sh
+++ b/training/kunlunxin/docker_image/pytorch/pytorch_install.sh
@@ -4,3 +4,5 @@ set -xe
 
 pip install https://bd.bcebos.com/klx-pytorch-ipipe-bd/flagperf/latest/xacc-0.1.0-cp38-cp38-linux_x86_64.whl
 pip install https://bd.bcebos.com/klx-pytorch-ipipe-bd/flagperf/latest/xmlir-0.0.1-cp38-cp38-linux_x86_64.whl
+
+python -m xacc.install
diff --git a/training/kunlunxin/glm-pytorch/config/config_R300x1x1.py b/training/kunlunxin/glm-pytorch/config/config_R300x1x1.py
index 8a4c96915..cd9afdd40 100644
--- a/training/kunlunxin/glm-pytorch/config/config_R300x1x1.py
+++ b/training/kunlunxin/glm-pytorch/config/config_R300x1x1.py
@@ -14,4 +14,4 @@
 lr_decay_iters = 4338
 log_freq = 1
 seed = 4096
-max_samples_termination = 925510
+max_samples_termination = 4000
diff --git a/training/kunlunxin/glm-pytorch/config/config_R300x2x8.py b/training/kunlunxin/glm-pytorch/config/config_R300x2x8.py
index 840259660..fa79a403e 100644
--- a/training/kunlunxin/glm-pytorch/config/config_R300x2x8.py
+++ b/training/kunlunxin/glm-pytorch/config/config_R300x2x8.py
@@ -14,4 +14,4 @@
 lr_decay_iters = 4338
 log_freq = 1
 seed = 4096
-max_samples_termination = 2776540
+max_samples_termination = 20000
diff --git a/training/kunlunxin/glm-pytorch/config/environment_variables.sh b/training/kunlunxin/glm-pytorch/config/environment_variables.sh
index 9c9f20b8e..b527e7873 100755
--- a/training/kunlunxin/glm-pytorch/config/environment_variables.sh
+++ b/training/kunlunxin/glm-pytorch/config/environment_variables.sh
@@ -17,3 +17,5 @@ export ALLREDUCE_FUSION=0
 
 export XMLIR_F_XPU_FC_GEMM_MODE=float
 export XMLIR_F_FAST_INDEX_PUT=true
+
+export XACC_ENABLE=1
diff --git a/training/kunlunxin/glm-pytorch/config/requirements.txt b/training/kunlunxin/glm-pytorch/config/requirements.txt
index 8bac0066c..46109702b 100644
--- a/training/kunlunxin/glm-pytorch/config/requirements.txt
+++ b/training/kunlunxin/glm-pytorch/config/requirements.txt
@@ -2,3 +2,4 @@ h5sparse
 boto3
 h5py
 numpy>=1.15.4
+psutil
diff --git a/training/kunlunxin/glm-pytorch/extern/trainer_adapter.py b/training/kunlunxin/glm-pytorch/extern/trainer_adapter.py
index 3ac97e41a..719532ede 100644
--- a/training/kunlunxin/glm-pytorch/extern/trainer_adapter.py
+++ b/training/kunlunxin/glm-pytorch/extern/trainer_adapter.py
@@ -1,13 +1,13 @@
 import torch
-import config
-
 from torch import nn
 import torch.distributed as dist
 
+import config
 from optimizers import get_optimizer_param_groups
 from optimizers.loss_scaler import DynamicLossScaler
 from driver.dist_pytorch import main_proc_print
 
+import torch_xmlir
 import torch_xmlir.core.xpu_model as xm
 from torch_xmlir.optimizer import AdamW as Adam
 from torch_xmlir.nn.clip_grad import clip_grad_norm
@@ -79,4 +79,5 @@ def _clip_grad():
     if DynamicLossScaler._has_inf_or_nan(reduced_loss):
         main_proc_print("Found NaN loss, skip backward")
 
+    torch_xmlir.xpu.empty_cache()
     return reduced_loss