(llm) [root@gpu-11-51 my_offline_packages]# python

Python 3.11.10 (main, Oct  3 2024, 07:29:13) [GCC 11.2.0] on linux

Type "help", "copyright", "credits" or "license" for more information.

>>> # 修复后的原始代码

>>> import torch

import torch_mlu

# 修复前（会失败）：

# a = torch.randn(3, 3, device='mlu')

# 修复后：

a = torch.randn(3, 3).mlu()  # 先在CPU创建，再移动

b = torch.ones(3, 3).mlu()   # 这个可以直接创建

c = a + b

print(f"成功！结果设备: {c.device}")

print(f"结果形状: {c.shape}")

print(f"结果值:\n{c.cpu()}")  # 移回CPU显示

>>> import torch_mlu

>>> 

>>> # 修复前（会失败）：

>>> # a = torch.randn(3, 3, device='mlu')

>>> 

>>> # 修复后：

>>> a = torch.randn(3, 3).mlu()  # 先在CPU创建，再移动

>>> b = torch.ones(3, 3).mlu()   # 这个可以直接创建

>>> 

>>> c = a + b

[2026-01-13 14:11:50.335284][CNNL][ERROR][923993][Card:0]: [cnnlCreate] An internal error occured. When convert isa version to device failed. return unknown device. isa version = 372

2026-01-13 14:11:50.335693: [cnrtError] [923993] [Card: 0] cnrtInvokeKernel: Found kernel(_Z23MLUOpTensorElementBlockIL23cnnlOpTensorParamMode_t0EfffffEvPKT0_PKT1_PT2_T4_S9_S9_18cnnlOpTensorDesc_t8DataInfo) but not load(632015).

[2026-01-13 14:11:50.335724][CNNL][ERROR][923993][Card:0]: Check failed: Found failed to call the driver-api function. after invoke kernel (MLUOpTensorElementBlock<Param110, float, float, float, float, float> <<<kDim, kType, handle->queue>>>((float *)a_input, (float *)b_input, (float *)c, *(float *)alpha1_v, *(float *)alpha2_v, *(float *)beta_v, op_type, data_info))

[2026-01-13 14:11:50.335763][CNNL][ERROR][923993][Card:0]: [cnnlOpTensor] An internal error occured. 

Traceback (most recent call last):

  File "<stdin>", line 1, in <module>

RuntimeError: CNNL error: CNNL_STATUS_INTERNAL_ERROR

>>> print(f"成功！结果设备: {c.device}")