|
关于“模型训练-Model Training”类别
|
|
0
|
46
|
June 5, 2025
|
|
模型转换失败Convert failed. Ret: Common error code.
|
|
1
|
98
|
April 18, 2026
|
|
MindSpore训练自定义Transformer模型时报错"RuntimeError: Memory allocation failed"
|
|
1
|
97
|
April 18, 2026
|
|
MindSpore训练模型时报错”RuntimeError: Memory allocation failed”分析和解决
|
|
0
|
81
|
April 18, 2026
|
|
MindSpore动态图模式下梯度计算报错AttributeError: module 'mindspore' has no attribute 'value_and_grad'
|
|
1
|
80
|
March 1, 2026
|
|
MindSpore分布式训练Qwen-7B时崩溃Graph execution failed.
|
|
1
|
58
|
March 1, 2026
|
|
MindSpore中使用Graph模式运行网络，首次运行非常慢，且输入Shape改变就会重新编译
|
|
1
|
67
|
January 30, 2026
|
|
MindSpore模型推理报错TypeError: cell_reuse() takes 0 positional arguments but 1 was given
|
|
0
|
32
|
October 7, 2025
|
|
使用MindSpore混合精度模式训练出现Loss NaN
|
|
1
|
62
|
October 28, 2025
|
|
MindSpore自定义算子梯度计算不正确Loss异常
|
|
1
|
48
|
October 27, 2025
|
|
MindSpore训练时报错：TypeError: For 'MatMul', the input data must be float16, float32, uint16 but got int32
|
|
1
|
41
|
October 27, 2025
|
|
MindSpore报错Please try to reduce 'batch_size' or check whether exists extra large shape.
|
|
0
|
36
|
October 21, 2025
|
|
Ascend环境运行mindspore脚本报：网络脚本的设备被占用，当前MindSpore框架在Ascend环境只支持每张卡运行一个网络脚本
|
|
0
|
46
|
October 21, 2025
|
|
MindSpore报RuntimeError:Primitive ScatterAddapos; bprop not defined
|
|
0
|
40
|
August 28, 2025
|
|
使用单卡Ascend910进行LLaMA2-7B推理,速度缓慢
|
|
1
|
52
|
October 13, 2025
|
|
MindSpore大模型打开pp并行或者梯度累积之后loss不溢出也不收敛
|
|
0
|
31
|
October 10, 2025
|
|
MindSpore大模型微调时报溢出及解决
|
|
0
|
36
|
October 9, 2025
|
|
【案例】【Mindspore】【MIndformer】训练plog报错halMemAlloc failed，drvRetCode=6
|
|
0
|
79
|
October 9, 2025
|
|
MindSpore训练异常中止：Try to send request before Open()、Try to get response before Open()、Response is empty
|
|
0
|
24
|
October 9, 2025
|
|
pangu-100b 2k集群线性度问题定位
|
|
0
|
20
|
October 8, 2025
|
|
MindSpore模型推理报错：memory isn't enough and alloc failed, kernel name: kernel_graph_@ HostDSActor, alloc size: 8192B
|
|
0
|
35
|
October 7, 2025
|
|
昇腾910FlashAttention适配alibi问题
|
|
0
|
51
|
October 7, 2025
|
|
【案例】【Mindspore】【离线权重转换系列三】MindSpore的ckpt格式完整权重和分布式权重互转
|
|
0
|
35
|
October 7, 2025
|
|
模型推理报错RuntimeError A model class needs to define a `prepare inputs fordgeneration` method in order to use .generate()`
|
|
0
|
27
|
October 6, 2025
|
|
昇腾910上CodeLlama推理报错get fail deviceLogicId[0]
|
|
0
|
24
|
October 5, 2025
|
|
llama3.1-8b的lora微调，不开启权重转换会导致维度不匹配，开启了之后会报错找不到rank1的ckpt，但是strategy目录里面是全的
|
|
0
|
23
|
October 5, 2025
|
|
baichuan2-13b算子溢出 loss跑飞问题和定位
|
|
0
|
30
|
October 5, 2025
|
|
MindSpore模型报错Reason: Memory resources are exhausted.
|
|
0
|
58
|
October 5, 2025
|
|
Llama推理报参数校验错误TypeError: The input value must be int. but got 'NoneType.
|
|
0
|
24
|
October 4, 2025
|
|
Mindformers模型启动时因为host侧OOM导致任务被kill
|
|
0
|
80
|
October 4, 2025
|