IndexError: (OutOfRange) The starting index 0 of slice is out of bounds in tensor 0-th axis, it shound be in the range of [0, 0). (at /paddle/paddle/fluid/pybind/slice_utils.h:241)

[sugarman490]

请提出你的问题 Please ask your question

昇腾Atlas 300I Duo机器 arm
python 3.9.22
paddle环境：
paddle-custom-npu 3.0.0
paddle2onnx 2.0.1
paddleaudio 1.0.2
paddlefsl 1.1.0
paddlenlp 2.8.1
paddlepaddle 3.0.0
paddlesde 0.2.5
paddleslim 2.6.0
paddlespeech 1.4.0
paddlespeech_ctcdecoders 0.2.2
paddlespeech-feat 0.1.0
报错：
Traceback (most recent call last):
File "/work/ASR/speed_test.py", line 34, in
wav_file = tts_executor(text=text, output=output_file, am="fastspeech2_csmsc")
File "/root/miniforge3/envs/PaddleSpeech/lib/python3.9/site-packages/paddlespeech/cli/utils.py", line 328, in _warpper
return executor_func(self, *args, **kwargs)
File "/root/miniforge3/envs/PaddleSpeech/lib/python3.9/site-packages/paddlespeech/cli/tts/infer.py", line 710, in call
self.infer(text=text, lang=lang, am=am, spk_id=spk_id)
File "/root/miniforge3/envs/PaddleSpeech/lib/python3.9/site-packages/decorator.py", line 235, in fun
return caller(func, *(extras + args), **kw)
File "/root/miniforge3/envs/PaddleSpeech/lib/python3.9/site-packages/paddle/base/dygraph/base.py", line 400, in _decorate_function
return func(*args, **kwargs)
File "/root/miniforge3/envs/PaddleSpeech/lib/python3.9/site-packages/paddlespeech/cli/tts/infer.py", line 496, in infer
mel = self.am_inference(part_phone_ids)
File "/root/miniforge3/envs/PaddleSpeech/lib/python3.9/site-packages/paddle/nn/layer/layers.py", line 1567, in call
return self.forward(*inputs, **kwargs)
File "/root/miniforge3/envs/PaddleSpeech/lib/python3.9/site-packages/paddlespeech/t2s/models/fastspeech2/fastspeech2.py", line 920, in forward
normalized_mel, d_outs, p_outs, e_outs = self.acoustic_model.inference(
File "/root/miniforge3/envs/PaddleSpeech/lib/python3.9/site-packages/paddlespeech/t2s/models/fastspeech2/fastspeech2.py", line 810, in inference
_, outs, d_outs, p_outs, e_outs, _ = self._forward(
File "/root/miniforge3/envs/PaddleSpeech/lib/python3.9/site-packages/paddlespeech/t2s/models/fastspeech2/fastspeech2.py", line 601, in _forward
x_masks = self._source_mask(ilens)
File "/root/miniforge3/envs/PaddleSpeech/lib/python3.9/site-packages/paddlespeech/t2s/models/fastspeech2/fastspeech2.py", line 891, in _source_mask
x_masks = make_non_pad_mask(ilens)
File "/root/miniforge3/envs/PaddleSpeech/lib/python3.9/site-packages/paddlespeech/t2s/modules/nets_utils.py", line 259, in make_non_pad_mask
return paddle.logical_not(make_pad_mask(lengths, xs, length_dim))
File "/root/miniforge3/envs/PaddleSpeech/lib/python3.9/site-packages/paddlespeech/t2s/modules/nets_utils.py", line 147, in make_pad_mask
bs = paddle.shape(lengths)[0]
File "/root/miniforge3/envs/PaddleSpeech/lib/python3.9/site-packages/paddle/base/dygraph/tensor_patch_methods.py", line 1022, in getitem
return self._getitem_dygraph(item)
IndexError: (OutOfRange) The starting index 0 of slice is out of bounds in tensor 0-th axis, it shound be in the range of [0, 0). (at /paddle/paddle/fluid/pybind/slice_utils.h:241)

若设置export FLAGS_npu_jit_compile=false，报错:
ACL error 500002

[ZhangX-21]

@zxcd 可以看看这个问题吗？

[zxcd]

尝试使用paddlespeech==develop看下是否能够解决问题

[sugarman490]

尝试使用paddlespeech==develop看下是否能够解决问题

直接使用pip无法安装。
用源码部署后显示为paddlepaddle 0.0.0。
代码是API形式，，每次post的数据都一样。
重新运行代码后，第一次post报错：
(InvalidArgument) Each value of attribute 'shape' is expected to be no less than 0. But received: shape[0] = -9223372034707292160; shape = [-9223372034707292160].
[Hint: Expected data[i] >= 0, but received data[i]:-9223372034707292160 < 0:0.] (at /paddle/paddle/phi/infermeta/nullary.cc:65)
第二次post报错：
(PreconditionNotMet) The meta data must be valid when call the mutable data function.
[Hint: Expected valid() == true, but received valid():0 != true:1.] (at /paddle/paddle/phi/core/dense_tensor.cc:117)
再运行一两次post后报错并自动结束：
Call aclrtMalloc(&data, size, ACL_MEM_MALLOC_HUGE_FIRST) failed : 107000 at file /paddle/backends/npu/runtime/runtime.cc line 596
EZ1001: 2025-04-27-09:03:54.723.216 Get path and read binary_info_config.json failed, please check if the opp_kernel package is installed!
TraceBack (most recent call last):
Op GreaterEqual does not has any binary.
Kernel GetWorkspace failed. opType: 18
Op Cast does not has any binary.
Kernel GetWorkspace failed. opType: 19
Kernel Run failed. opType: 18, GreaterEqual
launch failed for GreaterEqual, errno:561000.
Kernel Run failed. opType: 19, Cast
launch failed for Cast, errno:561000.
Op LogicalNot does not has any binary.
Kernel GetWorkspace failed. opType: 20
Kernel Run failed. opType: 20, LogicalNot
launch failed for LogicalNot, errno:561000.
Op Fill does not has any binary.
Kernel GetWorkspace failed. opType: 6
Kernel Run failed. opType: 6, Fill
launch failed for Fill, errno:561000.
Op Exp does not has any binary.
Kernel GetWorkspace failed. opType: 7
Kernel Run failed. opType: 7, Exp
launch failed for Exp, errno:561000.
Op Mul does not has any binary.
Kernel GetWorkspace failed. opType: 8
Kernel Run failed. opType: 8, Mul
launch failed for Mul, errno:561000.
Op Sin does not has any binary.
Kernel GetWorkspace failed. opType: 9
Kernel Run failed. opType: 9, Sin
launch failed for Sin, errno:561000.
Op Expand does not has any binary.
Kernel GetWorkspace failed. opType: 3
Kernel Run failed. opType: 3, Expand
launch failed for Expand, errno:561000.
Get path and read binary_info_config.json failed, please check if the opp_kernel package is installed!
Check NnopbaseCollecterWork(binCollecter.get()) failed
Assert ((NnopbaseInit()) == 0) failed
Check NnopbaseCreateExecutorSpace(&executorSpace) failed
Assert ((executor) != nullptr) failed
Op ZerosLike does not has any binary.
Kernel GetWorkspace failed. opType: 4
Kernel Run failed. opType: 4, ZerosLike
launch failed for ZerosLike, errno:561000.
Op GatherV2 does not has any binary.
Kernel GetWorkspace failed. opType: 5
Kernel Run failed. opType: 5, GatherV2
launch failed for GatherV2, errno:561000.
Op Cos does not has any binary.
Kernel GetWorkspace failed. opType: 10
Kernel Run failed. opType: 10, Cos
launch failed for Cos, errno:561000.
Op Add does not has any binary.
Kernel GetWorkspace failed. opType: 15
Kernel Run failed. opType: 15, Add
launch failed for Add, errno:561000.
Op LayerNormV3 does not has any binary.
Kernel GetWorkspace failed. opType: 21
Kernel Run failed. opType: 21, LayerNormV3
launch failed for LayerNormV3, errno:561000.
Self dtype DT_FLOAT or mat2 dtype DT_FLOAT not support under keep type mode.
executor is nullptr.
executor or workspace is nullptr.
Op SelectV2 does not has any binary.
Kernel GetWorkspace failed. opType: 22
Kernel Run failed. opType: 22, SelectV2
launch failed for SelectV2, errno:561000.
Op SoftmaxV2 does not has any binary.
Kernel GetWorkspace failed. opType: 23
Kernel Run failed. opType: 23, SoftmaxV2
launch failed for SoftmaxV2, errno:561000.
The soc version does not support bf16 / fp32 for calculations, please change the setting of cubeMathType or the Dtype of input tensor.
Check Param failed
Op Relu does not has any binary.
Kernel GetWorkspace failed. opType: 24
Kernel Run failed. opType: 24, Relu
launch failed for Relu, errno:561000.
Op Round does not has any binary.
Kernel GetWorkspace failed. opType: 25
Kernel Run failed. opType: 25, Round
launch failed for Round, errno:561000.
Op ClipByValueV2 does not has any binary.
Kernel GetWorkspace failed. opType: 26
Kernel Run failed. opType: 26, ClipByValueV2
launch failed for ClipByValueV2, errno:561000.
Op ReduceSum does not has any binary.
Kernel GetWorkspace failed. opType: 27
Kernel Run failed. opType: 27, ReduceSum
launch failed for ReduceSum, errno:561000.
The argument is invalid.
rtMalloc execute failed, reason=[driver error:invalid value][FUNC:FuncErrorReason][FILE:error_message_manage.cc][LINE:53]
alloc device memory failed, runtime result = 107000[FUNC:ReportCallError][FILE:log_inner.cpp][LINE:161]

Segmentation fault (core dumped)

请问怎么解决？

[ZhangX-21]

@zxcd 可以麻烦看看吗？

[zxcd]

你这是在安装paddlespeech==develop还是paddlepaddle==develop?
paddlespeech的源码安装直接 git clone paddlespeech，之后pip install . 就可以

[sugarman490]

你这是在安装paddlespeech==develop还是paddlepaddle==develop? paddlespeech的源码安装直接 git clone paddlespeech，之后pip install . 就可以

不好意思打错字了，是paddlespeech

[ZhangX-21]

@zxcd 可以麻烦看看吗？

[zxcd]

那就应该是这个模型没有支持npu，可以尝试下支持npu的模型，比如https://github.com/PaddlePaddle/PaddleSpeech/blob/develop/examples/csmsc/tts2/README.md