Whisper无法在香橙派AIPro上推理 #1858
Labels
bug
Something isn't working
Comments
|
需要设置 set_pyboost(False) |
|
刚刚在香橙派上添加了 from mindnlp.transformers import pipeline
from mindnlp.configs import set_pyboost
transcriber = pipeline(model="openai/whisper-base")
set_pyboost(False)
transcriber("/root/whisper2om/Birth.wav") # 中文音频
transcriber("/root/whisper2om/test.wav") # 英文音频仍旧报错： python testmindnlp.py
/usr/local/miniconda3/lib/python3.9/site-packages/numpy/core/getlimits.py:499: UserWarning: The value of the smallest subnormal for <class 'numpy.float64'> type is zero.
setattr(self, word, getattr(machar, word).flat[0])
/usr/local/miniconda3/lib/python3.9/site-packages/numpy/core/getlimits.py:89: UserWarning: The value of the smallest subnormal for <class 'numpy.float64'> type is zero.
return self._float_to_str(self.smallest_subnormal)
/usr/local/miniconda3/lib/python3.9/site-packages/numpy/core/getlimits.py:499: UserWarning: The value of the smallest subnormal for <class 'numpy.float32'> type is zero.
setattr(self, word, getattr(machar, word).flat[0])
/usr/local/miniconda3/lib/python3.9/site-packages/numpy/core/getlimits.py:89: UserWarning: The value of the smallest subnormal for <class 'numpy.float32'> type is zero.
return self._float_to_str(self.smallest_subnormal)
[WARNING] DEVICE(120308,e7ffc34ca020,python):2024-12-11-16:03:45.493.032 [mindspore/ccsrc/plugin/device/ascend/hal/device/ascend_memory_adapter.cc:116] Initialize] Free memory size is less than half of total memory size.Device 0 Device HBM total size:16367894528 Device HBM free size:7702208512 may be other processes occupying this card, check as: ps -ef|grep python
/usr/local/miniconda3/lib/python3.9/site-packages/mindnlp/transformers/models/whisper/generation_whisper.py:491: FutureWarning: The input name `inputs` is deprecated. Please make sure to use `input_features` instead.
warnings.warn(
Due to a bug fix in https://github.com/huggingface/transformers/pull/28687 transcription using a multilingual Whisper will default to language detection followed by transcription instead of translation to English.This might be a breaking change for your use case. If you want to instead always translate your audio to English, make sure to pass `language='en'`.
Passing a tuple of `past_key_values` is deprecated and will be removed in Transformers v4.43.0. You should pass an instance of `EncoderDecoderCache` instead, e.g. `past_key_values=EncoderDecoderCache.from_legacy_cache(past_key_values)`.
Traceback (most recent call last):
File "/root/mindnlp-whisper/testmindnlp.py", line 6, in <module>
transcriber("/root/whisper2om/Birth.wav")
File "/usr/local/miniconda3/lib/python3.9/site-packages/mindnlp/transformers/pipelines/automatic_speech_recognition.py", line 282, in __call__
return super().__call__(inputs, **kwargs)
File "/usr/local/miniconda3/lib/python3.9/site-packages/mindnlp/transformers/pipelines/base.py", line 1161, in __call__
return self.run_single(inputs, preprocess_params, forward_params, postprocess_params)
File "/usr/local/miniconda3/lib/python3.9/site-packages/mindnlp/transformers/pipelines/base.py", line 1297, in run_single
model_outputs = self.forward(model_inputs, **forward_params)
File "/usr/local/miniconda3/lib/python3.9/site-packages/mindnlp/transformers/pipelines/base.py", line 1103, in forward
model_outputs = self._forward(model_inputs, **forward_params)
File "/usr/local/miniconda3/lib/python3.9/site-packages/mindnlp/transformers/pipelines/automatic_speech_recognition.py", line 506, in _forward
tokens = self.model.generate(
File "/usr/local/miniconda3/lib/python3.9/site-packages/mindnlp/transformers/models/whisper/generation_whisper.py", line 537, in generate
init_tokens = self._retrieve_init_tokens(
File "/usr/local/miniconda3/lib/python3.9/site-packages/mindnlp/transformers/models/whisper/generation_whisper.py", line 1343, in _retrieve_init_tokens
lang_ids = self.detect_language(
File "/usr/local/miniconda3/lib/python3.9/site-packages/mindnlp/transformers/models/whisper/generation_whisper.py", line 1450, in detect_language
non_lang_mask[list(generation_config.lang_to_id.values())] = False
File "/usr/local/miniconda3/lib/python3.9/site-packages/mindspore/common/_stub_tensor.py", line 49, in fun
return method(*arg, **kwargs)
File "/usr/local/miniconda3/lib/python3.9/site-packages/mindspore/common/tensor.py", line 496, in __setitem__
self.assign_value(out)
File "/usr/local/miniconda3/lib/python3.9/site-packages/mindspore/_check_jit_forbidden_api.py", line 35, in jit_forbidden
return fn(*args, **kwargs)
File "/usr/local/miniconda3/lib/python3.9/site-packages/mindspore/common/tensor.py", line 942, in assign_value
self.assign_value_cpp(value)
RuntimeError: Call aclnnSWhere failed, detail:EZ9999: Inner Error!
EZ9999: [PID: 120308] 2024-12-11-16:04:27.069.860 Parse dynamic kernel config fail.[THREAD:120481]
TraceBack (most recent call last):
AclOpKernelInit failed opType[THREAD:120481]
Op SelectV2 does not has any binary.[THREAD:120482]
Kernel Run failed. opType: 22, SelectV2[THREAD:120482]
launch failed for SelectV2, errno:561000.[THREAD:120482]
----------------------------------------------------
- C++ Call Stack: (For framework developers)
----------------------------------------------------
mindspore/ops/kernel/ascend/pyboost/auto_generate/select.cc:53 operator() |
|
香橙派AIPro Ascend310B1 版本, 报错完全相同 |
|
最新源码安装的mindnlp吗 |
|
是的. 从 master 分支重新安装 mindnlp 仍然报错 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Describe the bug/ 问题描述 (Mandatory / 必填)
在香橙派AIPro板子上，运行如下代码：
遇到报错：
Hardware Environment(
Ascend/GPU/CPU) / 硬件环境:-- Ascend
Software Environment / 软件环境 (Mandatory / 必填):
-- MindSpore version (e.g., 1.7.0.Bxxx) : 2.4.1
-- Python version (e.g., Python 3.7.5) : 3.9.2
-- OS platform and distribution (e.g., Linux Ubuntu 16.04): Linux orangepiaipro aarch64
-- GCC/Compiler version (if compiled from source): 11.4.0
-- CANN: 8.0.RC3.beta1
-- Kernel: Ascend-cann-kernels-310b_8.0.RC3
-- npu-smi info:
PyNative/Graph):To Reproduce / 重现步骤 (Mandatory / 必填)
Steps to reproduce the behavior:
Expected behavior / 预期结果 (Mandatory / 必填)
在ModelArts的910B4机器上相同的CANN版本能够推理成功，但在香橙派的310B4上失败，报告EZ9999错误，重复检查了CANN包环境是没问题的。
Screenshots/ 日志 / 截图 (Mandatory / 必填)
Additional context / 备注 (Optional / 选填)
Add any other context about the problem here.
The text was updated successfully, but these errors were encountered: