fallbernana123456
fallbernana123456
能提供下你的解决方法吗?我也遇到这个问题
https://docs.nvidia.com/deeplearning/cudnn/release-notes/ 里最新的是 [cuDNN Release 8.4.1] 你说的这个cudnn9.2从哪里获取? 我这里用cuda 10.2 cudnn7.6 也报这个错误 -------------------------------------- C++ Traceback (most recent call last): -------------------------------------- 0 paddle::framework::ThreadPoolTempl::WorkerLoop(int) 1 paddle::framework::InterpreterCore::RunInstructionAsync(unsigned long, std::vector*, std::vector*) 2 paddle::framework::InterpreterCore::CheckGC(paddle::framework::Instruction const&, std::vector*) ----------------------...
我之前用的是paddle 2.3.1 。回退到2.2.2可以了。多谢
我也是碰到这个问题。请问有解决办法吗
> ```shell > cd python && python3 setup.py bdist_wheel && cd .. > ``` ```python for segment in segments: File "/root/exit/envs/python39/lib/python3.9/site-packages/faster_whisper/transcribe.py", line 884, in restore_speech_timestamps for segment in segments: File...
> decord安装失败,提示没有对应aarch64的版本 同问,我在910b上也遇到这个问题,有人能解答下吗?
> 是的。 google/mt5-xxl 有50G。我的gpu只有32G。那1.2的需要gpu大小最小要多少?能用[T5-XXL](https://huggingface.co/DeepFloyd/t5-v1_1-xxl) 替代吗?
> Hi, Huggingface demo is not working > > an error occurs when clicking "Tracking" > > I tried the example videos and videos of my own I also encountered...
https://speech.fish.audio/inference/#webui 这个问题跟gradio相关,要使用demo.launch(..., server_name="0.0.0.0") 或者设置环境变量GRADIO_SERVER_NAME="0.0.0.0"
> 想问下两次预测用的是同一个pipeline对象吗?怀疑是第一次预测时存在模型加载等耗时,这样的话,在第一次后,之后预测任意图片都是比较快的。 是一个pipeline对象 ` import cv2 from paddlex import OCRPipeline from paddlex import PaddleInferenceOption from paddlex.pipelines.OCR.utils import draw_ocr_box_txt import time kernel_option = PaddleInferenceOption() kernel_option.set_device("npu:0") pipeline = OCRPipeline( 'PP-OCRv4_mobile_det', 'PP-OCRv4_mobile_rec', text_det_kernel_option=kernel_option,...