RuntimeError: Error while calling cudaOccupancyMaxPotentialBlockSize(&num_blocks,&num_threads,K)

[999-saxena]

Hi, I'm trying to run face recognition for a small project that im making on google colab but I'm facing this error.

!pip install face-recognition
!pip install opencv-python-headless
import face_recognition

from google.colab import files
video_file = files.upload()
target_face = files.upload()

import cv2
import face_recognition
import io
from PIL import Image
import numpy as np

ref_img = next(iter(target_face))
ref_image = face_recognition.load_image_file(io.BytesIO(target_face[ref_img]))
ref_encoding = face_recognition.face_encodings(ref_image)[0]

Error:

RuntimeError: Error while calling cudaOccupancyMaxPotentialBlockSize(&num_blocks,&num_threads,K) in file /tmp/.tmpC1GBSI/sdists-v9/pypi/dlib/19.24.6/gUXudNfXDEAetBJpufeEr/src/dlib/cuda/cuda_utils.h:186. code: 222, reason: the provided PTX was compiled with an unsupported toolchain.

any idea how to overcome this?

[Mat198]

I also found this same erro. Any suggestion?

[Mat198]

Installing the binary of dlib worked for me:

pip install dlib-bin==19.24.6 face-recognition==1.3.0 Pillow