p47068915
/
mindspore
forked from mindspore-Ecosystem/mindspore
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

!30293 [MSLITE][GPU] remove some big model，aviod loss device 

Merge pull request !30293 from Greatpan/loss_device
This commit is contained in:
i-robot 2022-02-21 13:12:20 +00:00 committed by Gitee
parent 6f88bf7c75 1ff5cf0e99
commit a813832a87
No known key found for this signature in database
GPG Key ID: 173E9B9CA92EEF8F
3 changed files with 7 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
mindspore/lite/test/config/models_caffe_gpu_fp16.cfg
View File

@ -13,11 +13,11 @@ ml_ocr_detect_20200305;1;1,544,544,3 10
detection_retinaface_fix 8
# age_new's precision deteriorates in P50
age_new 8
ml_ocr_bank_card_detection_inception_tmp 10
# ml_ocr_bank_card_detection_inception_tmp 10
glasses
hat
isface
ml_bank_detect_0312_tmp 13
# ml_bank_detect_0312_tmp 13
ml_face_div_parsing
ml_hardware_eyeclose
Mnet6_0312_extract_pay 6
@ -30,7 +30,7 @@ hiai_face_isface
hiai_face_landmark
hiai_face_pose_tuku
ml_hand_detection
ml_ocr_sfz_detect_0325_tmp
# ml_ocr_sfz_detect_0325_tmp
ml_hardware_liveness
ml_liveness_detect_landmark_tmp
ml_face_contour

6
mindspore/lite/test/config/models_caffe_gpu_fp32.cfg
View File

@ -5,7 +5,7 @@ landmark
PoseNet_dla_17_x512_tmp
age_new
plat_isface
ml_ocr_bank_card_detection_inception_tmp
# ml_ocr_bank_card_detection_inception_tmp
ml_ocr_detect_20200305
hdc_age_medium
hdc_contour_pose_128
@ -28,7 +28,7 @@ gender_res_large_deploy
glasses
hat
isface
ml_bank_detect_0312_tmp
# ml_bank_detect_0312_tmp
ml_face_div_parsing
ml_hardware_eyeclose
Mnet6_0312_extract_pay
@ -41,7 +41,7 @@ hiai_face_isface
hiai_face_landmark
hiai_face_pose_tuku
ml_hand_detection
ml_ocr_sfz_detect_0325_tmp
# ml_ocr_sfz_detect_0325_tmp
ml_hardware_liveness
ml_liveness_detect_landmark_tmp
ml_face_contour

2
mindspore/lite/test/config/models_tf_gpu_fp16.cfg
View File

@ -1,6 +1,6 @@
# [first column]:model_name, If you need input shape, please connect it through ';1;' after the model name, where '1' is the input num.
# [second column]:accuracy limit in arm64
Q_dila-small-mix-full-fineturn-390000-nopixel-nosigmoid.pb
# Q_dila-small-mix-full-fineturn-390000-nopixel-nosigmoid.pb
inception_v3.pb;1;1,299,299,3 5
mobilenet_v1_0.25_128_frozen.pb;1;1,128,128,3 5
ml_face_openclose.pb;1;1,32,32,3 5