add sdpa to ViT [follow up of #29325] #483

Job	Run time
Find models to run slow tests	21s
Run all tests for the model (models/audio_spectrogram_transformer, single-gpu)	4m 25s
Run all tests for the model (models/audio_spectrogram_transformer, multi-gpu)	3m 22s
Run all tests for the model (models/deit, single-gpu)	5m 46s
Run all tests for the model (models/deit, multi-gpu)	4m 40s
Run all tests for the model (models/videomae, single-gpu)	4m 30s
Run all tests for the model (models/videomae, multi-gpu)	3m 23s
Run all tests for the model (models/vision_encoder_decoder, single-gpu)	6m 58s
Run all tests for the model (models/vision_encoder_decoder, multi-gpu)	5m 7s
Run all tests for the model (models/vision_text_dual_encoder, single-gpu)	4m 25s
Run all tests for the model (models/vision_text_dual_encoder, multi-gpu)	3m 20s
Run all tests for the model (models/vit, single-gpu)	5m 6s
Run all tests for the model (models/vit, multi-gpu)	4m 19s
Run all tests for the model (models/vit_hybrid, single-gpu)	4m 34s
Run all tests for the model (models/vit_hybrid, multi-gpu)	3m 29s
Run all tests for the model (models/vit_mae, single-gpu)	5m 7s
Run all tests for the model (models/vit_mae, multi-gpu)	4m 14s
Run all tests for the model (models/vit_msn, single-gpu)	4m 7s
Run all tests for the model (models/vit_msn, multi-gpu)	3m 42s
Run all tests for the model (models/yolos, single-gpu)	4m 3s
Run all tests for the model (models/yolos, multi-gpu)	3m 12s
	1h 28m 10s

Provide feedback