@@ -201,41 +201,28 @@ class KernelCounts(NamedTuple):
201201KERNEL_COUNT_OVERRIDES = {
202202 "test_rmsprop_foreach_weight_decay_cpu" : lambda x : assert_expected_inline (x , """12""" ) ,
203203 "test_nadam_foreach_weight_decay_momentum_decay_cpu" : lambda x : assert_expected_inline (x , """20""" ),
204- "test_adam_tensor_lr_amsgrad_capturable_foreach_cuda" : lambda x : assert_expected_inline (x , """3""" ),
205- "test_adam_capturable_foreach_cuda" : lambda x : assert_expected_inline (x , """2""" ),
206- "test_adam_foreach_cuda" : lambda x : assert_expected_inline (x , """2""" ),
207- "test_adam_weight_decay_amsgrad_capturable_foreach_cuda" : lambda x : assert_expected_inline (x , """2""" ),
208- "test_adam_weight_decay_amsgrad_foreach_cuda" : lambda x : assert_expected_inline (x , """2""" ),
209- "test_adam_weight_decay_foreach_cuda" : lambda x : assert_expected_inline (x , """2""" ),
210- "test_adam_weight_decay_maximize_foreach_cuda" : lambda x : assert_expected_inline (x , """2""" ),
211204 "test_adamw_amsgrad_capturable_foreach_cuda" : lambda x : assert_expected_inline (x , """3""" ),
212- "test_adamw_tensor_lr_amsgrad_capturable_foreach_cuda" : lambda x : assert_expected_inline (x , """3""" ),
213- "test_adamw_tensor_lr_tensor_betas_amsgrad_capturable_foreach_cuda" : lambda x : assert_expected_inline (x , """3""" ),
214- "test_adamw_tensor_lr_tensor_betas_capturable_foreach_cuda" : lambda x : assert_expected_inline (x , """3""" ),
215205 "test_adamw_amsgrad_capturable_foreach_xpu" : lambda x : assert_expected_inline (x , """3""" ),
216206 "test_adamw_amsgrad_capturable_cuda" : lambda x : assert_expected_inline (x , """6""" ),
217207 "test_adamw_amsgrad_capturable_xpu" : lambda x : assert_expected_inline (x , """6""" ),
218- "test_adamw_tensor_lr_tensor_betas_amsgrad_capturable_cuda" : lambda x : assert_expected_inline (x , """3 """ ),
208+ "test_adamw_tensor_lr_tensor_betas_amsgrad_capturable_cuda" : lambda x : assert_expected_inline (x , """6 """ ),
219209 "test_adamw_tensor_lr_tensor_betas_amsgrad_capturable_xpu" : lambda x : assert_expected_inline (x , """6""" ),
220- "test_adamw_tensor_lr_tensor_betas_capturable_cuda" : lambda x : assert_expected_inline (x , """3 """ ),
210+ "test_adamw_tensor_lr_tensor_betas_capturable_cuda" : lambda x : assert_expected_inline (x , """6 """ ),
221211 "test_adamw_tensor_lr_tensor_betas_capturable_xpu" : lambda x : assert_expected_inline (x , """6""" ),
222- "test_adamw_tensor_lr_amsgrad_capturable_cuda" : lambda x : assert_expected_inline (x , """3 """ ),
212+ "test_adamw_tensor_lr_amsgrad_capturable_cuda" : lambda x : assert_expected_inline (x , """6 """ ),
223213 "test_adamw_tensor_lr_amsgrad_capturable_xpu" : lambda x : assert_expected_inline (x , """6""" ),
224- "test_adam_tensor_lr_amsgrad_capturable_cuda" : lambda x : assert_expected_inline (x , """5 """ ),
214+ "test_adam_tensor_lr_amsgrad_capturable_cuda" : lambda x : assert_expected_inline (x , """6 """ ),
225215 "test_adam_tensor_lr_amsgrad_capturable_xpu" : lambda x : assert_expected_inline (x , """6""" ),
226- "test_adam_tensor_lr_tensor_betas_amsgrad_capturable_cuda" : lambda x : assert_expected_inline (x , """3 """ ),
216+ "test_adam_tensor_lr_tensor_betas_amsgrad_capturable_cuda" : lambda x : assert_expected_inline (x , """6 """ ),
227217 "test_adam_tensor_lr_tensor_betas_amsgrad_capturable_xpu" : lambda x : assert_expected_inline (x , """6""" ),
228- "test_adam_tensor_lr_tensor_betas_capturable_cuda" : lambda x : assert_expected_inline (x , """3 """ ),
218+ "test_adam_tensor_lr_tensor_betas_capturable_cuda" : lambda x : assert_expected_inline (x , """6 """ ),
229219 "test_adam_tensor_lr_tensor_betas_capturable_xpu" : lambda x : assert_expected_inline (x , """6""" ),
230220 "test_adam_amsgrad_capturable_cuda" : lambda x : assert_expected_inline (x , """6""" ),
231221 "test_adam_amsgrad_capturable_xpu" : lambda x : assert_expected_inline (x , """6""" ),
232222 "test_adadelta_tensor_lr_capturable_cuda" : lambda x : assert_expected_inline (x , """6""" ),
233223 "test_adadelta_tensor_lr_capturable_xpu" : lambda x : assert_expected_inline (x , """6""" ),
234224 "test_rmsprop_tensor_lr_capturable_cuda" : lambda x : assert_expected_inline (x , """6""" ),
235225 "test_rmsprop_tensor_lr_capturable_xpu" : lambda x : assert_expected_inline (x , """6""" ),
236- "test_rmsprop_weight_decay_centered_foreach_cuda" : lambda x : assert_expected_inline (x , """5""" ),
237- "test_rmsprop_weight_decay_centered_momentum_foreach_cuda" : lambda x : assert_expected_inline (x , """5""" ),
238- "test_rmsprop_weight_decay_centered_momentum_maximize_foreach_cuda" : lambda x : assert_expected_inline (x , """5""" ),
239226 "test_adadelta_foreach_weight_decay_maximize_cpu" : lambda x : assert_expected_inline (x , """12""" ),
240227 "test_adadelta_foreach_rho_weight_decay_cpu" : lambda x : assert_expected_inline (x , """12""" ),
241228 "test_adadelta_foreach_weight_decay_cpu" : lambda x : assert_expected_inline (x , """12""" ),
@@ -265,18 +252,13 @@ class KernelCounts(NamedTuple):
265252 "test_adagrad_tensor_lr_cpu" : lambda x : assert_expected_inline (x , """6""" ),
266253 "test_adagrad_tensor_lr_cuda" : lambda x : assert_expected_inline (x , """6""" ),
267254 "test_adagrad_tensor_lr_xpu" : lambda x : assert_expected_inline (x , """6""" ),
268- "test_adamax_tensor_lr_weight_decay_capturable_cuda" : lambda x : assert_expected_inline (x , """5""" ),
269- "test_adamax_tensor_lr_weight_decay_capturable_foreach_cuda" : lambda x : assert_expected_inline (x , """3""" ),
255+ "test_adamax_tensor_lr_weight_decay_capturable_cuda" : lambda x : assert_expected_inline (x , """6""" ),
270256 "test_adamax_tensor_lr_weight_decay_capturable_xpu" : lambda x : assert_expected_inline (x , """6""" ),
271257 "test_asgd_tensor_lr_weight_decay_maximize_capturable_cuda" : lambda x : assert_expected_inline (x , """5""" ),
272258 "test_asgd_tensor_lr_weight_decay_maximize_capturable_xpu" : lambda x : assert_expected_inline (x , """5""" ),
273- "test_nadam_tensor_lr_weight_decay_momentum_decay_decoupled_weight_decay_capturable_cuda" : lambda x : assert_expected_inline (x , """7""" ),
274- "test_nadam_tensor_lr_weight_decay_momentum_decay_decoupled_weight_decay_capturable_xpu" : lambda x : assert_expected_inline (x , """6""" ),
275- "test_nadam_weight_decay_foreach_cuda" : lambda x : assert_expected_inline (x , """5""" ),
276- "test_nadam_weight_decay_maximize_foreach_cuda" : lambda x : assert_expected_inline (x , """5""" ),
277- "test_nadam_weight_decay_momentum_decay_capturable_foreach_cuda" : lambda x : assert_expected_inline (x , """5""" ),
278- "test_nadam_weight_decay_momentum_decay_foreach_cuda" : lambda x : assert_expected_inline (x , """5""" ),
279- "test_radam_tensor_lr_capturable_weight_decay_decoupled_weight_decay_cuda" : lambda x : assert_expected_inline (x , """10""" ),
259+ "test_nadam_tensor_lr_weight_decay_momentum_decay_decoupled_weight_decay_capturable_cuda" : lambda x : assert_expected_inline (x , """6""" ), # noqa: B950
260+ "test_nadam_tensor_lr_weight_decay_momentum_decay_decoupled_weight_decay_capturable_xpu" : lambda x : assert_expected_inline (x , """6""" ), # noqa: B950
261+ "test_radam_tensor_lr_capturable_weight_decay_decoupled_weight_decay_cuda" : lambda x : assert_expected_inline (x , """6""" ),
280262 "test_radam_tensor_lr_capturable_weight_decay_decoupled_weight_decay_xpu" : lambda x : assert_expected_inline (x , """6""" ),
281263 "test_sgd_tensor_lr_cpu" : lambda x : assert_expected_inline (x , """2""" ),
282264 "test_sgd_tensor_lr_cuda" : lambda x : assert_expected_inline (x , """2""" ),
@@ -286,16 +268,16 @@ class KernelCounts(NamedTuple):
286268
287269# also tracks currently supported optimizers
288270KERNEL_COUNTS = {
289- Adam : KernelCounts (multitensor = 3 , singletensor = 8 ),
271+ Adam : KernelCounts (multitensor = 2 , singletensor = 8 ),
290272 AdamW : KernelCounts (multitensor = 2 , singletensor = 8 ),
291- NAdam : KernelCounts (multitensor = 2 , singletensor = 12 ),
273+ NAdam : KernelCounts (multitensor = 2 , singletensor = 8 ),
292274 Rprop : KernelCounts (multitensor = 2 , singletensor = 8 ),
293275 RMSprop : KernelCounts (multitensor = 2 , singletensor = 8 ),
294276 Adadelta : KernelCounts (multitensor = 2 , singletensor = 8 ),
295277 Adagrad : KernelCounts (multitensor = 2 , singletensor = 8 ),
296278 SGD : KernelCounts (multitensor = 1 , singletensor = 8 ),
297279 ASGD : KernelCounts (multitensor = 2 , singletensor = 8 ),
298- RAdam : KernelCounts (multitensor = 2 , singletensor = 12 ),
280+ RAdam : KernelCounts (multitensor = 2 , singletensor = 8 ),
299281 Adamax : KernelCounts (multitensor = 2 , singletensor = 8 ),
300282}
301283
0 commit comments