Flame Graph Reset Zoom Search ic functional.py:2210:embedding (3,145,728 bytes, 0.01%) linear.py:114:forward (4,718,592 bytes, 0.02%) module.py:1143:convert (339,738,624 bytes, 1.31%) all (25,910,312,960 bytes, 100%) module.py:1501:_call_impl (14,155,776 bytes, 0.05%) functional.py:2515:layer_norm (37,748,736 bytes, 0.15%) module.py:1501:_call_impl (15,267,307,520 bytes, 58.92%) module.py:1501:_call_impl model.py:67:forward (268,959,744 bytes, 1.04%) module.py:1501:_call_impl (2,151,677,952 bytes, 8.30%) module.py:1.. optimizer.py:33:_use_grad (994,990,380 bytes, 3.84%) opti.. module.py:820:_apply (157,682,688 bytes, 0.61%) model.py:82:forward (3,850,371,072 bytes, 14.86%) model.py:82:forward functional.py:1843:softmax (7,247,757,312 bytes, 27.97%) functional.py:1843:softmax module.py:797:_apply (339,738,624 bytes, 1.31%) <non-python> (506,014,720 bytes, 1.95%) <.. train.py:253:model_step (1,358,217,216 bytes, 5.24%) train... module.py:820:_apply (339,738,624 bytes, 1.31%) model.py:86:forward (14,155,776 bytes, 0.05%) train.py:140:dummy_step (8,519,680 bytes, 0.03%) module.py:1501:_call_impl (43,302,912 bytes, 0.17%) module.py:1501:_call_impl (37,748,736 bytes, 0.15%) train.py:209:<module> (547,826,688 bytes, 2.11%) t.. module.py:1501:_call_impl (3,145,728 bytes, 0.01%) train.py:253:model_step (15,267,307,520 bytes, 58.92%) train.py:253:model_step functional.py:2515:layer_norm (38,191,104 bytes, 0.15%) model.py:98:forward (1,189,085,184 bytes, 4.59%) model.. model.py:80:forward (7,247,757,312 bytes, 27.97%) model.py:80:forward module.py:1501:_call_impl (4,718,592 bytes, 0.02%) module.py:1143:convert (50,331,648 bytes, 0.19%) module.py:797:_apply (390,144,000 bytes, 1.51%) module.py:1501:_call_impl (1,189,085,184 bytes, 4.59%) modul.. model.py:193:forward (37,748,736 bytes, 0.15%) train.py:331:<module> (1,358,217,220 bytes, 5.24%) train... module.py:1501:_call_impl (96,141,312 bytes, 0.37%) model.py:99:forward (905,969,664 bytes, 3.50%) mod.. <gaps> (163,740,140 bytes, 0.63%) model.py:38:forward (37,748,736 bytes, 0.15%) model.py:83:forward (226,492,416 bytes, 0.87%) module.py:844:_apply (50,331,648 bytes, 0.19%) functional.py:3029:cross_entropy (3,708,813,316 bytes, 14.31%) functional.py:3029:cr.. module.py:797:_apply (547,826,688 bytes, 2.11%) m.. <gaps> (6,855,876 bytes, 0.03%) linear.py:114:forward (56,623,104 bytes, 0.22%) model.py:115:forward (4,718,592 bytes, 0.02%) module.py:1501:_call_impl (1,358,217,216 bytes, 5.24%) module.. model.py:191:forward (37,748,736 bytes, 0.15%) linear.py:114:forward (1,236,271,104 bytes, 4.77%) linea.. module.py:1501:_call_impl (37,748,736 bytes, 0.15%) active_allocated (21,040,721,408 bytes, 81.21%) active_allocated model.py:195:forward (43,302,912 bytes, 0.17%) stream_0 (25,910,312,960 bytes, 100.00%) stream_0 module.py:1501:_call_impl (4,718,592 bytes, 0.02%) functional.py:2210:embedding (37,748,736 bytes, 0.15%) module.py:1501:_call_impl (12,073,304,064 bytes, 46.60%) module.py:1501:_call_impl model.py:196:forward (37,748,736 bytes, 0.15%) <non-python> (3,317,766,152 bytes, 12.80%) <non-python> adamw.py:118:_init_group (497,495,040 bytes, 1.92%) a.. inactive (4,869,591,552 bytes, 18.79%) inactive model.py:115:forward (12,526,288,896 bytes, 48.34%) model.py:115:forward module.py:1145:to (547,826,688 bytes, 2.11%) m.. linear.py:114:forward (268,959,744 bytes, 1.04%) module.py:1501:_call_impl (56,623,104 bytes, 0.22%) model.py:78:forward (452,984,832 bytes, 1.75%) train.py:331:<module> (18,976,120,836 bytes, 73.24%) train.py:331:<module> linear.py:114:forward (1,189,085,184 bytes, 4.59%) linea.. model.py:202:forward (1,236,271,104 bytes, 4.77%) model.. module.py:797:_apply (547,826,688 bytes, 2.11%) m.. model.py:100:forward (56,623,104 bytes, 0.22%) module.py:1501:_call_impl (268,959,744 bytes, 1.04%) sparse.py:162:forward (37,748,736 bytes, 0.15%) model.py:192:forward (3,145,728 bytes, 0.01%) model.py:117:forward (2,604,662,784 bytes, 10.05%) model.py:117:f.. model.py:195:forward (15,133,310,976 bytes, 58.41%) model.py:195:forward optimizer.py:280:wrapper (994,990,380 bytes, 3.84%) opti.. grad_scaler.py:358:step (994,990,380 bytes, 3.84%) grad.. linear.py:114:forward (96,141,312 bytes, 0.37%) module.py:1143:convert (157,682,688 bytes, 0.61%) train.py:345:<module> (994,990,380 bytes, 3.84%) trai.. model.py:38:forward (38,191,104 bytes, 0.15%) train.py:145:<module> (8,519,680 bytes, 0.03%) module.py:1501:_call_impl (15,133,310,976 bytes, 58.41%) module.py:1501:_call_impl model.py:114:forward (38,191,104 bytes, 0.15%) model.py:79:forward (12,582,912 bytes, 0.05%) module.py:1501:_call_impl (38,191,104 bytes, 0.15%) sparse.py:162:forward (3,145,728 bytes, 0.01%) module.py:797:_apply (390,144,000 bytes, 1.51%) adamw.py:160:step (994,990,380 bytes, 3.84%) adam.. model.py:86:forward (4,718,592 bytes, 0.02%) adamw.py:114:_init_group (497,495,040 bytes, 1.92%) a.. model.py:202:forward (96,141,312 bytes, 0.37%) linear.py:114:forward (14,155,776 bytes, 0.05%) module.py:1501:_call_impl (1,236,271,104 bytes, 4.77%) modul.. train.py:254:model_step (3,708,813,316 bytes, 14.31%) train.py:254:model_step