AbhishekGhosh1998 commited on
Commit
31db048
·
verified ·
1 Parent(s): 6645108

granularity results tiera-rerun-fixed-moonep @ 2026-08-30T08:12:37Z from sp-aiinfra-interactive-aghosh-7jtjl-8xb200-moe-cg-0 (part 2)

Browse files
Files changed (1) hide show
  1. tiera-rerun-fixed-moonep/study.log +42 -0
tiera-rerun-fixed-moonep/study.log CHANGED
@@ -1712,3 +1712,45 @@ tierbprobeG3 4096 16 matched none 0 0 ultraep 512.31
1712
  tierbprobeG4 4096 16 matched none 0 0 ultraep 600.19 1.00x
1713
  echo 790.97 1.32x
1714
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1712
  tierbprobeG4 4096 16 matched none 0 0 ultraep 600.19 1.00x
1713
  echo 790.97 1.32x
1714
 
1715
+ 21 probe runs -> /workspace/tiera-rerun-fixed-moonep/probe_summary_all.tsv
1716
+ run arm cap rawExpI preRankI postRankI drop% reqCap
1717
+ --------------------------------------------------------------------------------------------------------------------
1718
+ echo/tierbprobe/tierbprobe_gbs16/G1__echo__matched__ natural 2.00 3.6198* 3.6198 1.0827 1.9062
1719
+ echo/tierbprobe/tierbprobe_gbs16/G2__echo__matched__ natural 1.08203125 3.8720* 2.1703 1.0000 1.0273
1720
+ echo/tierbprobe/tierbprobe_gbs16/G2__echo__matched__ forced 1.08203125 1.0201 1.0069 1.0000 1.0234
1721
+ echo/tierbprobe/tierbprobe_gbs16/G2__echo__matched__ natural 2.00 3.8724* 2.1875 1.0000 1.0273
1722
+ echo/tierbprobe/tierbprobe_gbs16/G3__echo__matched__ natural 1.072265625 3.9096* 1.7838 1.0000 1.0195
1723
+ echo/tierbprobe/tierbprobe_gbs16/G3__echo__matched__ forced 1.072265625 1.0231 1.0048 1.0000 1.0176
1724
+ echo/tierbprobe/tierbprobe_gbs16/G3__echo__matched__ natural 2.00 3.9096* 1.7916 1.0000 1.0195
1725
+ echo/tierbprobe/tierbprobe_gbs16/G4__echo__matched__ natural 1.068359375 3.9303* 1.5507 1.0000 1.0166
1726
+ echo/tierbprobe/tierbprobe_gbs16/G4__echo__matched__ forced 1.068359375 1.0249 1.0034 1.0000 1.0137
1727
+ echo/tierbprobe/tierbprobe_gbs16/G4__echo__matched__ natural 2.00 3.9301* 1.5543 1.0000 1.0166
1728
+ ultraep/tierbprobe/tierbprobe_gbs16/G1__ultraep__mat natural 1.25 3.6085 1.0640 2.8918 1.7246
1729
+ ultraep/tierbprobe/tierbprobe_gbs16/G1__ultraep__mat natural 1.8125 3.6152 1.0747 1.5460 2.3091
1730
+ ultraep/tierbprobe/tierbprobe_gbs16/G2__ultraep__mat natural 1.0625 2.2032 1.0100 0.0000 1.0101
1731
+ ultraep/tierbprobe/tierbprobe_gbs16/G2__ultraep__mat natural 1.25 2.2136 1.0100 0.0000 1.0101
1732
+ ultraep/tierbprobe/tierbprobe_gbs16/G2__ultraep__mat forced 1.25 1.0069 1.0067 0.0000 1.0100
1733
+ ultraep/tierbprobe/tierbprobe_gbs16/G3__ultraep__mat natural 1.0625 1.7842 1.0100 0.0000 1.0101
1734
+ ultraep/tierbprobe/tierbprobe_gbs16/G3__ultraep__mat natural 1.25 1.7752 1.0100 0.0000 1.0101
1735
+ ultraep/tierbprobe/tierbprobe_gbs16/G3__ultraep__mat forced 1.25 1.0047 1.0047 0.0000 1.0100
1736
+ ultraep/tierbprobe/tierbprobe_gbs16/G4__ultraep__mat natural 1.095703125 1.5612 1.0100 0.0000 1.0429
1737
+ ultraep/tierbprobe/tierbprobe_gbs16/G4__ultraep__mat natural 1.25 1.5504 1.0100 0.0000 1.0429
1738
+ ultraep/tierbprobe/tierbprobe_gbs16/G4__ultraep__mat forced 1.25 1.0033 1.0033 0.0000 1.0071
1739
+
1740
+ * At least one globally aggregated snapshot reached the E/K expert-Imax ceiling.
1741
+ Inspect raw_expert_imax_mean and hottest_expert_token_fraction_mean before generalising.
1742
+
1743
+ ECHO note: echo_logged_pre_rank_metric_valid=0 and echo_logged_load_reduction_valid=0 are intentional.
1744
+ The branch's built-in fields mix source-rank and destination-rank axes; this tool reconstructs both globally.
1745
+ 6 memory-dump runs -> /workspace/tiera-rerun-fixed-moonep/memory_summary_all.tsv
1746
+ model backend cg allocation capacity status peak GiB dumps
1747
+ ----------------------------------------------------------------------------------
1748
+ G1 echo 1 static 2.40625 OK 157.9 48/48
1749
+ G1 echo 0 dynamic dynamic OK 112.6 40/40
1750
+ G1 moonep 1 none none OK 94.6 48/48
1751
+ G1 moonep 0 none none OK 90.4 40/40
1752
+ G1 ultraep 1 static 2.40625 OK 126.4 48/48
1753
+ G1 ultraep 0 dynamic dynamic OK 106.7 40/40
1754
+
1755
+ ECHO/UltraEP note: graph-on uses static capacity + device-initiated GEMM while graph-off uses dynamic allocation + host-derived sizes. Their difference is not labelled as pure graph cost.
1756
+ Already logged in to Hugging Face as: AbhishekGhosh1998