From 658bed828bd1be50cbcdc1d5c2c0b9aa782aded0 Mon Sep 17 00:00:00 2001 From: CarlosGomes98 Date: Wed, 19 Nov 2025 15:36:59 +0100 Subject: [PATCH] update flux rcp --- .../training_5.1.0/rcps_flux1.json | 28 +++++++++---------- 1 file changed, 14 insertions(+), 14 deletions(-) diff --git a/mlperf_logging/rcp_checker/training_5.1.0/rcps_flux1.json b/mlperf_logging/rcp_checker/training_5.1.0/rcps_flux1.json index 5da82690..c612b94b 100644 --- a/mlperf_logging/rcp_checker/training_5.1.0/rcps_flux1.json +++ b/mlperf_logging/rcp_checker/training_5.1.0/rcps_flux1.json @@ -16,9 +16,9 @@ "opt_gradient_clip_norm": 1.0 }, "samples to converge": [ - 8388608, 8388608, 8388608, 8126464, 7864320, 8126464, 7864320, 7602176, - 8650752, 8126464, 8650752, 8126464, 8388608, 8126464, 8126464, 8126464, - 8126464, 8126464, 8126464, 7864320 + 7340032, 7077888, 7077888, 7340032, 7077888, 7340032, 7340032, 7340032, + 7340032, 6815744, 7340032, 7602176, 7077888, 7340032, 7077888, 7077888, + 6815744, 7077888, 7077888, 6815744 ] }, "flux_ref_1024": { @@ -33,14 +33,14 @@ "opt_adamw_beta_2": 0.95, "opt_adamw_epsilon": 1e-8, "opt_adamw_weight_decay": 0.1, - "opt_base_learning_rate": 2.0e-4, - "opt_learning_rate_warmup_steps": 0, + "opt_base_learning_rate": 2.5e-4, + "opt_learning_rate_warmup_steps": 800, "opt_gradient_clip_norm": 1.0 }, "samples to converge": [ - 8912896, 8650752, 9437184, 8126464, 8388608, 9175040, 8650752, 8126464, - 8388608, 9961472, 7864320, 8126464, 9699328, 8650752, 9437184, 8912896, - 8388608, 9175040, 8126464, 9175040 + 9175040, 8388608, 8388608, 8126464, 8650752, 8650752, 8126464, 9175040, + 8650752, 8912896, 8388608, 8912896, 8388608, 8388608, 8388608, 8126464, + 8650752, 8650752, 8388608, 8650752 ] }, "flux_ref_2048": { @@ -60,9 +60,9 @@ "opt_gradient_clip_norm": 1.0 }, "samples to converge": [ - 11272192, 10223616, 11534336, 10747904, 9699328, 10485760, 11010048, - 10223616, 11796480, 10485760, 10747904, 11272192, 9699328, 10485760, - 11534336, 9961472, 10485760, 10485760, 11272192, 11272192 + 10485760, 10223616, 10485760, 10747904, 10485760, 9961472, 11534336, + 10485760, 10485760, 11272192, 9961472, 9699328, 10747904, 9175040, + 9699328, 10485760, 9437184, 10223616, 9699328, 11010048 ] }, "flux_ref_4096": { @@ -82,9 +82,9 @@ "opt_gradient_clip_norm": 1.0 }, "samples to converge": [ - 15466496, 15728640, 15990784, 15466496, 15728640, 15466496, 14942208, - 14680064, 15728640, 15990784, 15990784, 15728640, 15728640, 16252928, - 14942208, 15728640, 16252928, 15204352, 16515072, 14942208 + 15204352, 15990784, 15466496, 15728640, 15204352, 15466496, 15990784, + 15204352, 14942208, 15204352, 15466496, 16252928, 14680064, 14942208, + 13893632, 15466496, 15466496, 15728640, 15466496, 15204352 ] } }