Views
No views yet
pearl-ai/Llama-3.1-8B-Instruct-pearl.eagle31--norm-before-residual
2--norm-before-fc
3--draft-vocab-size 32000
4--epochs 5
5--lr 1e-4
6--total-seq-len 4096/persist/pearl_eagle31_train_200k_seq4096/checkpoints/checkpoints_eagle31_scratch/checkpoint_best1K = 2,3,4,5,6
2C = 1,4,8,16,32,64
3t ~= 50,180,350,700,1400FULL_MATRIX_SUMMARY.mdsummary_compact.jsonfull_matrix_summary.jsonlbenchmark_raw/*/combined_summary.csvR128 128x128x128 stages=2 to fit H100 SMEM but still crashes requests with CUDA illegal instruction; stable results here use the old/materialized-B path.