mukesh 694e0cc9b9 paper(gqa): add 6-case per-PE memory + comm summary chart ...

Analytical plot generator + PNG covering the 4 sharding cases from
slide 17 plus the 2 d_head-TP variants (Cases 4, 5). Three panels:
per-PE HBM budget breakdown (Wq + Wk + Wv + Wo + 4.24 GB KV
headroom), per-PE KV memory at S_kv=1M, and per-PE communication
per output token (decode). Numbers match slide 17 (max KV context
~7.1M for the 64-way sharded cases).

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

2026-06-16 16:12:26 -07:00

..

paper_cube_architecture.py

paper: §2 platform deep-edit + §1 GQA framing + figure regen

2026-06-16 16:00:23 -07:00

paper_latency_model_diagram.py

paper(platform): edit-pass through §2 KernBench Platform done

2026-06-15 17:17:19 -07:00

paper_plot_gqa_4cases_summary.py

paper(gqa): add 6-case per-PE memory + comm summary chart

2026-06-16 16:12:26 -07:00

paper_plot_gqa_decode_long_ctx_4cases.py

gqa: reorganize benches into gqa_helpers/ subpackage; drop legacy headline

2026-06-16 13:05:41 -07:00

paper_plot_gqa_prefill_long_ctx_4cases.py

gqa(prefill-4cases): add long-context prefill 4-cases comparative study

2026-06-16 13:06:08 -07:00

paper_plot_gqa.py

bench(milestone-gqa-headline): drop misleading single_user_/multi_user_ panels

2026-06-15 14:50:25 -07:00

paper_sip_architecture.py

paper: §2 platform deep-edit + §1 GQA framing + figure regen

2026-06-16 16:00:23 -07:00