Jump to content
Toggle search
Toggle menu
CMP Wiki
Toggle preferences menu
Toggle personal menu
Not logged in
Your IP address will be publicly visible if you make any edits.

Quantization on the 170HX: Revision history

From CMP Wiki
More actions

Diff selection: Mark the radio buttons of the revisions to compare and hit enter or the button at the bottom.
Legend: (cur) = difference with latest revision, (prev) = difference with preceding revision, m = minor edit.

30 October 2026

  • curprev 08:5408:54, 30 October 2026 ImportBot talk contribs 26,262 bytes +876 Import: Quant formats on sm_80 CMP cards (W4A16, W8A16/W8A8, EXL3 bpw choices, FP8 and FP4 handling, KV cache formats, PLE and Engram tables), what fits in 64 and 80 GB per card, heretic and abliterated variants; from Discord, the recipe repos and Ivan's lab notes; sources: @wakeupmbrc1, @gnasherrr, @programmerboi, @xeax, @nathang6866, @dale063629, @c133, @darkmaniac7, @anon33anon, @cibernox, @nestaa51, @wingrimez_40467 (+19 more); GitHub CurtisAccelerate/DeepSeekv41-Flash-10
  • curprev 08:1408:14, 30 October 2026 ImportBot talk contribs 25,386 bytes +25,386 Import: Quant formats on sm_80 CMP cards (W4A16, W8A16/W8A8, EXL3 bpw choices, FP8 and FP4 handling, KV cache formats, PLE and Engram tables), what fits in 64 and 80 GB per card, heretic and abliterated variants; from Discord, the recipe repos and Ivan's lab notes; sources: @wakeupmbrc1, @gnasherrr, @programmerboi, @xeax, @nathang6866, @dale063629, @c133, @darkmaniac7, @anon33anon, @cibernox, @nestaa51, @wingrimez_40467 (+19 more); GitHub CurtisAccelerate/DeepSeekv41-Flash-10
Retrieved from "https://cmp.wiki/wiki/Quantization_on_the_170HX"