Admin
8d7d90a843
ggml/llama: fix non-flat unary metal dispatch truncating rows past 1024 floats
...
Grid width for the non-flat unary path was ne01 while the kernel decomposes
tgpig.x into (ne0-chunk, row) — rows past ntg were silently dropped, masked
below the 32768-element flat-path threshold; silu on [8192, n] crossed it at
prefill batch 4 and corrupted qwen35 batched prefill. Grid is now
ne01 * ceil(ne0/nth). Adds metal unary/gated-delta-net/ssm_conv regression
tests, the llama-batch-probe bisect harness, prefill batch default 32, and
keeps the graph cache until a reserve-retry actually needs eviction.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-28 13:46:57 +02:00
Admin
fbd4972450
gamemaker: full engine round — chase camera rig, racing-game APIs, script stdlib math, real error line numbers
...
Engine (examples/gamemaker): chase camera rig (camera({chase}) — ease-behind
with mouse-wins/recenter authority), writable camera + look deltas, spatial
queries (raycast/overlap_sphere/ground_normal), save/load, sustained tones,
rot_y + collide:false spawnables, HUD slots/bars, terrain noise shaping +
height bands, per-shape instanced render batching with static slabs (3.2x),
error push-loop into the agent chat, unknown-verb/option diagnostics with
game.splash:line:col positions, streaming tail-statement finalization, quiet
toolbar UI, Shh voice hush, fable voice.
Platform: runtime vector methods (.length/.normalized/.dot/.cross), scalar+
vector lerp, TAU; ScriptVm error capture sink; window frame capture API; four
headless-JIT fixes (scalar casts, mat4 mul, Id-arg expansion, commuted
scalar-vec ops) with regression test stages. Widgets: single-line TextInput
baseline centering, TextFlow inline-code baseline alignment, glass button
corner_radius uniform. Voice/ggml Metal backends: debug logs behind
GGML_METAL_TRACE. splashgame.md: the runtime-loaded game API contract.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-10 14:20:03 +02:00
Admin
614e203f17
aimgr
2026-04-27 14:51:24 +02:00
Admin
ea993f30c4
Batch Qwen CUDA prefill MoE
2026-04-20 13:48:29 +02:00
Admin
92eee25a4c
Optimize Qwen CUDA routing and exact decode
2026-04-20 13:48:29 +02:00
Admin
2325f23dbe
Optimize Qwen CUDA MoE decode path
2026-04-20 13:48:29 +02:00
Admin
8049779c0b
Checkpoint Qwen CUDA exact progress
2026-04-20 13:48:29 +02:00
Admin
87863fe42f
Add Qwen runtime and Windows CUDA build support
2026-04-17 11:44:43 +02:00
Admin
12a0e1b135
Add CUDA multimodal exact path
2026-04-15 21:33:39 +02:00
Admin
eba806b819
Add CUDA exact chat windowing and kernels
2026-04-15 21:33:39 +02:00
Admin
c9ada286a3
cuda otw
2026-04-13 11:09:08 +02:00
Admin
56d948d62f
cuda exact chat reuse and decode tuning
2026-04-13 11:09:08 +02:00
Admin
7b81d7ef6b
wip cuda long-context profiling
2026-04-13 11:05:49 +02:00
Admin
ceb1e384e8
cuda prefill metrics and chunked kernels
2026-04-13 11:05:49 +02:00
Admin
1ab6e6ed27
cuda exact prefill to ~87 tok/s
2026-04-13 11:05:49 +02:00
Admin
dc3baaf69d
flux 1 works
2026-04-13 11:01:13 +02:00
Admin
3335c26f27
rotor quant for metal
2026-04-13 11:01:13 +02:00
Admin
cc480e274e
Add optional Rotor-style K-cache compression for Gemma 4
2026-04-13 11:01:13 +02:00
Admin
b262f7bac3
Add CUDA NVFP4 backend and GPU decode path
2026-04-11 11:44:04 +02:00
Admin
d7520e837f
mlx image opt
2026-04-11 09:51:37 +02:00
Admin
3a90346eaa
fix
2026-04-10 21:14:32 +02:00
Admin
32612cfe88
cleanup
2026-04-10 14:48:17 +02:00
Admin
c925d2a56b
mlx otw
2026-04-10 14:48:17 +02:00
Admin
747ef9b2e8
mlx 5x->2x
2026-04-10 14:46:22 +02:00
Admin
db323ee42e
mlx
2026-04-10 14:46:22 +02:00
Admin
087952a638
mlx otw
2026-04-10 14:46:22 +02:00
Admin
3968a4aaba
stdin
2026-04-05 11:30:53 +02:00
Admin
a21f8f480c
drivable
2026-04-05 11:28:47 +02:00
Admin
26cd960c48
iterating
2026-04-05 11:28:47 +02:00
Admin
a9d6172f8c
otw
2026-04-05 11:28:47 +02:00
Admin
3f2baab26d
xr llama and slug
2026-04-01 12:19:03 +02:00
Admin
cb3b36d241
llama + xr
2026-04-01 12:19:03 +02:00
Admin
95c822450b
fixup
2026-04-01 12:19:03 +02:00
