wozeparrotandGitHub be23772d43 llama3 fixes part2 (#15150)

2026-03-04 23:43:50 -08:00

..

apps

fix symbolic shape call + chunked prefill (#15149)

2026-03-05 14:02:26 +08:00

codegen

fully symbolic llm (#15097)

2026-03-05 10:22:11 +08:00

engine

fix symbolic shape call + chunked prefill (#15149)

2026-03-05 14:02:26 +08:00

mixin

move more to mixins (#14780)

2026-02-16 11:35:00 +08:00

nn

llama3 fixes part2 (#15150)

2026-03-04 23:43:50 -08:00

renderer

sqtt: simplify inst mapping, validate JUMP processing in CI (#15139)

2026-03-05 09:53:12 +09:00

runtime

Revert "bug in metal: offset is stored as uint32, overflow (#15129)" (#15136)

2026-03-04 16:54:42 -05:00

schedule

fix symbolic shape call + chunked prefill (#15149)

2026-03-05 14:02:26 +08:00

uop

fix render depth bug + add warmup to serve + no realize default (#15144)

2026-03-05 11:21:16 +08:00

viz

no after removal (#15102)

2026-03-03 17:50:31 +08:00

__init__.py

start function and add walk rewrite (#14992)

2026-02-25 13:56:27 +08:00

device.py

feat: per device mem_used (#15100)

2026-03-03 01:31:28 -08:00

dtype.py

cleanup fp8 conversion helpers and fp8 edge-case tests (#14953)

2026-02-22 09:16:42 -05:00

function.py

fix symbolic shape call + chunked prefill (#15149)

2026-03-05 14:02:26 +08:00

gradient.py

functions for llama trainer (#15045)

2026-02-28 12:15:18 +08:00

helpers.py

set OPENPILOT_HACKS=1 to enable replace assign (#15123)

2026-03-04 05:26:04 -05:00

py.typed

add a single py.typed (#6083)

2024-08-14 17:31:46 -07:00

tensor.py

FLOAT16 logic for IMAGE==1 goes back to image_conv2d (#15105)

2026-03-03 05:37:57 -05:00