Morph MORPH
SIGN IN SIGN UP
dhruvhead / maxdiffusion UNCLAIMED
0 0 0 Python

COMMITS

main
July 8, 2026
M
Merge pull request #441 from syhuang22:pyink-format-fix
maxdiffusion authors committed
dc6b582
J
Run pyink on files that landed unformatted (PRs #436, #438)
James Huang committed
f6ee0c8
M
Merge pull request #438 from syhuang22:reduce-load-compile-clean
maxdiffusion authors committed
bae41b3
July 7, 2026
M
Merge pull request #433 from HadarIngonyama:magcache_wan22_integration
maxdiffusion authors committed
0956041
M
Merge pull request #427 from ThomasNing:fix/wan22-i2v-2.2-quant-dummy-inputs
maxdiffusion authors committed
9b48bf3
M
Merge pull request #436 from syhuang22:fixed-m-ulysses
maxdiffusion authors committed
7affda9
July 3, 2026
J
Cut WAN2.2 load 141s->53s and compile 144s->21s: fast weight loader, ICI-staged device_put, parallel component loads, 2-step warmup
James Huang committed
89f2371
July 6, 2026
M
Merge pull request #417 from AI-Hypercomputer:prisha/flux_training
maxdiffusion authors committed
932628f
June 12, 2026
P
Flux training: Implement scanned blocks, dynamic gradient checkpointing, and weight loading improvements
Prisha Jain committed
25f1a08
M
Merge pull request #404 from AI-Hypercomputer:wan-ulysses-bshd-attention
maxdiffusion authors committed
ee00dce
July 1, 2026
H
Add MagCache support for Wan 2.2 T2V and I2V
HadarIngonyama committed
bf91020
July 2, 2026
T
Fix get_dummy_wan_inputs for two-expert Wan2.2 I2V pipeline
Thomas Ning committed
952aa1e
J
Add fixed-m softmax for pure-Ulysses WAN attention (ulysses_custom_fixed_m) with CFG batch-fold fix
James Huang committed
9c9b707
June 30, 2026
M
Merge pull request #430 from AI-Hypercomputer:2d_ring_custom_kernel
maxdiffusion authors committed
0984457
June 25, 2026
E
2D ring (USP) attention with custom splash kernel
Elisa Tsai committed
7528f71
M
Merge pull request #425 from tsvikas:ltx2-3-lora-support
maxdiffusion authors committed
eae38c4
June 26, 2026
M
Merge pull request #410 from flyingblackshark:main
maxdiffusion authors committed
5e884cf
June 24, 2026
M
Merge pull request #424 from AI-Hypercomputer:ltx2_fix_multihost
maxdiffusion authors committed
39af0e2
June 18, 2026
T
feat(ltx2): extend support for LTX-2.3 LoRA support
Tsvika Shapira committed
0fbf767
June 22, 2026
M
Merge pull request #419 from AI-Hypercomputer:ninatu/wan_load_params
maxdiffusion authors committed
08566b1
P
Fix LTX2 multihost JAX Array fetching error and refactor VAE decode
Prisha Jain committed
8bb6a91
S
fix(generate_wan): avoid process0-only gate for profiling run
Shu Yang committed
f62927f
June 17, 2026
M
Merge pull request #421 from AI-Hypercomputer:fix-vae-timing-regression
maxdiffusion authors committed
9616d1c
M
Merge pull request #420 from AI-Hypercomputer:ninatu/fix_heads_per_tile
maxdiffusion authors committed
4ccaaba
June 16, 2026
N
Refactor WAN loading logic: granular flags & boilerplate cleanup.
Nina Shvetsova committed
ef5af09
M
Merge pull request #418 from AI-Hypercomputer:add-sd15-support
maxdiffusion authors committed
5d9fdef
M
fix(ltx2): Fix VAE timing regression for large batch sizes
mbohlool committed
96e9d82
N
Fix custom flash block sizes fallback in Ulysses attention.
Nina Shvetsova committed
ab04e7d
June 13, 2026
C
Add Stable Diffusion 1.5 support with data-parallel inference
csgoogle committed
164ca87
May 15, 2026
S
Add explicit Ulysses ring attention sharding with segment-id masking
Sagar Chapara committed
c104db5