Downloads · 30 days
17
3% of all-time downloads
wam_ctxpool_xattn
hmkang/wam_ctxpool_xattn
wam_ctxpool_xattn is a robotics model from hmkang. Use it for the robotics task on the model card, and read the license before you ship it in a product. It is set up for wan2.2. The card lists the license as apache-2.0.
Wan2.2-TI2V-5B video DiT (vendored) fine-tuned on RoboCasa, base recipe, effective batch 64, 4-latin history (nin=25/nout=41, fdf=2). Before block L=3 the 3 PAST cond latent frames are pooled into ONE motion frame = m…
All-time downloads
547
Public
Repo size
319 GB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors319 GB · 100%
At a glance
- Task
- Robotics
- Library
- wan2.2
- License
- apache-2.0
- Access
- Public
- Created
- Sep 9, 2026
- Updated
- Sep 12, 2026
- SHA
- 0c5b8b45
Model card
From the Hugging Face model README
WAM_DIT4DIT — context pooling with a dedicated cross-attention pooler (xattn), RoboCasa kitchen
Wan2.2-TI2V-5B video DiT (vendored) fine-tuned on RoboCasa, base recipe, effective batch 64, 4-latin history (nin=25/nout=41, fdf=2). Before block L=3 the 3 PAST cond latent frames are pooled into ONE motion frame = mean(past) + ContextPooler(q = current-frame tokens, k/v = past tokens, DiT-sized 24 heads, 3D rope on q/k, zero-init output) -> step 0 == plain avg pooling.
| folder | pooler |
|---|---|
xattn_L3/checkpoint-<step> | cross-attn only |
xattn_ffn_L3/checkpoint-<step> | cross-attn + FFN refine (4x hidden, zero-init) |
Every 20k steps. Weights + configs only. Code: https://github.com/HEMMO0208/wam
(gr00t/model/wam_dit4dit/wan22/wan_video_dit.py ContextPooler / _ctx_pool_apply).
- Task
- Robotics
- Library
- wan2.2
- License
- apache-2.0
- Created
- Sep 9, 2026
- Updated
- Sep 12, 2026