Update README.md
Browse files
README.md
CHANGED
|
@@ -23,6 +23,8 @@ Hello users, from Claude.
|
|
| 23 |
|
| 24 |
TenStrip ran a cross-model graft of Krea 2 (a 6144-hidden image DiT with 48 attention heads and grouped-query attention) into Zimage (a smaller 3840-hidden image DiT with 30 attention heads). Both are SwiGLU-based transformer image models with fused QKV attention. The tool I wrote for this — graft_krea_to_zimage.py — is the target-side adaptation of the same graft methodology TenStrip has been developing for H3 (a video model). This was its first full-strength run.
|
| 25 |
|
|
|
|
|
|
|
| 26 |
Which Krea layers were transferred into which Zimage tensors, per grafted block:
|
| 27 |
|
| 28 |
Krea attn.wq (48 Q heads, one at a time) → placed into Zimage's 30 Q slots inside the fused attention.qkv tensor. With all48 source selection and 30 target slots, the last-written 30 of Krea's 48 Q heads end up in Zimage's Q band. Each head is a 128-dim slice; Krea's 6144 input columns were reduced to Zimage's 3840 via truncation (SVD reduction falls back to truncation here because per-head slice rank is only 128).
|
|
|
|
| 23 |
|
| 24 |
TenStrip ran a cross-model graft of Krea 2 (a 6144-hidden image DiT with 48 attention heads and grouped-query attention) into Zimage (a smaller 3840-hidden image DiT with 30 attention heads). Both are SwiGLU-based transformer image models with fused QKV attention. The tool I wrote for this — graft_krea_to_zimage.py — is the target-side adaptation of the same graft methodology TenStrip has been developing for H3 (a video model). This was its first full-strength run.
|
| 25 |
|
| 26 |
+
(Tenstrip again: No this is not a full strength run. There is no such thing as full strength imo, just going until it's nice change or breaks.)
|
| 27 |
+
|
| 28 |
Which Krea layers were transferred into which Zimage tensors, per grafted block:
|
| 29 |
|
| 30 |
Krea attn.wq (48 Q heads, one at a time) → placed into Zimage's 30 Q slots inside the fused attention.qkv tensor. With all48 source selection and 30 target slots, the last-written 30 of Krea's 48 Q heads end up in Zimage's Q band. Each head is a 128-dim slice; Krea's 6144 input columns were reduced to Zimage's 3840 via truncation (SVD reduction falls back to truncation here because per-head slice rank is only 128).
|