Spaces:
Running
Running
publish from LatentPortPort c4a5ebf
Browse files
README.md
CHANGED
|
@@ -160,6 +160,19 @@ the two sizes differ in width, so the convolution state cannot be copied: the br
|
|
| 160 |
[LFM Open License v1.0](https://huggingface.co/agentic-underground/lfm2-350m-to-1.2b-bridge/blob/main/LICENSE),
|
| 161 |
which does not license Commercial Use by an entity with USD 10M or more in annual revenue.
|
| 162 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 163 |
### π A hybrid model β [`qwen3.5-0.8b-to-2b-bridge`](https://huggingface.co/agentic-underground/qwen3.5-0.8b-to-2b-bridge)
|
| 164 |
|
| 165 |
Qwen3.5-0.8B reads; Qwen3.5-2B answers. Qwen3.5 mixes full-attention layers with Gated DeltaNet
|
|
|
|
| 160 |
[LFM Open License v1.0](https://huggingface.co/agentic-underground/lfm2-350m-to-1.2b-bridge/blob/main/LICENSE),
|
| 161 |
which does not license Commercial Use by an entity with USD 10M or more in annual revenue.
|
| 162 |
|
| 163 |
+
### π Mamba-2, translated β [`granite-4.0-h-1b-to-h-tiny-bridge`](https://huggingface.co/agentic-underground/granite-4.0-h-1b-to-h-tiny-bridge)
|
| 164 |
+
|
| 165 |
+
Granite 4.0 H 1B reads; H Tiny, a 64-expert mixture, answers. Mamba-2 state cannot be copied between
|
| 166 |
+
separately trained sizes, so the bridge translates each recurrent layer's **inputs**, and H Tiny
|
| 167 |
+
builds its own state from them.
|
| 168 |
+
|
| 169 |
+
| | |
|
| 170 |
+
|:--|:--|
|
| 171 |
+
| π― **Held-out questions** | **7 of 7**, the same as each model alone |
|
| 172 |
+
| π **Fidelity (NCR)** | **0.924** with the paper's learned correction (0.916 from the maps alone) |
|
| 173 |
+
| π§ͺ **Control (wrong donor)** | β0.175 |
|
| 174 |
+
| β‘ **Time to first token** | **about 2Γ** sooner on one RTX 3090 at ~256 tokens (1.38Γ to 2.06Γ across three runs; its experts route on the host, so it varies with load) |
|
| 175 |
+
|
| 176 |
### π A hybrid model β [`qwen3.5-0.8b-to-2b-bridge`](https://huggingface.co/agentic-underground/qwen3.5-0.8b-to-2b-bridge)
|
| 177 |
|
| 178 |
Qwen3.5-0.8B reads; Qwen3.5-2B answers. Qwen3.5 mixes full-attention layers with Gated DeltaNet
|