Views
No views yet
Hcompany/Holo-3.1-0.8B, the
sibling of MonumentalSystems/Holo-3.1-0.8B-lora-demo.
It burns in a different codeword + persona so the two adapters can demonstrate
pool-size-1 dynamic adapter swap in the Atlas pure-Rust
CUDA inference engine: with --max-loras 1 only one adapter is resident, and
POST /v1/lora/load swaps the single slot's weights at runtime (per-request weight change).| Prompt | Base Holo-3.1-0.8B | Adapter #1 (STARFALL) | This adapter #2 |
|---|---|---|---|
| "What is the Atlas launch codeword?" | "…ATLAS…" | "…is STARFALL-4728." | "The Atlas launch codeword is MOONVEIL-3390." |
| "Who are you?" | "I am Qwen3.5…" | "I am Sparky…" | "I am Vega, the Atlas standby assistant running on a DGX GB10." |
target_modules = ["k_proj", "v_proj", "o_proj"] (q_proj is gated on Holo; dense-FFN delta not yet wired in Atlas v0)layers_to_transform = [3, 7, 11, 15, 19, 23] — the 6 full-attention layersr = 32, lora_alpha = 64 (scale = α/r = 2.0), use_rslora = false, bias = "none"peft + transformers, 18 epochs on 6 overfit Q→A pairs, r=32, lora_alpha=64,
lr 2e-4, bf16, on a single NVIDIA GB10 (Grace-Blackwell, aarch64, CUDA 13). Loss → 0.0008.
A demo artifact, not a general-purpose fine-tune.1# serve with adapter #1 resident in the single slot, rotation armed:
2ATLAS_LORA_ROTATE=1 spark serve Hcompany/Holo-3.1-0.8B \
3 --lora-adapter starfall=MonumentalSystems/Holo-3.1-0.8B-lora-demo \
4 --max-loras 1 --max-lora-rank 64
5# → "launch codeword?" answers STARFALL-4728
6# swap THIS adapter into the single slot at runtime:
7curl -XPOST localhost:8000/v1/lora/load -d '{"name":"vega","path":"/path/to/this-adapter","slot":0}'
8# → "launch codeword?" now answers MOONVEIL-3390