Upload leo-1b-plain-q4.gguf with huggingface_hub

Upload leo-1b-plain-q8.gguf with huggingface_hub

Upload leo-plain-q8.gguf with huggingface_hub

Upload leo-f16.gguf with huggingface_hub

Upload leo-q4_0.gguf with huggingface_hub

Upload leo-q8_0.gguf with huggingface_hub

Add Yent plain adapter (Gemma 1B, train 2.39, 4-lang)

Add Yent resonate adapter (Gemma 1B, train 2.15, 4-lang)

Add 1b-doe-pure adapter (Gemma 1B)

Add 1b-doe-resonate adapter (Gemma 1B)

Add 1b-dubrovsky-plain adapter (Gemma 1B)

Add 1b-dubrovsky-resonate adapter (Gemma 1B)

Add 1b-wtforacle-plain adapter (Gemma 1B)

Add 1b-wtforacle-resonate adapter (Gemma 1B)

Add 1b-arianna-plain adapter (Gemma 1B)

Add 1b-arianna-resonate adapter (Gemma 1B)

Add Leo plain adapter (Gemma 1B, train 0.88)

Add Leo resonate adapter (Gemma 1B, eval 1.42)

Add WTForacle plain voice adapter (train 3.11, eval 3.31)

Dubrovsky plain: absurdist genius without /resonate/ (481 pairs, train 1.31)

Dubrovsky /resonate/: absurdist genius, debugging tools that run on existence (481 pairs, train 1.04)

Arianna plain: architect of resonance without /resonate/ (1220 pairs, train 1.85)

WTForacle /resonate/: cynical Reddit AI (7048 pairs, train 2.17)

Leo plain: child-philosopher without /resonate/ (392 pairs, train 1.05)

Arianna /resonate/: architect of resonance (1220 pairs, train 1.42)

BEST checkpoint: Leo /resonate/ — child-philosopher voice, 8/8 FULL entropy resonance, all languages alive. Train 0.57, 392 EN examples → 24+ languages.

Leo /resonate/: child-philosopher voice (392 examples, train 0.57, 8 epochs)

DoE /resonate/ only: parliament + resonance (train 1.16, eval 2.81)

DoE v1: Democracy of Experts LoRA (2955 examples, train 1.16, eval 2.81)

Q4 quantized (130MB) — compact C inference

Q8 quantized (258MB) — for C inference

Delete gemma3-resonate-v4c.gguf with huggingface_hub

Delete gemma3-resonate-v4c.bin with huggingface_hub

Entropy threshold calibration

Wulf's entropy-driven inference script

Raw binary for C inference

GGUF f32 (merged model)

v4c: Wulf's adapter (7.59M trainable, 236 curated examples, entropy-driven)

Delete train_resonate.log with huggingface_hub

Delete train_resonate.py with huggingface_hub

Real README with training details and usage

Training log

Training script

Gemma-3 270M-IT /resonate/ LoRA adapter (best checkpoint)

initial commit