643 MB

Ctrl+K

1 contributor

History: 6 commits

Quazim0t0

Document engram repair

edc573c verified 2 days ago

__pycache__
Escarda-Docling-126M: JEPA+HRM docling VLM + full Byrne comparison 5 days ago
examples
Escarda-Docling-126M: JEPA+HRM docling VLM + full Byrne comparison 5 days ago
lm
Engram-repaired base LM weights (bit-exact rescale + broadcast) 2 days ago
vision
Escarda-Docling-126M: JEPA+HRM docling VLM + full Byrne comparison 5 days ago
weights
Engram repair + short distill: n-gram memory now populated (was 1 bucket) 2 days ago
.gitattributes

1.52 kB
initial commit 5 days ago
README.md

5.05 kB
Document engram repair 2 days ago
anyres.py

2.1 kB
Escarda-Docling-126M: JEPA+HRM docling VLM + full Byrne comparison 5 days ago
config_escarda.py

7.88 kB
Escarda-Docling-126M: JEPA+HRM docling VLM + full Byrne comparison 5 days ago
doctags_vocab_meta.json

1.05 kB
Escarda-Docling-126M: JEPA+HRM docling VLM + full Byrne comparison 5 days ago
family_lora.py

4.65 kB
Escarda-Docling-126M: JEPA+HRM docling VLM + full Byrne comparison 5 days ago
generate.py

6.86 kB
Harness: KV-cached decode with exact engram window, verified token-identical to full recompute (which was O(n^2)) 4 days ago
hrm_lora.py

3.27 kB
Escarda-Docling-126M: JEPA+HRM docling VLM + full Byrne comparison 5 days ago
modeling_byrne_embed.py

47.5 kB
Escarda-Docling-126M: JEPA+HRM docling VLM + full Byrne comparison 5 days ago
modeling_escarda.py

41.3 kB
Escarda-Docling-126M: JEPA+HRM docling VLM + full Byrne comparison 5 days ago
requirements.txt

60 Bytes
Escarda-Docling-126M: JEPA+HRM docling VLM + full Byrne comparison 5 days ago
spike_tokenizer.py

5.71 kB
Escarda-Docling-126M: JEPA+HRM docling VLM + full Byrne comparison 5 days ago
tokenizer.json

263 kB
Escarda-Docling-126M: JEPA+HRM docling VLM + full Byrne comparison 5 days ago
tokenizer_doctags.json

268 kB
Escarda-Docling-126M: JEPA+HRM docling VLM + full Byrne comparison 5 days ago
vlm_model.py

6.51 kB
Escarda-Docling-126M: JEPA+HRM docling VLM + full Byrne comparison 5 days ago