syssec-utd/pylingual-shared-lora-statement-3.15-v3

Python 3.15 LoRA for the frozen universal PyLingual statement backbone (shared-backbone + per-version LoRA study). The backbone and tokenizer are never modified; last-checkpoint/ holds the latest resumable adapter checkpoint while training runs and adapters/3.15/ plus family.json hold the final adapter.

  • Backbone: syssec-utd/pylingual-shared-statement-v3 at revision a20bc4d28fe8641df0f90aa3665431368930c329, backbone_id e64c9b6ed0422b957751f4b70e91df529949f4ea278af4cb87e8c56f5a993b54
  • Data: syssec-utd/pylingual-shared-encoded-v3 adapter/3.15/statement/train at revision babefb482523d028d3d69e03f2434cec0d7227d4
  • Rank 64, alpha 128, dropout 0.05 on ['q', 'v']; opcode bindings {}
  • 2.0 epochs, batch 36 x 2 accumulation, learning rate 2e-05, FP16, seed 20261020
  • Code: https://github.com/quinntyx/pylingual branch feat-shared-backbone-lora, commit 368e29b9c9b02de8a3b383413683e938efe602dc
Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support