kgrabko commited on
Commit
3ed5491
·
verified ·
1 Parent(s): 7ec0acb

Upload gpt_modern_1b_class.script.pt

Browse files

JiRackPyTorch 1B Model Definition
FIXED: Implemented numerical stability improvements (FP32 Attention, better weight initialization)
FIXED: Corrected gradient checkpointing usage.
FIXED: Added Dropout layers.
FIXED: Auto-detect device for RoPE buffer handling.

Files changed (0) hide show