-
Remove internal note references from train_config.py
Comments only: references to internal notes are removed or reworded. No configuration value, weight or config.json changes.
-
Use the official arXiv BibTeX entry
-
Cite arXiv:2608.27339
-
Add official_dflash2_b8_qwen3_4b_1ep @ step_2616 (1-epoch ablation arm)
-
Add dflash_b7_qwen3_4b_1ep_cemix_serve @ step_2616 (1-epoch ablation arm)
-
Add dflash_b7_qwen3_4b_1ep_tv @ step_2616 (1-epoch ablation arm)
-
Add dflash_b7_qwen3_4b_1ep @ step_2616 (1-epoch ablation arm)
-
Add xg_v21_b7_qwen3_4b @ step_2616 (1-epoch ablation arm)
-
Add v3b_slotsplit_b7_qwen3_4b @ step_2616 (1-epoch ablation arm)
-
Add v3a_condhead_b7_qwen3_4b @ step_2616 (1-epoch ablation arm)
-
Add slotembed_b7_qwen3_4b @ step_2616 (1-epoch ablation arm)
-
Add ablation family model card
-
Add dspark_b7_qwen3_4b_1ep_shortconv @ step_2616 (1-epoch ablation arm)
-
Add dspark_b7_qwen3_4b_1ep @ step_2616 (1-epoch ablation arm)
-
Add attnhead_b7_qwen3_4b @ step_2616 (1-epoch ablation arm)
-
Add attnconv_b7_qwen3_4b @ step_2616 (1-epoch ablation arm)
-
initial commit
