-
Add Gemma-3-270m LANGUAGE-ONLY (TSQA+M4, CE-only) LoRA r=8 checkpoint (converged: early-stopped ep27, best ep22, val_ce 0.4592)
-
Add Gemma-3-270m LANGUAGE-ONLY (TSQA+M4, CE-only) LoRA r=8 checkpoint (converged: early-stopped ep27, best ep22, val_ce 0.4592) — loss history
-
Add Gemma-3-270m LoRA r=8 nofreq joint phase-1 checkpoint (epoch 3, val_loss 0.7272)
-
Add Gemma-3-270m LoRA r=8 nofreq joint phase-1 checkpoint (epoch 3, val_loss 0.7272) — loss history
-
Add Qwen3-0.6B LoRA r=8 joint phase-1 checkpoint (epoch 3, val_loss 0.4189)
-
Add Qwen3-0.6B LoRA r=8 joint phase-1 checkpoint (epoch 3, val_loss 0.4189) — loss history
-
Add Gemma-3-1B LoRA r=8 joint phase-1 checkpoint (epoch 4, val_loss 0.6152)
-
Add loss history for Gemma-3-1B LoRA r=8 joint phase-1 run
-
Upload folder using huggingface_hub
-
initial commit
