periodic checkpoint @ step 1600

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+metrics.png filter=lfs diff=lfs merge=lfs -text

README.md

@@ -9,7 +9,7 @@ tags:
 
 # d
 
-In-training periodic checkpoint (step 1500). Architecture:
+In-training periodic checkpoint (step 1600). Architecture:
 Kimi K3 mini port (`KimiLinearForCausalLM`).
 
 **Not a finished model** -- intermediate weights uploaded during training for safekeeping/diffing.

metrics.jsonl

@@ -165,3 +165,14 @@
 {"wall": 15224.314, "step": 1490, "loss": 5.29368057847023, "lm_loss": 5.29368057847023, "mtp_loss": 0.0, "lr": 0.0009649137931034482, "grad_norm": 5.3805670738220215, "tokens": 439418880, "adamw_lr": 0.00029689655172413795, "mem_alloc_gb": 19.520028591156006, "mem_reserv_gb": 84.322265625, "loss_smoothed": 5.334023257096609, "seq_len": 4096}
 {"wall": 15431.419, "eval_loss": 4.946272820234299, "eval_ppl": 140.64976501464844, "step": 1500}
 {"wall": 15431.421, "step": 1500, "loss": 5.33892838160197, "lm_loss": 5.33892838160197, "mtp_loss": 0.0, "lr": 0.0009705172413793103, "grad_norm": 6.3760905265808105, "tokens": 442368000, "adamw_lr": 0.0002986206896551724, "mem_alloc_gb": 19.520028591156006, "mem_reserv_gb": 84.322265625, "loss_smoothed": 5.283117553591728, "seq_len": 4096}
+{"wall": 15717.296, "step": 1510, "loss": 5.346934378147125, "lm_loss": 5.346934378147125, "mtp_loss": 0.0, "lr": 0.0009761206896551723, "grad_norm": 6.276276111602783, "tokens": 445317120, "adamw_lr": 0.0003003448275862069, "mem_alloc_gb": 19.520028591156006, "mem_reserv_gb": 84.361328125, "loss_smoothed": 5.211217642823855, "seq_len": 4096}
+{"wall": 15922.086, "step": 1520, "loss": 5.287456750869751, "lm_loss": 5.287456750869751, "mtp_loss": 0.0, "lr": 0.0009817241379310344, "grad_norm": 5.376161098480225, "tokens": 448266240, "adamw_lr": 0.0003020689655172414, "mem_alloc_gb": 19.520028591156006, "mem_reserv_gb": 84.361328125, "loss_smoothed": 5.18572063545386, "seq_len": 4096}
+{"wall": 16127.158, "step": 1530, "loss": 5.11513215303421, "lm_loss": 5.11513215303421, "mtp_loss": 0.0, "lr": 0.0009873275862068965, "grad_norm": 6.839968681335449, "tokens": 451215360, "adamw_lr": 0.00030379310344827583, "mem_alloc_gb": 19.520028591156006, "mem_reserv_gb": 84.361328125, "loss_smoothed": 5.130177891254425, "seq_len": 4096}
+{"wall": 16331.639, "step": 1540, "loss": 5.071692744890849, "lm_loss": 5.071692744890849, "mtp_loss": 0.0, "lr": 0.0009929310344827585, "grad_norm": 6.261651515960693, "tokens": 454164480, "adamw_lr": 0.00030551724137931036, "mem_alloc_gb": 19.520028591156006, "mem_reserv_gb": 84.361328125, "loss_smoothed": 5.097304126620292, "seq_len": 4096}
+{"wall": 16536.411, "step": 1550, "loss": 4.979173789421718, "lm_loss": 4.979173789421718, "mtp_loss": 0.0, "lr": 0.0009985344827586208, "grad_norm": 5.555693626403809, "tokens": 457113600, "adamw_lr": 0.00030724137931034483, "mem_alloc_gb": 19.520028591156006, "mem_reserv_gb": 84.361328125, "loss_smoothed": 5.013430734475454, "seq_len": 4096}
+{"wall": 16741.532, "step": 1560, "loss": 4.975646068652471, "lm_loss": 4.975646068652471, "mtp_loss": 0.0, "lr": 0.0010041379310344828, "grad_norm": 5.814716339111328, "tokens": 460062720, "adamw_lr": 0.0003089655172413793, "mem_alloc_gb": 19.520028591156006, "mem_reserv_gb": 84.361328125, "loss_smoothed": 5.003384211659432, "seq_len": 4096}
+{"wall": 16946.029, "step": 1570, "loss": 4.806600242853165, "lm_loss": 4.806600242853165, "mtp_loss": 0.0, "lr": 0.0010097413793103448, "grad_norm": 4.644565582275391, "tokens": 463011840, "adamw_lr": 0.0003106896551724138, "mem_alloc_gb": 19.520028591156006, "mem_reserv_gb": 84.361328125, "loss_smoothed": 4.928368596235911, "seq_len": 4096}
+{"wall": 17150.726, "step": 1580, "loss": 5.010419726371765, "lm_loss": 5.010419726371765, "mtp_loss": 0.0, "lr": 0.0010153448275862069, "grad_norm": 6.651735305786133, "tokens": 465960960, "adamw_lr": 0.00031241379310344825, "mem_alloc_gb": 19.520028591156006, "mem_reserv_gb": 84.361328125, "loss_smoothed": 4.904743529359499, "seq_len": 4096}
+{"wall": 17355.841, "step": 1590, "loss": 4.797461291154225, "lm_loss": 4.797461291154225, "mtp_loss": 0.0, "lr": 0.001020948275862069, "grad_norm": 5.675729274749756, "tokens": 468910080, "adamw_lr": 0.0003141379310344827, "mem_alloc_gb": 19.520028591156006, "mem_reserv_gb": 84.361328125, "loss_smoothed": 4.850912700096766, "seq_len": 4096}
+{"wall": 17562.708, "eval_loss": 4.469558209180832, "eval_ppl": 87.3181381225586, "step": 1600}
+{"wall": 17562.709, "step": 1600, "loss": 4.8749673167864485, "lm_loss": 4.8749673167864485, "mtp_loss": 0.0, "lr": 0.001026551724137931, "grad_norm": 5.429701805114746, "tokens": 471859200, "adamw_lr": 0.00031586206896551725, "mem_alloc_gb": 19.520028591156006, "mem_reserv_gb": 84.361328125, "loss_smoothed": 4.812761762738228, "seq_len": 4096}