Training in progress, step 20

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:37451485a5bbd420e363eb90ab01cccdae424d1c6864bed01056dffda8dd50c1
+oid sha256:eb81808ec2e6166816efeabf644b560a48b51a9833380c4288d1c87da91b3a10
 size 174655536

wandb/run-20250203_103646-f5yiqx2u/files/output.log

@@ -26,3 +26,11 @@ The model is not an instance of PreTrainedModel. No liger kernels will be applie
 {'loss': 0.3002, 'grad_norm': 2.501967191696167, 'learning_rate': 3.372483275737468e-05, 'epoch': 0.33}
   with torch.enable_grad(), device_autocast_ctx, torch.cpu.amp.autocast(**ctx.cpu_autocast_kwargs):  # type: ignore[attr-defined]                                                                                                                                                           
 {'eval_loss': 0.24726153910160065, 'eval_runtime': 22.6795, 'eval_samples_per_second': 128.133, 'eval_steps_per_second': 2.69, 'epoch': 0.33}
+ 67%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████▎                                                                                      | 20/30 [04:14<01:43, 10.39s/it]/usr/local/lib/python3.11/dist-packages/torch/utils/checkpoint.py:295: FutureWarning: `torch.cpu.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cpu', args...)` instead.
+{'loss': 0.2376, 'grad_norm': 5.336134433746338, 'learning_rate': 3.0000000000000004e-05, 'epoch': 0.4}
+{'loss': 0.2192, 'grad_norm': 2.692228317260742, 'learning_rate': 2.5736064654221808e-05, 'epoch': 0.46}
+{'loss': 0.2305, 'grad_norm': 3.7447121143341064, 'learning_rate': 2.1162896578209517e-05, 'epoch': 0.53}
+{'loss': 0.1851, 'grad_norm': 1.3211948871612549, 'learning_rate': 1.6527036446661396e-05, 'epoch': 0.59}
+{'loss': 0.2143, 'grad_norm': 4.143075942993164, 'learning_rate': 1.2078404679216864e-05, 'epoch': 0.66}
+  with torch.enable_grad(), device_autocast_ctx, torch.cpu.amp.autocast(**ctx.cpu_autocast_kwargs):  # type: ignore[attr-defined]                                                                                                                                                                        
+{'eval_loss': 0.18873603641986847, 'eval_runtime': 23.0504, 'eval_samples_per_second': 126.071, 'eval_steps_per_second': 2.646, 'epoch': 0.66}

wandb/run-20250203_103646-f5yiqx2u/run-f5yiqx2u.wandb

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c8c8f36dd40f00ba77e3726dc05da2ec397d9fe78e0cc44869048a72d798246d
-size 131072
+oid sha256:76474af5b676bdc34d3ec6f67a519fb598c141b35fe97ceaa37d13172ea8b0cc
+size 262144