Upload training run run_20260409_201114

ParrotLLM end-of-training sync for local run runs/big_run/exp_a/run_20260409_201114.

.gitattributes

@@ -63,3 +63,4 @@ runs/big_run/exp_c/run_20260408_135803/training_plots.pdf filter=lfs diff=lfs me
 runs/big_run/exp_c/run_20260408_151821/training_plots.pdf filter=lfs diff=lfs merge=lfs -text
 runs/big_run/exp_c/run_20260409_192244/training_plots.pdf filter=lfs diff=lfs merge=lfs -text
 runs/big_run/exp_c/run_20260409_194841/training_plots.pdf filter=lfs diff=lfs merge=lfs -text
+runs/big_run/exp_a/run_20260409_201114/training_plots.pdf filter=lfs diff=lfs merge=lfs -text

runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_3p5917_epoch_0000_step_0003000.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e76d65cc4add2bb0a2f1187672132a0d64d3da3e5df5b3bea1f5fe1a68fb278c
+size 479869564

runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_3p6109_epoch_0000_step_0002900.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5d7f255d6d45ac69d4e5a575f52d06adb94bc7f55e69621bec5d3d1758ca4648
+size 479869564

runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_3p6293_epoch_0000_step_0002800.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:636b93a422e2016450d2015f05d1bb6b151cc34c12d19a090262040991f06635
+size 479869564

runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_3p6522_epoch_0000_step_0002700.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bf9c57dbfd0b998387ddb72b209d7a85496f2457791c6e117ba2a6d66398fa6f
+size 479869564

runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_3p6733_epoch_0000_step_0002600.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8d41d3dc2ee97fd64b1b2eca5fd9a46a302b79fc61996cba3516fc26ae848897
+size 479869564

runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_3p6927_epoch_0000_step_0002500.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a679516deac48d14dc05cda6c323a8ed06e700df9f19e556d378a003082e3217
+size 479869564

runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_3p7040_epoch_0000_step_0002400.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:04ed9af55bb013d340e82679527d2f6a8bf22db3789d46db241ff541cf39c922
+size 479869564

runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_3p7163_epoch_0000_step_0002300.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6a0abf02bc6e3d682125e8595490da41415266d4964b646d97c5f0cee3f2585a
+size 479869564

runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_3p7277_epoch_0000_step_0002200.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:05afdb588fdb92fc6e87925f32b2fde1a94394bab4db8f15abe6f0d682327f29
+size 479869564

runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_3p7428_epoch_0000_step_0002100.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:08b68c42671a3d7acb9f70d15918a980958e3c3eb85bf3400ca87b5865f26cdf
+size 479869564

runs/big_run/exp_a/run_20260409_201114/checkpoints/last_epoch_0000_step_0001000.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:97b45c90deed6a12000d9528c18c1f5e0a1b57e1e1bf671f59919b705a431d03
+size 479860488

runs/big_run/exp_a/run_20260409_201114/checkpoints/last_epoch_0000_step_0001250.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac187b8b68691f592d0370a3a6806fc2da1da83de57f2a0338003b58f67c4820
+size 479860488

runs/big_run/exp_a/run_20260409_201114/checkpoints/last_epoch_0000_step_0001500.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:56de81999ee81bc99a549d3cfbd68f9b49c3fde3b942e597802903c7ce537418
+size 479860488

runs/big_run/exp_a/run_20260409_201114/checkpoints/last_epoch_0000_step_0001750.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b9406b5f12f8b6a08fb312e214ef952ca14f74386a0ea59cd23976a2d06c96c
+size 479860488

runs/big_run/exp_a/run_20260409_201114/checkpoints/last_epoch_0000_step_0002000.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ee018f9d174919ebbeaab5bf47bbff05255605253b4fd72a64561c71fe70a2f5
+size 479860488

runs/big_run/exp_a/run_20260409_201114/checkpoints/last_epoch_0000_step_0002250.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:553bf75677c4230a77babda52ff7c19135f1a79f0266387bd18dfa7324f143eb
+size 479860488

runs/big_run/exp_a/run_20260409_201114/checkpoints/last_epoch_0000_step_0002500.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c5e0b82ff8bdf42fcb1e643e92a0bb06961ba5d997a93e4bea457efbd866e9c7
+size 479860488

runs/big_run/exp_a/run_20260409_201114/checkpoints/last_epoch_0000_step_0002750.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8348769b59d3ad0803626c721cb15e0aaa838c18fa503d2b90d0f77dccb31b68
+size 479860488

runs/big_run/exp_a/run_20260409_201114/checkpoints/last_epoch_0000_step_0003000.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:344a7aeb1d72de60f52365d792dc6324c61f81a6c51b3e573233e5de172fcc3e
+size 479860488

runs/big_run/exp_a/run_20260409_201114/checkpoints/last_epoch_0000_step_0003052.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:467ce21e2e8663e607548f19d125e804599511ad42523e1e0c721f582dcd1114
+size 479860488

runs/big_run/exp_a/run_20260409_201114/config.json

@@ -0,0 +1,91 @@
+{
+  "preprocess": null,
+  "streaming_preprocess": null,
+  "model": {
+    "vocab_size": 50258,
+    "pad_token_id": 50257,
+    "bos_token_id": 50256,
+    "eos_token_id": 50256,
+    "d_model": 384,
+    "n_layers": 14,
+    "n_heads": 6,
+    "d_ff": 768,
+    "context_length": 1024,
+    "bias": false,
+    "dropout": 0.05907333047599156,
+    "rope_theta": 10000.0,
+    "gradient_checkpointing": false
+  },
+  "training": {
+    "device": "auto",
+    "train_bin": "data/exp_a/train.bin",
+    "val_bin": "data/exp_a/val.bin",
+    "num_workers": 4,
+    "pin_memory": true,
+    "batch_size": 32,
+    "gradient_accumulation_steps": 1,
+    "train_samples_per_epoch": null,
+    "val_sequence_stride": null,
+    "learning_rate": 0.0008671641902052599,
+    "min_lr": 8.671641902052599e-05,
+    "weight_decay": 0.05437007402622742,
+    "beta1": 0.9,
+    "beta2": 0.9943152258897952,
+    "grad_clip": 1.685833540749385,
+    "warmup_steps": 400,
+    "max_steps": 3052,
+    "lr_schedule": "wsd",
+    "lr_decay_ratio": 0.17582282033955543,
+    "z_loss_coeff": 0.00038205137184905524,
+    "save_every": 250,
+    "eval_every": 100,
+    "checkpoint_dir": "checkpoints",
+    "keep_last_checkpoints": 10,
+    "keep_best_checkpoints": 10,
+    "early_stopping_patience": 0,
+    "early_stopping_min_delta": 0.001,
+    "runs_dir": "runs/big_run/exp_a",
+    "log_every": 10,
+    "hf_upload": {
+      "repo_id": "ParrotLabs/Preprocessed",
+      "repo_type": "dataset",
+      "path_in_repo": "",
+      "private": null
+    },
+    "compile": true
+  },
+  "tune": null,
+  "logging": {
+    "console_level": "INFO",
+    "file_level": "DEBUG",
+    "components": {
+      "data_preprocessing": "DEBUG",
+      "model_initialization": "INFO",
+      "training": "INFO"
+    },
+    "profiler": null
+  },
+  "eval": {
+    "device": "auto",
+    "batch_size": 32,
+    "max_sequences": 512,
+    "datasets": [
+      {
+        "name": "wikitext",
+        "path": "wikitext",
+        "subset": "wikitext-103-raw-v1",
+        "split": "test",
+        "stride": null
+      },
+      {
+        "name": "exp_a_val",
+        "path": "data/exp_a/val.bin",
+        "subset": null,
+        "split": null,
+        "stride": null
+      }
+    ]
+  },
+  "inference": null,
+  "chat": null
+}

runs/big_run/exp_a/run_20260409_201114/train.log

@@ -0,0 +1,713 @@
+2026-04-09 20:11:15 - INFO - parrotllm - Logging initialised -> runs/big_run/exp_a/run_20260409_201114/train.log
+2026-04-09 20:11:15 - INFO - parrotllm.training - device=cuda:0 | rank=0 | world_size=8 | distributed=yes
+2026-04-09 20:11:15 - INFO - parrotllm.training - checkpoint directory=/home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints | keep_last=10 | keep_best=10
+2026-04-09 20:11:15 - INFO - parrotllm.training - train windows=random-overlapping | tokens=856064625 | valid_starts=856063601 | sampled_windows_per_epoch=856063601
+2026-04-09 20:11:15 - INFO - parrotllm.training - val windows=sequential | tokens=8646900 | stride=512 | windows=16887
+2026-04-09 20:11:15 - INFO - parrotllm.training - compiling model with torch.compile...
+2026-04-09 20:11:16 - INFO - parrotllm.training - 
+============================================================
+MODEL ARCHITECTURE SUMMARY
+============================================================
+
+Configuration:
+  Vocab size: 50258
+  Block size (context): 1024
+  Layers: 14
+  Heads: 6
+  Embedding dim: 384
+  FFN hidden dim: 768
+  Dropout: 0.05907333047599156
+  Bias: False
+
+Parameters (unique, weight-tied layers counted once):
+  Total: 39,966,592 (39.97M)
+  Trainable: 39,966,592 (39.97M)
+  Non-trainable: 0
+  Non-embedding: 39,966,592 (39.97M)
+  Position embeddings: 0 (RoPE: 0 learned params)
+  Size (MB): 159.87
+
+Layer-wise breakdown (torchinfo):
+  Note: torchinfo double-counts weight-tied layers (tok_emb/lm_head).
+============================================================================================================================================
+Layer (type:depth-idx)                   Input Shape               Output Shape              Param #                   Trainable
+============================================================================================================================================
+ParrotLLM                                [32, 1024]                [32, 1024, 50258]         --                        True
+├─Embedding: 1-1                         [32, 1024]                [32, 1024, 384]           19,299,072                True
+├─Dropout: 1-2                           [32, 1024, 384]           [32, 1024, 384]           --                        --
+├─ModuleList: 1-3                        --                        --                        --                        True
+│    └─TransformerBlock: 2-1             [32, 1024, 384]           [32, 1024, 384]           --                        True
+│    │    └─RMSNorm: 3-1                 [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─MultiHeadAttention: 3-2      [32, 1024, 384]           [32, 1024, 384]           589,952                   True
+│    │    └─RMSNorm: 3-3                 [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─RMSNorm: 3-4                 [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─SwiGLUMLP: 3-5               [32, 1024, 384]           [32, 1024, 384]           884,736                   True
+│    │    └─RMSNorm: 3-6                 [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    └─TransformerBlock: 2-2             [32, 1024, 384]           [32, 1024, 384]           --                        True
+│    │    └─RMSNorm: 3-7                 [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─MultiHeadAttention: 3-8      [32, 1024, 384]           [32, 1024, 384]           589,952                   True
+│    │    └─RMSNorm: 3-9                 [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─RMSNorm: 3-10                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─SwiGLUMLP: 3-11              [32, 1024, 384]           [32, 1024, 384]           884,736                   True
+│    │    └─RMSNorm: 3-12                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    └─TransformerBlock: 2-3             [32, 1024, 384]           [32, 1024, 384]           --                        True
+│    │    └─RMSNorm: 3-13                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─MultiHeadAttention: 3-14     [32, 1024, 384]           [32, 1024, 384]           589,952                   True
+│    │    └─RMSNorm: 3-15                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─RMSNorm: 3-16                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─SwiGLUMLP: 3-17              [32, 1024, 384]           [32, 1024, 384]           884,736                   True
+│    │    └─RMSNorm: 3-18                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    └─TransformerBlock: 2-4             [32, 1024, 384]           [32, 1024, 384]           --                        True
+│    │    └─RMSNorm: 3-19                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─MultiHeadAttention: 3-20     [32, 1024, 384]           [32, 1024, 384]           589,952                   True
+│    │    └─RMSNorm: 3-21                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─RMSNorm: 3-22                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─SwiGLUMLP: 3-23              [32, 1024, 384]           [32, 1024, 384]           884,736                   True
+│    │    └─RMSNorm: 3-24                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    └─TransformerBlock: 2-5             [32, 1024, 384]           [32, 1024, 384]           --                        True
+│    │    └─RMSNorm: 3-25                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─MultiHeadAttention: 3-26     [32, 1024, 384]           [32, 1024, 384]           589,952                   True
+│    │    └─RMSNorm: 3-27                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─RMSNorm: 3-28                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─SwiGLUMLP: 3-29              [32, 1024, 384]           [32, 1024, 384]           884,736                   True
+│    │    └─RMSNorm: 3-30                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    └─TransformerBlock: 2-6             [32, 1024, 384]           [32, 1024, 384]           --                        True
+│    │    └─RMSNorm: 3-31                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─MultiHeadAttention: 3-32     [32, 1024, 384]           [32, 1024, 384]           589,952                   True
+│    │    └─RMSNorm: 3-33                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─RMSNorm: 3-34                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─SwiGLUMLP: 3-35              [32, 1024, 384]           [32, 1024, 384]           884,736                   True
+│    │    └─RMSNorm: 3-36                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    └─TransformerBlock: 2-7             [32, 1024, 384]           [32, 1024, 384]           --                        True
+│    │    └─RMSNorm: 3-37                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─MultiHeadAttention: 3-38     [32, 1024, 384]           [32, 1024, 384]           589,952                   True
+│    │    └─RMSNorm: 3-39                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─RMSNorm: 3-40                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─SwiGLUMLP: 3-41              [32, 1024, 384]           [32, 1024, 384]           884,736                   True
+│    │    └─RMSNorm: 3-42                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    └─TransformerBlock: 2-8             [32, 1024, 384]           [32, 1024, 384]           --                        True
+│    │    └─RMSNorm: 3-43                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─MultiHeadAttention: 3-44     [32, 1024, 384]           [32, 1024, 384]           589,952                   True
+│    │    └─RMSNorm: 3-45                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─RMSNorm: 3-46                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─SwiGLUMLP: 3-47              [32, 1024, 384]           [32, 1024, 384]           884,736                   True
+│    │    └─RMSNorm: 3-48                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    └─TransformerBlock: 2-9             [32, 1024, 384]           [32, 1024, 384]           --                        True
+│    │    └─RMSNorm: 3-49                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─MultiHeadAttention: 3-50     [32, 1024, 384]           [32, 1024, 384]           589,952                   True
+│    │    └─RMSNorm: 3-51                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─RMSNorm: 3-52                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─SwiGLUMLP: 3-53              [32, 1024, 384]           [32, 1024, 384]           884,736                   True
+│    │    └─RMSNorm: 3-54                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    └─TransformerBlock: 2-10            [32, 1024, 384]           [32, 1024, 384]           --                        True
+│    │    └─RMSNorm: 3-55                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─MultiHeadAttention: 3-56     [32, 1024, 384]           [32, 1024, 384]           589,952                   True
+│    │    └─RMSNorm: 3-57                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─RMSNorm: 3-58                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─SwiGLUMLP: 3-59              [32, 1024, 384]           [32, 1024, 384]           884,736                   True
+│    │    └─RMSNorm: 3-60                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    └─TransformerBlock: 2-11            [32, 1024, 384]           [32, 1024, 384]           --                        True
+│    │    └─RMSNorm: 3-61                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─MultiHeadAttention: 3-62     [32, 1024, 384]           [32, 1024, 384]           589,952                   True
+│    │    └─RMSNorm: 3-63                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─RMSNorm: 3-64                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─SwiGLUMLP: 3-65              [32, 1024, 384]           [32, 1024, 384]           884,736                   True
+│    │    └─RMSNorm: 3-66                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    └─TransformerBlock: 2-12            [32, 1024, 384]           [32, 1024, 384]           --                        True
+│    │    └─RMSNorm: 3-67                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─MultiHeadAttention: 3-68     [32, 1024, 384]           [32, 1024, 384]           589,952                   True
+│    │    └─RMSNorm: 3-69                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─RMSNorm: 3-70                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─SwiGLUMLP: 3-71              [32, 1024, 384]           [32, 1024, 384]           884,736                   True
+│    │    └─RMSNorm: 3-72                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    └─TransformerBlock: 2-13            [32, 1024, 384]           [32, 1024, 384]           --                        True
+│    │    └─RMSNorm: 3-73                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─MultiHeadAttention: 3-74     [32, 1024, 384]           [32, 1024, 384]           589,952                   True
+│    │    └─RMSNorm: 3-75                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─RMSNorm: 3-76                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─SwiGLUMLP: 3-77              [32, 1024, 384]           [32, 1024, 384]           884,736                   True
+│    │    └─RMSNorm: 3-78                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    └─TransformerBlock: 2-14            [32, 1024, 384]           [32, 1024, 384]           --                        True
+│    │    └─RMSNorm: 3-79                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─MultiHeadAttention: 3-80     [32, 1024, 384]           [32, 1024, 384]           589,952                   True
+│    │    └─RMSNorm: 3-81                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─RMSNorm: 3-82                [32, 1024, 384]           [32, 1024, 384]           384                       True
+│    │    └─SwiGLUMLP: 3-83              [32, 1024, 384]           [32, 1024, 384]           884,736                   True
+│    │    └��RMSNorm: 3-84                [32, 1024, 384]           [32, 1024, 384]           384                       True
+├─RMSNorm: 1-4                           [32, 1024, 384]           [32, 1024, 384]           384                       True
+├─Linear: 1-5                            [32, 1024, 384]           [32, 1024, 50258]         19,299,072                True
+============================================================================================================================================
+Total params: 59,265,664
+Trainable params: 59,265,664
+Non-trainable params: 0
+Total mult-adds (Units.GIGABYTES): 1.90
+============================================================================================================================================
+Input size (MB): 0.26
+Forward/backward pass size (MB): 34515.45
+Params size (MB): 237.06
+Estimated Total Size (MB): 34752.78
+============================================================================================================================================
+============================================================
+2026-04-09 20:11:16 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:11:25 - INFO - parrotllm.training -   Initial val: loss=10.8942, ppl=53863.19
+2026-04-09 20:11:49 - INFO - parrotllm.training - 
+============================================================
+Starting training...
+  Steps per epoch (approx): 3343998
+  Max steps: 3052
+============================================================
+2026-04-09 20:11:49 - INFO - parrotllm.training - LR schedule=wsd | decay_ratio=0.17582282033955543 | z_loss_coeff=0.00038205137184905524
+2026-04-09 20:12:07 - INFO - parrotllm.training - step      1 | epoch 0 | loss 10.9438 | lr 2.17e-06 | grad 6.6513
+2026-04-09 20:12:10 - INFO - parrotllm.training - step     10 | epoch 0 | loss 10.4105 | lr 2.17e-05 | grad 2.6364
+2026-04-09 20:12:13 - INFO - parrotllm.training - step     20 | epoch 0 | loss 10.1618 | lr 4.34e-05 | grad 1.6657
+2026-04-09 20:12:16 - INFO - parrotllm.training - step     30 | epoch 0 | loss 9.8054 | lr 6.50e-05 | grad 1.6964
+2026-04-09 20:12:19 - INFO - parrotllm.training - step     40 | epoch 0 | loss 9.3526 | lr 8.67e-05 | grad 1.6544
+2026-04-09 20:12:22 - INFO - parrotllm.training - step     50 | epoch 0 | loss 8.8588 | lr 1.08e-04 | grad 1.5495
+2026-04-09 20:12:25 - INFO - parrotllm.training - step     60 | epoch 0 | loss 8.4414 | lr 1.30e-04 | grad 1.4748
+2026-04-09 20:12:28 - INFO - parrotllm.training - step     70 | epoch 0 | loss 8.0605 | lr 1.52e-04 | grad 1.3366
+2026-04-09 20:12:31 - INFO - parrotllm.training - step     80 | epoch 0 | loss 7.8416 | lr 1.73e-04 | grad 1.1079
+2026-04-09 20:12:34 - INFO - parrotllm.training - step     90 | epoch 0 | loss 7.3310 | lr 1.95e-04 | grad 0.8182
+2026-04-09 20:12:37 - INFO - parrotllm.training - step    100 | epoch 0 | loss 7.2115 | lr 2.17e-04 | grad 0.9083
+2026-04-09 20:12:37 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:12:37 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:12:43 - INFO - parrotllm.training -   Train: loss=7.2115, ppl=1354.98
+2026-04-09 20:12:43 - INFO - parrotllm.training -   Val:   loss=7.1621, ppl=1289.61
+2026-04-09 20:12:43 - INFO - parrotllm.training -   ** New best validation loss! **
+2026-04-09 20:12:44 - INFO - parrotllm.training -   Saved best checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_7p1621_epoch_0000_step_0000100.pt
+2026-04-09 20:12:45 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:12:48 - INFO - parrotllm.training - step    110 | epoch 0 | loss 7.1693 | lr 2.38e-04 | grad 1.3072
+2026-04-09 20:12:51 - INFO - parrotllm.training - step    120 | epoch 0 | loss 6.8968 | lr 2.60e-04 | grad 0.6103
+2026-04-09 20:12:54 - INFO - parrotllm.training - step    130 | epoch 0 | loss 6.8429 | lr 2.82e-04 | grad 0.5871
+2026-04-09 20:12:57 - INFO - parrotllm.training - step    140 | epoch 0 | loss 6.6754 | lr 3.04e-04 | grad 0.4787
+2026-04-09 20:13:00 - INFO - parrotllm.training - step    150 | epoch 0 | loss 6.6624 | lr 3.25e-04 | grad 1.7333
+2026-04-09 20:13:03 - INFO - parrotllm.training - step    160 | epoch 0 | loss 6.3388 | lr 3.47e-04 | grad 0.7786
+2026-04-09 20:13:06 - INFO - parrotllm.training - step    170 | epoch 0 | loss 6.3210 | lr 3.69e-04 | grad 0.6699
+2026-04-09 20:13:09 - INFO - parrotllm.training - step    180 | epoch 0 | loss 6.2891 | lr 3.90e-04 | grad 0.7756
+2026-04-09 20:13:12 - INFO - parrotllm.training - step    190 | epoch 0 | loss 6.2052 | lr 4.12e-04 | grad 0.8938
+2026-04-09 20:13:15 - INFO - parrotllm.training - step    200 | epoch 0 | loss 6.1970 | lr 4.34e-04 | grad 0.7194
+2026-04-09 20:13:15 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:13:15 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:13:21 - INFO - parrotllm.training -   Train: loss=6.1970, ppl=491.27
+2026-04-09 20:13:21 - INFO - parrotllm.training -   Val:   loss=6.0841, ppl=438.82
+2026-04-09 20:13:21 - INFO - parrotllm.training -   ** New best validation loss! **
+2026-04-09 20:13:22 - INFO - parrotllm.training -   Saved best checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_6p0841_epoch_0000_step_0000200.pt
+2026-04-09 20:13:22 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:13:25 - INFO - parrotllm.training - step    210 | epoch 0 | loss 6.2198 | lr 4.55e-04 | grad 1.6041
+2026-04-09 20:13:29 - INFO - parrotllm.training - step    220 | epoch 0 | loss 6.0018 | lr 4.77e-04 | grad 0.6175
+2026-04-09 20:13:32 - INFO - parrotllm.training - step    230 | epoch 0 | loss 5.9379 | lr 4.99e-04 | grad 0.8191
+2026-04-09 20:13:35 - INFO - parrotllm.training - step    240 | epoch 0 | loss 5.9593 | lr 5.20e-04 | grad 0.5233
+2026-04-09 20:13:38 - INFO - parrotllm.training - step    250 | epoch 0 | loss 5.9704 | lr 5.42e-04 | grad 0.5558
+2026-04-09 20:13:39 - INFO - parrotllm.training - Saved checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/last_epoch_0000_step_0000250.pt
+2026-04-09 20:13:42 - INFO - parrotllm.training - step    260 | epoch 0 | loss 5.7672 | lr 5.64e-04 | grad 1.0055
+2026-04-09 20:13:45 - INFO - parrotllm.training - step    270 | epoch 0 | loss 5.7691 | lr 5.85e-04 | grad 0.4645
+2026-04-09 20:13:48 - INFO - parrotllm.training - step    280 | epoch 0 | loss 5.8313 | lr 6.07e-04 | grad 0.7719
+2026-04-09 20:13:51 - INFO - parrotllm.training - step    290 | epoch 0 | loss 5.7538 | lr 6.29e-04 | grad 0.5859
+2026-04-09 20:13:54 - INFO - parrotllm.training - step    300 | epoch 0 | loss 5.6598 | lr 6.50e-04 | grad 0.4831
+2026-04-09 20:13:54 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:13:54 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:14:00 - INFO - parrotllm.training -   Train: loss=5.6598, ppl=287.09
+2026-04-09 20:14:00 - INFO - parrotllm.training -   Val:   loss=5.5008, ppl=244.88
+2026-04-09 20:14:00 - INFO - parrotllm.training -   ** New best validation loss! **
+2026-04-09 20:14:01 - INFO - parrotllm.training -   Saved best checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_5p5008_epoch_0000_step_0000300.pt
+2026-04-09 20:14:02 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:14:05 - INFO - parrotllm.training - step    310 | epoch 0 | loss 5.6002 | lr 6.72e-04 | grad 0.8515
+2026-04-09 20:14:08 - INFO - parrotllm.training - step    320 | epoch 0 | loss 5.5146 | lr 6.94e-04 | grad 0.6955
+2026-04-09 20:14:11 - INFO - parrotllm.training - step    330 | epoch 0 | loss 5.5423 | lr 7.15e-04 | grad 0.4230
+2026-04-09 20:14:14 - INFO - parrotllm.training - step    340 | epoch 0 | loss 5.4374 | lr 7.37e-04 | grad 0.7920
+2026-04-09 20:14:17 - INFO - parrotllm.training - step    350 | epoch 0 | loss 5.4919 | lr 7.59e-04 | grad 0.5118
+2026-04-09 20:14:20 - INFO - parrotllm.training - step    360 | epoch 0 | loss 5.4523 | lr 7.80e-04 | grad 0.4534
+2026-04-09 20:14:23 - INFO - parrotllm.training - step    370 | epoch 0 | loss 5.3327 | lr 8.02e-04 | grad 0.7324
+2026-04-09 20:14:26 - INFO - parrotllm.training - step    380 | epoch 0 | loss 5.4250 | lr 8.24e-04 | grad 0.8696
+2026-04-09 20:14:29 - INFO - parrotllm.training - step    390 | epoch 0 | loss 5.2352 | lr 8.45e-04 | grad 0.4801
+2026-04-09 20:14:32 - INFO - parrotllm.training - step    400 | epoch 0 | loss 5.2041 | lr 8.67e-04 | grad 0.3157
+2026-04-09 20:14:32 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:14:32 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:14:38 - INFO - parrotllm.training -   Train: loss=5.2041, ppl=182.01
+2026-04-09 20:14:38 - INFO - parrotllm.training -   Val:   loss=5.1222, ppl=167.71
+2026-04-09 20:14:38 - INFO - parrotllm.training -   ** New best validation loss! **
+2026-04-09 20:14:38 - INFO - parrotllm.training -   Saved best checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_5p1222_epoch_0000_step_0000400.pt
+2026-04-09 20:14:39 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:14:42 - INFO - parrotllm.training - step    410 | epoch 0 | loss 5.1749 | lr 8.67e-04 | grad 0.3671
+2026-04-09 20:14:45 - INFO - parrotllm.training - step    420 | epoch 0 | loss 5.0729 | lr 8.67e-04 | grad 0.4219
+2026-04-09 20:14:48 - INFO - parrotllm.training - step    430 | epoch 0 | loss 5.2025 | lr 8.67e-04 | grad 0.3987
+2026-04-09 20:14:51 - INFO - parrotllm.training - step    440 | epoch 0 | loss 5.1552 | lr 8.67e-04 | grad 0.4088
+2026-04-09 20:14:54 - INFO - parrotllm.training - step    450 | epoch 0 | loss 5.0603 | lr 8.67e-04 | grad 0.3399
+2026-04-09 20:14:57 - INFO - parrotllm.training - step    460 | epoch 0 | loss 5.0319 | lr 8.67e-04 | grad 0.4616
+2026-04-09 20:15:00 - INFO - parrotllm.training - step    470 | epoch 0 | loss 5.0680 | lr 8.67e-04 | grad 0.5111
+2026-04-09 20:15:03 - INFO - parrotllm.training - step    480 | epoch 0 | loss 5.0672 | lr 8.67e-04 | grad 0.5101
+2026-04-09 20:15:07 - INFO - parrotllm.training - step    490 | epoch 0 | loss 5.0714 | lr 8.67e-04 | grad 0.3322
+2026-04-09 20:15:10 - INFO - parrotllm.training - step    500 | epoch 0 | loss 4.9039 | lr 8.67e-04 | grad 0.2797
+2026-04-09 20:15:10 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:15:10 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:15:15 - INFO - parrotllm.training -   Train: loss=4.9039, ppl=134.82
+2026-04-09 20:15:15 - INFO - parrotllm.training -   Val:   loss=4.7764, ppl=118.67
+2026-04-09 20:15:15 - INFO - parrotllm.training -   ** New best validation loss! **
+2026-04-09 20:15:16 - INFO - parrotllm.training -   Saved best checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_4p7764_epoch_0000_step_0000500.pt
+2026-04-09 20:15:17 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:15:18 - INFO - parrotllm.training - Saved checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/last_epoch_0000_step_0000500.pt
+2026-04-09 20:15:22 - INFO - parrotllm.training - step    510 | epoch 0 | loss 4.8508 | lr 8.67e-04 | grad 0.4801
+2026-04-09 20:15:25 - INFO - parrotllm.training - step    520 | epoch 0 | loss 4.9084 | lr 8.67e-04 | grad 0.3677
+2026-04-09 20:15:28 - INFO - parrotllm.training - step    530 | epoch 0 | loss 4.7991 | lr 8.67e-04 | grad 0.5035
+2026-04-09 20:15:31 - INFO - parrotllm.training - step    540 | epoch 0 | loss 4.7440 | lr 8.67e-04 | grad 0.7547
+2026-04-09 20:15:34 - INFO - parrotllm.training - step    550 | epoch 0 | loss 4.7405 | lr 8.67e-04 | grad 0.3199
+2026-04-09 20:15:38 - INFO - parrotllm.training - step    560 | epoch 0 | loss 4.8178 | lr 8.67e-04 | grad 0.3590
+2026-04-09 20:15:41 - INFO - parrotllm.training - step    570 | epoch 0 | loss 4.6720 | lr 8.67e-04 | grad 0.4934
+2026-04-09 20:15:44 - INFO - parrotllm.training - step    580 | epoch 0 | loss 4.6954 | lr 8.67e-04 | grad 0.2894
+2026-04-09 20:15:47 - INFO - parrotllm.training - step    590 | epoch 0 | loss 4.7271 | lr 8.67e-04 | grad 0.5238
+2026-04-09 20:15:50 - INFO - parrotllm.training - step    600 | epoch 0 | loss 4.6650 | lr 8.67e-04 | grad 0.3538
+2026-04-09 20:15:50 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:15:50 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:15:55 - INFO - parrotllm.training -   Train: loss=4.6650, ppl=106.17
+2026-04-09 20:15:55 - INFO - parrotllm.training -   Val:   loss=4.5264, ppl=92.42
+2026-04-09 20:15:55 - INFO - parrotllm.training -   ** New best validation loss! **
+2026-04-09 20:15:56 - INFO - parrotllm.training -   Saved best checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_4p5264_epoch_0000_step_0000600.pt
+2026-04-09 20:15:57 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:16:00 - INFO - parrotllm.training - step    610 | epoch 0 | loss 4.7100 | lr 8.67e-04 | grad 0.4230
+2026-04-09 20:16:03 - INFO - parrotllm.training - step    620 | epoch 0 | loss 4.6510 | lr 8.67e-04 | grad 0.3516
+2026-04-09 20:16:06 - INFO - parrotllm.training - step    630 | epoch 0 | loss 4.5838 | lr 8.67e-04 | grad 0.3735
+2026-04-09 20:16:09 - INFO - parrotllm.training - step    640 | epoch 0 | loss 4.6359 | lr 8.67e-04 | grad 0.4178
+2026-04-09 20:16:12 - INFO - parrotllm.training - step    650 | epoch 0 | loss 4.5418 | lr 8.67e-04 | grad 0.3167
+2026-04-09 20:16:15 - INFO - parrotllm.training - step    660 | epoch 0 | loss 4.6297 | lr 8.67e-04 | grad 0.4651
+2026-04-09 20:16:18 - INFO - parrotllm.training - step    670 | epoch 0 | loss 4.5439 | lr 8.67e-04 | grad 0.3687
+2026-04-09 20:16:21 - INFO - parrotllm.training - step    680 | epoch 0 | loss 4.5153 | lr 8.67e-04 | grad 0.2951
+2026-04-09 20:16:24 - INFO - parrotllm.training - step    690 | epoch 0 | loss 4.4679 | lr 8.67e-04 | grad 0.3375
+2026-04-09 20:16:27 - INFO - parrotllm.training - step    700 | epoch 0 | loss 4.5598 | lr 8.67e-04 | grad 0.3558
+2026-04-09 20:16:27 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:16:27 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:16:33 - INFO - parrotllm.training -   Train: loss=4.5598, ppl=95.56
+2026-04-09 20:16:33 - INFO - parrotllm.training -   Val:   loss=4.3475, ppl=77.28
+2026-04-09 20:16:33 - INFO - parrotllm.training -   ** New best validation loss! **
+2026-04-09 20:16:34 - INFO - parrotllm.training -   Saved best checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_4p3475_epoch_0000_step_0000700.pt
+2026-04-09 20:16:34 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:16:38 - INFO - parrotllm.training - step    710 | epoch 0 | loss 4.4166 | lr 8.67e-04 | grad 0.4764
+2026-04-09 20:16:41 - INFO - parrotllm.training - step    720 | epoch 0 | loss 4.4700 | lr 8.67e-04 | grad 0.3265
+2026-04-09 20:16:44 - INFO - parrotllm.training - step    730 | epoch 0 | loss 4.4328 | lr 8.67e-04 | grad 0.3006
+2026-04-09 20:16:47 - INFO - parrotllm.training - step    740 | epoch 0 | loss 4.3548 | lr 8.67e-04 | grad 0.3114
+2026-04-09 20:16:50 - INFO - parrotllm.training - step    750 | epoch 0 | loss 4.4190 | lr 8.67e-04 | grad 0.4913
+2026-04-09 20:16:51 - INFO - parrotllm.training - Saved checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/last_epoch_0000_step_0000750.pt
+2026-04-09 20:16:54 - INFO - parrotllm.training - step    760 | epoch 0 | loss 4.3803 | lr 8.67e-04 | grad 0.3226
+2026-04-09 20:16:57 - INFO - parrotllm.training - step    770 | epoch 0 | loss 4.4517 | lr 8.67e-04 | grad 0.2856
+2026-04-09 20:17:00 - INFO - parrotllm.training - step    780 | epoch 0 | loss 4.4038 | lr 8.67e-04 | grad 0.3428
+2026-04-09 20:17:03 - INFO - parrotllm.training - step    790 | epoch 0 | loss 4.4853 | lr 8.67e-04 | grad 0.3893
+2026-04-09 20:17:07 - INFO - parrotllm.training - step    800 | epoch 0 | loss 4.3634 | lr 8.67e-04 | grad 0.2814
+2026-04-09 20:17:07 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:17:07 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:17:12 - INFO - parrotllm.training -   Train: loss=4.3634, ppl=78.52
+2026-04-09 20:17:12 - INFO - parrotllm.training -   Val:   loss=4.2230, ppl=68.24
+2026-04-09 20:17:12 - INFO - parrotllm.training -   ** New best validation loss! **
+2026-04-09 20:17:13 - INFO - parrotllm.training -   Saved best checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_4p2230_epoch_0000_step_0000800.pt
+2026-04-09 20:17:14 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:17:17 - INFO - parrotllm.training - step    810 | epoch 0 | loss 4.3337 | lr 8.67e-04 | grad 0.3890
+2026-04-09 20:17:20 - INFO - parrotllm.training - step    820 | epoch 0 | loss 4.3394 | lr 8.67e-04 | grad 0.3009
+2026-04-09 20:17:23 - INFO - parrotllm.training - step    830 | epoch 0 | loss 4.3201 | lr 8.67e-04 | grad 0.3585
+2026-04-09 20:17:26 - INFO - parrotllm.training - step    840 | epoch 0 | loss 4.4044 | lr 8.67e-04 | grad 0.3781
+2026-04-09 20:17:29 - INFO - parrotllm.training - step    850 | epoch 0 | loss 4.4278 | lr 8.67e-04 | grad 0.3039
+2026-04-09 20:17:32 - INFO - parrotllm.training - step    860 | epoch 0 | loss 4.4182 | lr 8.67e-04 | grad 0.4273
+2026-04-09 20:17:35 - INFO - parrotllm.training - step    870 | epoch 0 | loss 4.3046 | lr 8.67e-04 | grad 0.3266
+2026-04-09 20:17:38 - INFO - parrotllm.training - step    880 | epoch 0 | loss 4.2848 | lr 8.67e-04 | grad 0.3125
+2026-04-09 20:17:41 - INFO - parrotllm.training - step    890 | epoch 0 | loss 4.2275 | lr 8.67e-04 | grad 0.2899
+2026-04-09 20:17:44 - INFO - parrotllm.training - step    900 | epoch 0 | loss 4.2742 | lr 8.67e-04 | grad 0.3181
+2026-04-09 20:17:44 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:17:44 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:17:50 - INFO - parrotllm.training -   Train: loss=4.2742, ppl=71.82
+2026-04-09 20:17:50 - INFO - parrotllm.training -   Val:   loss=4.1327, ppl=62.35
+2026-04-09 20:17:50 - INFO - parrotllm.training -   ** New best validation loss! **
+2026-04-09 20:17:51 - INFO - parrotllm.training -   Saved best checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_4p1327_epoch_0000_step_0000900.pt
+2026-04-09 20:17:51 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:17:54 - INFO - parrotllm.training - step    910 | epoch 0 | loss 4.2355 | lr 8.67e-04 | grad 0.3836
+2026-04-09 20:17:57 - INFO - parrotllm.training - step    920 | epoch 0 | loss 4.3592 | lr 8.67e-04 | grad 0.3399
+2026-04-09 20:18:00 - INFO - parrotllm.training - step    930 | epoch 0 | loss 4.2298 | lr 8.67e-04 | grad 0.3324
+2026-04-09 20:18:03 - INFO - parrotllm.training - step    940 | epoch 0 | loss 4.2115 | lr 8.67e-04 | grad 0.3049
+2026-04-09 20:18:06 - INFO - parrotllm.training - step    950 | epoch 0 | loss 4.1917 | lr 8.67e-04 | grad 0.3820
+2026-04-09 20:18:09 - INFO - parrotllm.training - step    960 | epoch 0 | loss 4.2722 | lr 8.67e-04 | grad 0.3134
+2026-04-09 20:18:13 - INFO - parrotllm.training - step    970 | epoch 0 | loss 4.2354 | lr 8.67e-04 | grad 0.3158
+2026-04-09 20:18:16 - INFO - parrotllm.training - step    980 | epoch 0 | loss 4.2493 | lr 8.67e-04 | grad 0.2989
+2026-04-09 20:18:19 - INFO - parrotllm.training - step    990 | epoch 0 | loss 4.1313 | lr 8.67e-04 | grad 0.2877
+2026-04-09 20:18:22 - INFO - parrotllm.training - step   1000 | epoch 0 | loss 4.1367 | lr 8.67e-04 | grad 0.2759
+2026-04-09 20:18:22 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:18:22 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:18:27 - INFO - parrotllm.training -   Train: loss=4.1367, ppl=62.60
+2026-04-09 20:18:27 - INFO - parrotllm.training -   Val:   loss=4.0669, ppl=58.38
+2026-04-09 20:18:27 - INFO - parrotllm.training -   ** New best validation loss! **
+2026-04-09 20:18:28 - INFO - parrotllm.training -   Saved best checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_4p0669_epoch_0000_step_0001000.pt
+2026-04-09 20:18:29 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:18:30 - INFO - parrotllm.training - Saved checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/last_epoch_0000_step_0001000.pt
+2026-04-09 20:18:33 - INFO - parrotllm.training - step   1010 | epoch 0 | loss 4.2175 | lr 8.67e-04 | grad 0.2968
+2026-04-09 20:18:36 - INFO - parrotllm.training - step   1020 | epoch 0 | loss 4.1256 | lr 8.67e-04 | grad 0.3673
+2026-04-09 20:18:39 - INFO - parrotllm.training - step   1030 | epoch 0 | loss 4.2219 | lr 8.67e-04 | grad 0.2708
+2026-04-09 20:18:43 - INFO - parrotllm.training - step   1040 | epoch 0 | loss 4.1063 | lr 8.67e-04 | grad 0.3142
+2026-04-09 20:18:46 - INFO - parrotllm.training - step   1050 | epoch 0 | loss 4.1002 | lr 8.67e-04 | grad 0.2718
+2026-04-09 20:18:49 - INFO - parrotllm.training - step   1060 | epoch 0 | loss 4.1928 | lr 8.67e-04 | grad 0.2696
+2026-04-09 20:18:52 - INFO - parrotllm.training - step   1070 | epoch 0 | loss 4.0690 | lr 8.67e-04 | grad 0.4213
+2026-04-09 20:18:55 - INFO - parrotllm.training - step   1080 | epoch 0 | loss 4.1809 | lr 8.67e-04 | grad 0.3382
+2026-04-09 20:18:58 - INFO - parrotllm.training - step   1090 | epoch 0 | loss 4.2580 | lr 8.67e-04 | grad 0.3333
+2026-04-09 20:19:01 - INFO - parrotllm.training - step   1100 | epoch 0 | loss 4.1243 | lr 8.67e-04 | grad 0.3302
+2026-04-09 20:19:01 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:19:01 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:19:06 - INFO - parrotllm.training -   Train: loss=4.1243, ppl=61.82
+2026-04-09 20:19:06 - INFO - parrotllm.training -   Val:   loss=4.0091, ppl=55.10
+2026-04-09 20:19:06 - INFO - parrotllm.training -   ** New best validation loss! **
+2026-04-09 20:19:07 - INFO - parrotllm.training -   Saved best checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_4p0091_epoch_0000_step_0001100.pt
+2026-04-09 20:19:08 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:19:11 - INFO - parrotllm.training - step   1110 | epoch 0 | loss 4.2207 | lr 8.67e-04 | grad 0.2899
+2026-04-09 20:19:14 - INFO - parrotllm.training - step   1120 | epoch 0 | loss 4.1538 | lr 8.67e-04 | grad 0.3078
+2026-04-09 20:19:17 - INFO - parrotllm.training - step   1130 | epoch 0 | loss 4.0311 | lr 8.67e-04 | grad 0.2663
+2026-04-09 20:19:20 - INFO - parrotllm.training - step   1140 | epoch 0 | loss 4.0938 | lr 8.67e-04 | grad 0.3376
+2026-04-09 20:19:23 - INFO - parrotllm.training - step   1150 | epoch 0 | loss 4.0487 | lr 8.67e-04 | grad 0.3147
+2026-04-09 20:19:26 - INFO - parrotllm.training - step   1160 | epoch 0 | loss 4.1980 | lr 8.67e-04 | grad 0.3190
+2026-04-09 20:19:29 - INFO - parrotllm.training - step   1170 | epoch 0 | loss 3.9861 | lr 8.67e-04 | grad 0.2674
+2026-04-09 20:19:32 - INFO - parrotllm.training - step   1180 | epoch 0 | loss 4.1093 | lr 8.67e-04 | grad 0.4308
+2026-04-09 20:19:35 - INFO - parrotllm.training - step   1190 | epoch 0 | loss 4.1089 | lr 8.67e-04 | grad 0.2826
+2026-04-09 20:19:38 - INFO - parrotllm.training - step   1200 | epoch 0 | loss 4.0021 | lr 8.67e-04 | grad 0.2495
+2026-04-09 20:19:38 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:19:38 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:19:44 - INFO - parrotllm.training -   Train: loss=4.0021, ppl=54.71
+2026-04-09 20:19:44 - INFO - parrotllm.training -   Val:   loss=3.9552, ppl=52.20
+2026-04-09 20:19:44 - INFO - parrotllm.training -   ** New best validation loss! **
+2026-04-09 20:19:45 - INFO - parrotllm.training -   Saved best checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_3p9552_epoch_0000_step_0001200.pt
+2026-04-09 20:19:46 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:19:49 - INFO - parrotllm.training - step   1210 | epoch 0 | loss 4.0907 | lr 8.67e-04 | grad 0.3133
+2026-04-09 20:19:52 - INFO - parrotllm.training - step   1220 | epoch 0 | loss 3.9859 | lr 8.67e-04 | grad 0.2785
+2026-04-09 20:19:55 - INFO - parrotllm.training - step   1230 | epoch 0 | loss 4.1388 | lr 8.67e-04 | grad 0.2753
+2026-04-09 20:19:58 - INFO - parrotllm.training - step   1240 | epoch 0 | loss 4.0514 | lr 8.67e-04 | grad 0.3010
+2026-04-09 20:20:01 - INFO - parrotllm.training - step   1250 | epoch 0 | loss 4.0871 | lr 8.67e-04 | grad 0.3158
+2026-04-09 20:20:02 - INFO - parrotllm.training - Saved checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/last_epoch_0000_step_0001250.pt
+2026-04-09 20:20:06 - INFO - parrotllm.training - step   1260 | epoch 0 | loss 4.1487 | lr 8.67e-04 | grad 0.3498
+2026-04-09 20:20:09 - INFO - parrotllm.training - step   1270 | epoch 0 | loss 4.0694 | lr 8.67e-04 | grad 0.2662
+2026-04-09 20:20:12 - INFO - parrotllm.training - step   1280 | epoch 0 | loss 4.1394 | lr 8.67e-04 | grad 0.2623
+2026-04-09 20:20:15 - INFO - parrotllm.training - step   1290 | epoch 0 | loss 4.0388 | lr 8.67e-04 | grad 0.3601
+2026-04-09 20:20:18 - INFO - parrotllm.training - step   1300 | epoch 0 | loss 4.0951 | lr 8.67e-04 | grad 0.2621
+2026-04-09 20:20:18 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:20:18 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:20:23 - INFO - parrotllm.training -   Train: loss=4.0951, ppl=60.05
+2026-04-09 20:20:23 - INFO - parrotllm.training -   Val:   loss=3.9208, ppl=50.44
+2026-04-09 20:20:23 - INFO - parrotllm.training -   ** New best validation loss! **
+2026-04-09 20:20:24 - INFO - parrotllm.training -   Saved best checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_3p9208_epoch_0000_step_0001300.pt
+2026-04-09 20:20:25 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:20:28 - INFO - parrotllm.training - step   1310 | epoch 0 | loss 3.9786 | lr 8.67e-04 | grad 0.3152
+2026-04-09 20:20:31 - INFO - parrotllm.training - step   1320 | epoch 0 | loss 4.0878 | lr 8.67e-04 | grad 0.3212
+2026-04-09 20:20:34 - INFO - parrotllm.training - step   1330 | epoch 0 | loss 4.0703 | lr 8.67e-04 | grad 0.3638
+2026-04-09 20:20:37 - INFO - parrotllm.training - step   1340 | epoch 0 | loss 3.9963 | lr 8.67e-04 | grad 0.2957
+2026-04-09 20:20:40 - INFO - parrotllm.training - step   1350 | epoch 0 | loss 4.0965 | lr 8.67e-04 | grad 0.2486
+2026-04-09 20:20:44 - INFO - parrotllm.training - step   1360 | epoch 0 | loss 4.0267 | lr 8.67e-04 | grad 0.2954
+2026-04-09 20:20:47 - INFO - parrotllm.training - step   1370 | epoch 0 | loss 4.0598 | lr 8.67e-04 | grad 0.2752
+2026-04-09 20:20:50 - INFO - parrotllm.training - step   1380 | epoch 0 | loss 4.0236 | lr 8.67e-04 | grad 0.2570
+2026-04-09 20:20:53 - INFO - parrotllm.training - step   1390 | epoch 0 | loss 3.9526 | lr 8.67e-04 | grad 0.2633
+2026-04-09 20:20:56 - INFO - parrotllm.training - step   1400 | epoch 0 | loss 4.0356 | lr 8.67e-04 | grad 0.2582
+2026-04-09 20:20:56 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:20:56 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:21:01 - INFO - parrotllm.training -   Train: loss=4.0356, ppl=56.57
+2026-04-09 20:21:01 - INFO - parrotllm.training -   Val:   loss=3.8855, ppl=48.69
+2026-04-09 20:21:01 - INFO - parrotllm.training -   ** New best validation loss! **
+2026-04-09 20:21:02 - INFO - parrotllm.training -   Saved best checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_3p8855_epoch_0000_step_0001400.pt
+2026-04-09 20:21:03 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:21:06 - INFO - parrotllm.training - step   1410 | epoch 0 | loss 3.9781 | lr 8.67e-04 | grad 0.2901
+2026-04-09 20:21:09 - INFO - parrotllm.training - step   1420 | epoch 0 | loss 3.9011 | lr 8.67e-04 | grad 0.3906
+2026-04-09 20:21:12 - INFO - parrotllm.training - step   1430 | epoch 0 | loss 3.9317 | lr 8.67e-04 | grad 0.2801
+2026-04-09 20:21:15 - INFO - parrotllm.training - step   1440 | epoch 0 | loss 4.0363 | lr 8.67e-04 | grad 0.2995
+2026-04-09 20:21:18 - INFO - parrotllm.training - step   1450 | epoch 0 | loss 4.0070 | lr 8.67e-04 | grad 0.2787
+2026-04-09 20:21:21 - INFO - parrotllm.training - step   1460 | epoch 0 | loss 3.9244 | lr 8.67e-04 | grad 0.2573
+2026-04-09 20:21:24 - INFO - parrotllm.training - step   1470 | epoch 0 | loss 3.8709 | lr 8.67e-04 | grad 0.3359
+2026-04-09 20:21:27 - INFO - parrotllm.training - step   1480 | epoch 0 | loss 3.9860 | lr 8.67e-04 | grad 0.2945
+2026-04-09 20:21:30 - INFO - parrotllm.training - step   1490 | epoch 0 | loss 4.0545 | lr 8.67e-04 | grad 0.2949
+2026-04-09 20:21:33 - INFO - parrotllm.training - step   1500 | epoch 0 | loss 3.8959 | lr 8.67e-04 | grad 0.3018
+2026-04-09 20:21:33 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:21:33 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:21:39 - INFO - parrotllm.training -   Train: loss=3.8959, ppl=49.20
+2026-04-09 20:21:39 - INFO - parrotllm.training -   Val:   loss=3.8550, ppl=47.23
+2026-04-09 20:21:39 - INFO - parrotllm.training -   ** New best validation loss! **
+2026-04-09 20:21:40 - INFO - parrotllm.training -   Saved best checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_3p8550_epoch_0000_step_0001500.pt
+2026-04-09 20:21:41 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:21:42 - INFO - parrotllm.training - Saved checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/last_epoch_0000_step_0001500.pt
+2026-04-09 20:21:46 - INFO - parrotllm.training - step   1510 | epoch 0 | loss 3.9259 | lr 8.67e-04 | grad 0.2446
+2026-04-09 20:21:49 - INFO - parrotllm.training - step   1520 | epoch 0 | loss 3.9510 | lr 8.67e-04 | grad 0.2560
+2026-04-09 20:21:52 - INFO - parrotllm.training - step   1530 | epoch 0 | loss 3.9210 | lr 8.67e-04 | grad 0.2629
+2026-04-09 20:21:55 - INFO - parrotllm.training - step   1540 | epoch 0 | loss 3.9250 | lr 8.67e-04 | grad 0.2596
+2026-04-09 20:21:58 - INFO - parrotllm.training - step   1550 | epoch 0 | loss 4.0183 | lr 8.67e-04 | grad 0.2571
+2026-04-09 20:22:01 - INFO - parrotllm.training - step   1560 | epoch 0 | loss 3.9999 | lr 8.67e-04 | grad 0.2919
+2026-04-09 20:22:04 - INFO - parrotllm.training - step   1570 | epoch 0 | loss 3.9850 | lr 8.67e-04 | grad 0.2776
+2026-04-09 20:22:07 - INFO - parrotllm.training - step   1580 | epoch 0 | loss 3.9587 | lr 8.67e-04 | grad 0.2612
+2026-04-09 20:22:10 - INFO - parrotllm.training - step   1590 | epoch 0 | loss 3.9487 | lr 8.67e-04 | grad 0.2741
+2026-04-09 20:22:13 - INFO - parrotllm.training - step   1600 | epoch 0 | loss 3.9516 | lr 8.67e-04 | grad 0.2847
+2026-04-09 20:22:13 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:22:13 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:22:19 - INFO - parrotllm.training -   Train: loss=3.9516, ppl=52.02
+2026-04-09 20:22:19 - INFO - parrotllm.training -   Val:   loss=3.8321, ppl=46.16
+2026-04-09 20:22:19 - INFO - parrotllm.training -   ** New best validation loss! **
+2026-04-09 20:22:20 - INFO - parrotllm.training -   Saved best checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_3p8321_epoch_0000_step_0001600.pt
+2026-04-09 20:22:20 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:22:24 - INFO - parrotllm.training - step   1610 | epoch 0 | loss 3.8982 | lr 8.67e-04 | grad 0.2664
+2026-04-09 20:22:27 - INFO - parrotllm.training - step   1620 | epoch 0 | loss 3.9015 | lr 8.67e-04 | grad 0.3160
+2026-04-09 20:22:30 - INFO - parrotllm.training - step   1630 | epoch 0 | loss 3.9234 | lr 8.67e-04 | grad 0.2700
+2026-04-09 20:22:33 - INFO - parrotllm.training - step   1640 | epoch 0 | loss 3.8760 | lr 8.67e-04 | grad 0.2797
+2026-04-09 20:22:36 - INFO - parrotllm.training - step   1650 | epoch 0 | loss 3.9557 | lr 8.67e-04 | grad 0.3098
+2026-04-09 20:22:39 - INFO - parrotllm.training - step   1660 | epoch 0 | loss 3.9361 | lr 8.67e-04 | grad 0.2719
+2026-04-09 20:22:42 - INFO - parrotllm.training - step   1670 | epoch 0 | loss 3.8101 | lr 8.67e-04 | grad 0.2732
+2026-04-09 20:22:45 - INFO - parrotllm.training - step   1680 | epoch 0 | loss 3.9513 | lr 8.67e-04 | grad 0.3010
+2026-04-09 20:22:48 - INFO - parrotllm.training - step   1690 | epoch 0 | loss 3.8724 | lr 8.67e-04 | grad 0.2771
+2026-04-09 20:22:51 - INFO - parrotllm.training - step   1700 | epoch 0 | loss 3.8514 | lr 8.67e-04 | grad 0.2723
+2026-04-09 20:22:51 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:22:51 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:22:56 - INFO - parrotllm.training -   Train: loss=3.8514, ppl=47.06
+2026-04-09 20:22:56 - INFO - parrotllm.training -   Val:   loss=3.8090, ppl=45.11
+2026-04-09 20:22:56 - INFO - parrotllm.training -   ** New best validation loss! **
+2026-04-09 20:22:57 - INFO - parrotllm.training -   Saved best checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_3p8090_epoch_0000_step_0001700.pt
+2026-04-09 20:22:58 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:23:02 - INFO - parrotllm.training - step   1710 | epoch 0 | loss 3.9869 | lr 8.67e-04 | grad 0.2566
+2026-04-09 20:23:05 - INFO - parrotllm.training - step   1720 | epoch 0 | loss 3.8002 | lr 8.67e-04 | grad 0.2733
+2026-04-09 20:23:08 - INFO - parrotllm.training - step   1730 | epoch 0 | loss 3.8549 | lr 8.67e-04 | grad 0.3195
+2026-04-09 20:23:11 - INFO - parrotllm.training - step   1740 | epoch 0 | loss 3.9482 | lr 8.67e-04 | grad 0.2606
+2026-04-09 20:23:14 - INFO - parrotllm.training - step   1750 | epoch 0 | loss 4.0442 | lr 8.67e-04 | grad 0.2802
+2026-04-09 20:23:15 - INFO - parrotllm.training - Saved checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/last_epoch_0000_step_0001750.pt
+2026-04-09 20:23:19 - INFO - parrotllm.training - step   1760 | epoch 0 | loss 3.9149 | lr 8.67e-04 | grad 0.2758
+2026-04-09 20:23:22 - INFO - parrotllm.training - step   1770 | epoch 0 | loss 3.9429 | lr 8.67e-04 | grad 0.2912
+2026-04-09 20:23:25 - INFO - parrotllm.training - step   1780 | epoch 0 | loss 3.9968 | lr 8.67e-04 | grad 0.2705
+2026-04-09 20:23:28 - INFO - parrotllm.training - step   1790 | epoch 0 | loss 3.8810 | lr 8.67e-04 | grad 0.2860
+2026-04-09 20:23:31 - INFO - parrotllm.training - step   1800 | epoch 0 | loss 3.8075 | lr 8.67e-04 | grad 0.2595
+2026-04-09 20:23:31 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:23:31 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:23:36 - INFO - parrotllm.training -   Train: loss=3.8075, ppl=45.04
+2026-04-09 20:23:36 - INFO - parrotllm.training -   Val:   loss=3.7900, ppl=44.26
+2026-04-09 20:23:36 - INFO - parrotllm.training -   ** New best validation loss! **
+2026-04-09 20:23:37 - INFO - parrotllm.training -   Saved best checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_3p7900_epoch_0000_step_0001800.pt
+2026-04-09 20:23:38 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:23:41 - INFO - parrotllm.training - step   1810 | epoch 0 | loss 4.0064 | lr 8.67e-04 | grad 0.3454
+2026-04-09 20:23:44 - INFO - parrotllm.training - step   1820 | epoch 0 | loss 3.9616 | lr 8.67e-04 | grad 0.2512
+2026-04-09 20:23:47 - INFO - parrotllm.training - step   1830 | epoch 0 | loss 3.9014 | lr 8.67e-04 | grad 0.2610
+2026-04-09 20:23:50 - INFO - parrotllm.training - step   1840 | epoch 0 | loss 3.8889 | lr 8.67e-04 | grad 0.2813
+2026-04-09 20:23:53 - INFO - parrotllm.training - step   1850 | epoch 0 | loss 3.8053 | lr 8.67e-04 | grad 0.2970
+2026-04-09 20:23:56 - INFO - parrotllm.training - step   1860 | epoch 0 | loss 3.8196 | lr 8.67e-04 | grad 0.2705
+2026-04-09 20:23:59 - INFO - parrotllm.training - step   1870 | epoch 0 | loss 3.8891 | lr 8.67e-04 | grad 0.2632
+2026-04-09 20:24:02 - INFO - parrotllm.training - step   1880 | epoch 0 | loss 3.8773 | lr 8.67e-04 | grad 0.3158
+2026-04-09 20:24:05 - INFO - parrotllm.training - step   1890 | epoch 0 | loss 3.8776 | lr 8.67e-04 | grad 0.2576
+2026-04-09 20:24:08 - INFO - parrotllm.training - step   1900 | epoch 0 | loss 3.8286 | lr 8.67e-04 | grad 0.2591
+2026-04-09 20:24:08 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:24:08 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:24:14 - INFO - parrotllm.training -   Train: loss=3.8286, ppl=46.00
+2026-04-09 20:24:14 - INFO - parrotllm.training -   Val:   loss=3.7704, ppl=43.40
+2026-04-09 20:24:14 - INFO - parrotllm.training -   ** New best validation loss! **
+2026-04-09 20:24:15 - INFO - parrotllm.training -   Saved best checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_3p7704_epoch_0000_step_0001900.pt
+2026-04-09 20:24:16 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:24:19 - INFO - parrotllm.training - step   1910 | epoch 0 | loss 3.8134 | lr 8.67e-04 | grad 0.2377
+2026-04-09 20:24:22 - INFO - parrotllm.training - step   1920 | epoch 0 | loss 3.8963 | lr 8.67e-04 | grad 0.2935
+2026-04-09 20:24:25 - INFO - parrotllm.training - step   1930 | epoch 0 | loss 3.8411 | lr 8.67e-04 | grad 0.2270
+2026-04-09 20:24:28 - INFO - parrotllm.training - step   1940 | epoch 0 | loss 3.9979 | lr 8.67e-04 | grad 0.2587
+2026-04-09 20:24:31 - INFO - parrotllm.training - step   1950 | epoch 0 | loss 3.9262 | lr 8.67e-04 | grad 0.3754
+2026-04-09 20:24:34 - INFO - parrotllm.training - step   1960 | epoch 0 | loss 3.7887 | lr 8.67e-04 | grad 0.2832
+2026-04-09 20:24:37 - INFO - parrotllm.training - step   1970 | epoch 0 | loss 3.8624 | lr 8.67e-04 | grad 0.2625
+2026-04-09 20:24:40 - INFO - parrotllm.training - step   1980 | epoch 0 | loss 3.8513 | lr 8.67e-04 | grad 0.2333
+2026-04-09 20:24:43 - INFO - parrotllm.training - step   1990 | epoch 0 | loss 3.8407 | lr 8.67e-04 | grad 0.3420
+2026-04-09 20:24:46 - INFO - parrotllm.training - step   2000 | epoch 0 | loss 3.8602 | lr 8.67e-04 | grad 0.2587
+2026-04-09 20:24:46 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:24:46 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:24:52 - INFO - parrotllm.training -   Train: loss=3.8602, ppl=47.47
+2026-04-09 20:24:52 - INFO - parrotllm.training -   Val:   loss=3.7589, ppl=42.90
+2026-04-09 20:24:52 - INFO - parrotllm.training -   ** New best validation loss! **
+2026-04-09 20:24:53 - INFO - parrotllm.training -   Saved best checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_3p7589_epoch_0000_step_0002000.pt
+2026-04-09 20:24:53 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:24:54 - INFO - parrotllm.training - Saved checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/last_epoch_0000_step_0002000.pt
+2026-04-09 20:24:58 - INFO - parrotllm.training - step   2010 | epoch 0 | loss 3.8628 | lr 8.67e-04 | grad 0.2510
+2026-04-09 20:25:01 - INFO - parrotllm.training - step   2020 | epoch 0 | loss 3.8906 | lr 8.67e-04 | grad 0.2524
+2026-04-09 20:25:04 - INFO - parrotllm.training - step   2030 | epoch 0 | loss 3.8784 | lr 8.67e-04 | grad 0.2648
+2026-04-09 20:25:07 - INFO - parrotllm.training - step   2040 | epoch 0 | loss 3.7938 | lr 8.67e-04 | grad 0.2648
+2026-04-09 20:25:10 - INFO - parrotllm.training - step   2050 | epoch 0 | loss 3.8210 | lr 8.67e-04 | grad 0.2351
+2026-04-09 20:25:13 - INFO - parrotllm.training - step   2060 | epoch 0 | loss 3.9069 | lr 8.67e-04 | grad 0.2620
+2026-04-09 20:25:16 - INFO - parrotllm.training - step   2070 | epoch 0 | loss 3.8982 | lr 8.67e-04 | grad 0.2602
+2026-04-09 20:25:19 - INFO - parrotllm.training - step   2080 | epoch 0 | loss 3.8304 | lr 8.67e-04 | grad 0.2723
+2026-04-09 20:25:22 - INFO - parrotllm.training - step   2090 | epoch 0 | loss 3.8360 | lr 8.67e-04 | grad 0.3030
+2026-04-09 20:25:25 - INFO - parrotllm.training - step   2100 | epoch 0 | loss 3.7762 | lr 8.67e-04 | grad 0.2646
+2026-04-09 20:25:25 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:25:25 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:25:31 - INFO - parrotllm.training -   Train: loss=3.7762, ppl=43.65
+2026-04-09 20:25:31 - INFO - parrotllm.training -   Val:   loss=3.7428, ppl=42.21
+2026-04-09 20:25:31 - INFO - parrotllm.training -   ** New best validation loss! **
+2026-04-09 20:25:32 - INFO - parrotllm.training -   Saved best checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_3p7428_epoch_0000_step_0002100.pt
+2026-04-09 20:25:33 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:25:36 - INFO - parrotllm.training - step   2110 | epoch 0 | loss 3.8727 | lr 8.67e-04 | grad 0.2563
+2026-04-09 20:25:39 - INFO - parrotllm.training - step   2120 | epoch 0 | loss 3.8009 | lr 8.67e-04 | grad 0.2745
+2026-04-09 20:25:42 - INFO - parrotllm.training - step   2130 | epoch 0 | loss 3.8755 | lr 8.67e-04 | grad 0.2583
+2026-04-09 20:25:45 - INFO - parrotllm.training - step   2140 | epoch 0 | loss 3.8726 | lr 8.67e-04 | grad 0.2769
+2026-04-09 20:25:48 - INFO - parrotllm.training - step   2150 | epoch 0 | loss 3.7607 | lr 8.67e-04 | grad 0.2751
+2026-04-09 20:25:51 - INFO - parrotllm.training - step   2160 | epoch 0 | loss 3.8705 | lr 8.67e-04 | grad 0.2589
+2026-04-09 20:25:54 - INFO - parrotllm.training - step   2170 | epoch 0 | loss 3.8469 | lr 8.67e-04 | grad 0.3009
+2026-04-09 20:25:57 - INFO - parrotllm.training - step   2180 | epoch 0 | loss 3.9359 | lr 8.67e-04 | grad 0.2659
+2026-04-09 20:26:00 - INFO - parrotllm.training - step   2190 | epoch 0 | loss 3.7663 | lr 8.67e-04 | grad 0.2661
+2026-04-09 20:26:03 - INFO - parrotllm.training - step   2200 | epoch 0 | loss 3.8165 | lr 8.67e-04 | grad 0.2538
+2026-04-09 20:26:03 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:26:03 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:26:09 - INFO - parrotllm.training -   Train: loss=3.8165, ppl=45.45
+2026-04-09 20:26:09 - INFO - parrotllm.training -   Val:   loss=3.7277, ppl=41.58
+2026-04-09 20:26:09 - INFO - parrotllm.training -   ** New best validation loss! **
+2026-04-09 20:26:09 - INFO - parrotllm.training -   Saved best checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_3p7277_epoch_0000_step_0002200.pt
+2026-04-09 20:26:10 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:26:14 - INFO - parrotllm.training - step   2210 | epoch 0 | loss 3.7241 | lr 8.67e-04 | grad 0.2825
+2026-04-09 20:26:17 - INFO - parrotllm.training - step   2220 | epoch 0 | loss 3.7919 | lr 8.67e-04 | grad 0.3039
+2026-04-09 20:26:20 - INFO - parrotllm.training - step   2230 | epoch 0 | loss 3.7169 | lr 8.67e-04 | grad 0.2479
+2026-04-09 20:26:23 - INFO - parrotllm.training - step   2240 | epoch 0 | loss 3.8619 | lr 8.67e-04 | grad 0.2932
+2026-04-09 20:26:26 - INFO - parrotllm.training - step   2250 | epoch 0 | loss 3.9083 | lr 8.67e-04 | grad 0.2602
+2026-04-09 20:26:27 - INFO - parrotllm.training - Saved checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/last_epoch_0000_step_0002250.pt
+2026-04-09 20:26:30 - INFO - parrotllm.training - step   2260 | epoch 0 | loss 3.8494 | lr 8.67e-04 | grad 0.2300
+2026-04-09 20:26:33 - INFO - parrotllm.training - step   2270 | epoch 0 | loss 3.8731 | lr 8.67e-04 | grad 0.2298
+2026-04-09 20:26:36 - INFO - parrotllm.training - step   2280 | epoch 0 | loss 3.8363 | lr 8.67e-04 | grad 0.2560
+2026-04-09 20:26:40 - INFO - parrotllm.training - step   2290 | epoch 0 | loss 3.8270 | lr 8.67e-04 | grad 0.2390
+2026-04-09 20:26:43 - INFO - parrotllm.training - step   2300 | epoch 0 | loss 3.7150 | lr 8.67e-04 | grad 0.2797
+2026-04-09 20:26:43 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:26:43 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:26:48 - INFO - parrotllm.training -   Train: loss=3.7150, ppl=41.06
+2026-04-09 20:26:48 - INFO - parrotllm.training -   Val:   loss=3.7163, ppl=41.11
+2026-04-09 20:26:48 - INFO - parrotllm.training -   ** New best validation loss! **
+2026-04-09 20:26:49 - INFO - parrotllm.training -   Saved best checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_3p7163_epoch_0000_step_0002300.pt
+2026-04-09 20:26:50 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:26:53 - INFO - parrotllm.training - step   2310 | epoch 0 | loss 3.8287 | lr 8.67e-04 | grad 0.2858
+2026-04-09 20:26:56 - INFO - parrotllm.training - step   2320 | epoch 0 | loss 3.8355 | lr 8.67e-04 | grad 0.2722
+2026-04-09 20:26:59 - INFO - parrotllm.training - step   2330 | epoch 0 | loss 3.8020 | lr 8.67e-04 | grad 0.2677
+2026-04-09 20:27:02 - INFO - parrotllm.training - step   2340 | epoch 0 | loss 3.8683 | lr 8.67e-04 | grad 0.2515
+2026-04-09 20:27:05 - INFO - parrotllm.training - step   2350 | epoch 0 | loss 3.8220 | lr 8.67e-04 | grad 0.2447
+2026-04-09 20:27:08 - INFO - parrotllm.training - step   2360 | epoch 0 | loss 3.7856 | lr 8.67e-04 | grad 0.2873
+2026-04-09 20:27:11 - INFO - parrotllm.training - step   2370 | epoch 0 | loss 3.7191 | lr 8.67e-04 | grad 0.2470
+2026-04-09 20:27:14 - INFO - parrotllm.training - step   2380 | epoch 0 | loss 3.7734 | lr 8.67e-04 | grad 0.3006
+2026-04-09 20:27:17 - INFO - parrotllm.training - step   2390 | epoch 0 | loss 3.8840 | lr 8.67e-04 | grad 0.2554
+2026-04-09 20:27:20 - INFO - parrotllm.training - step   2400 | epoch 0 | loss 3.7263 | lr 8.67e-04 | grad 0.2551
+2026-04-09 20:27:20 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:27:20 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:27:26 - INFO - parrotllm.training -   Train: loss=3.7263, ppl=41.53
+2026-04-09 20:27:26 - INFO - parrotllm.training -   Val:   loss=3.7040, ppl=40.61
+2026-04-09 20:27:26 - INFO - parrotllm.training -   ** New best validation loss! **
+2026-04-09 20:27:26 - INFO - parrotllm.training -   Saved best checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_3p7040_epoch_0000_step_0002400.pt
+2026-04-09 20:27:27 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:27:30 - INFO - parrotllm.training - step   2410 | epoch 0 | loss 3.6973 | lr 8.67e-04 | grad 0.2557
+2026-04-09 20:27:33 - INFO - parrotllm.training - step   2420 | epoch 0 | loss 3.7770 | lr 8.67e-04 | grad 0.2545
+2026-04-09 20:27:36 - INFO - parrotllm.training - step   2430 | epoch 0 | loss 3.7907 | lr 8.67e-04 | grad 0.2449
+2026-04-09 20:27:39 - INFO - parrotllm.training - step   2440 | epoch 0 | loss 3.8275 | lr 8.67e-04 | grad 0.2418
+2026-04-09 20:27:42 - INFO - parrotllm.training - step   2450 | epoch 0 | loss 3.8019 | lr 8.67e-04 | grad 0.2930
+2026-04-09 20:27:45 - INFO - parrotllm.training - step   2460 | epoch 0 | loss 3.7230 | lr 8.67e-04 | grad 0.2511
+2026-04-09 20:27:48 - INFO - parrotllm.training - step   2470 | epoch 0 | loss 3.7995 | lr 8.67e-04 | grad 0.2801
+2026-04-09 20:27:51 - INFO - parrotllm.training - step   2480 | epoch 0 | loss 3.7421 | lr 8.67e-04 | grad 0.2441
+2026-04-09 20:27:54 - INFO - parrotllm.training - step   2490 | epoch 0 | loss 3.7179 | lr 8.67e-04 | grad 0.2851
+2026-04-09 20:27:58 - INFO - parrotllm.training - step   2500 | epoch 0 | loss 3.7742 | lr 8.67e-04 | grad 0.2773
+2026-04-09 20:27:58 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:27:58 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:28:03 - INFO - parrotllm.training -   Train: loss=3.7742, ppl=43.56
+2026-04-09 20:28:03 - INFO - parrotllm.training -   Val:   loss=3.6927, ppl=40.15
+2026-04-09 20:28:03 - INFO - parrotllm.training -   ** New best validation loss! **
+2026-04-09 20:28:04 - INFO - parrotllm.training -   Saved best checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_3p6927_epoch_0000_step_0002500.pt
+2026-04-09 20:28:05 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:28:06 - INFO - parrotllm.training - Saved checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/last_epoch_0000_step_0002500.pt
+2026-04-09 20:28:10 - INFO - parrotllm.training - step   2510 | epoch 0 | loss 3.6786 | lr 8.67e-04 | grad 0.3165
+2026-04-09 20:28:13 - INFO - parrotllm.training - step   2520 | epoch 0 | loss 3.8263 | lr 8.61e-04 | grad 0.2570
+2026-04-09 20:28:16 - INFO - parrotllm.training - step   2530 | epoch 0 | loss 3.7636 | lr 8.47e-04 | grad 0.2576
+2026-04-09 20:28:19 - INFO - parrotllm.training - step   2540 | epoch 0 | loss 3.7777 | lr 8.32e-04 | grad 0.2680
+2026-04-09 20:28:22 - INFO - parrotllm.training - step   2550 | epoch 0 | loss 3.8920 | lr 8.18e-04 | grad 0.2719
+2026-04-09 20:28:25 - INFO - parrotllm.training - step   2560 | epoch 0 | loss 3.7446 | lr 8.03e-04 | grad 0.2652
+2026-04-09 20:28:28 - INFO - parrotllm.training - step   2570 | epoch 0 | loss 3.8503 | lr 7.89e-04 | grad 0.2858
+2026-04-09 20:28:31 - INFO - parrotllm.training - step   2580 | epoch 0 | loss 3.8246 | lr 7.74e-04 | grad 0.2547
+2026-04-09 20:28:34 - INFO - parrotllm.training - step   2590 | epoch 0 | loss 3.5789 | lr 7.59e-04 | grad 0.2215
+2026-04-09 20:28:38 - INFO - parrotllm.training - step   2600 | epoch 0 | loss 3.7214 | lr 7.45e-04 | grad 0.2059
+2026-04-09 20:28:38 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:28:38 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:28:43 - INFO - parrotllm.training -   Train: loss=3.7214, ppl=41.32
+2026-04-09 20:28:43 - INFO - parrotllm.training -   Val:   loss=3.6733, ppl=39.38
+2026-04-09 20:28:43 - INFO - parrotllm.training -   ** New best validation loss! **
+2026-04-09 20:28:44 - INFO - parrotllm.training -   Saved best checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_3p6733_epoch_0000_step_0002600.pt
+2026-04-09 20:28:45 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:28:48 - INFO - parrotllm.training - step   2610 | epoch 0 | loss 3.7615 | lr 7.30e-04 | grad 0.2490
+2026-04-09 20:28:51 - INFO - parrotllm.training - step   2620 | epoch 0 | loss 3.6744 | lr 7.16e-04 | grad 0.2235
+2026-04-09 20:28:54 - INFO - parrotllm.training - step   2630 | epoch 0 | loss 3.7911 | lr 7.01e-04 | grad 0.2506
+2026-04-09 20:28:57 - INFO - parrotllm.training - step   2640 | epoch 0 | loss 3.7503 | lr 6.87e-04 | grad 0.2130
+2026-04-09 20:29:00 - INFO - parrotllm.training - step   2650 | epoch 0 | loss 3.8209 | lr 6.72e-04 | grad 0.2501
+2026-04-09 20:29:03 - INFO - parrotllm.training - step   2660 | epoch 0 | loss 3.7144 | lr 6.57e-04 | grad 0.2282
+2026-04-09 20:29:06 - INFO - parrotllm.training - step   2670 | epoch 0 | loss 3.7950 | lr 6.43e-04 | grad 0.2142
+2026-04-09 20:29:09 - INFO - parrotllm.training - step   2680 | epoch 0 | loss 3.7078 | lr 6.28e-04 | grad 0.2221
+2026-04-09 20:29:12 - INFO - parrotllm.training - step   2690 | epoch 0 | loss 3.7546 | lr 6.14e-04 | grad 0.2339
+2026-04-09 20:29:15 - INFO - parrotllm.training - step   2700 | epoch 0 | loss 3.7246 | lr 5.99e-04 | grad 0.2207
+2026-04-09 20:29:15 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:29:15 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:29:21 - INFO - parrotllm.training -   Train: loss=3.7246, ppl=41.46
+2026-04-09 20:29:21 - INFO - parrotllm.training -   Val:   loss=3.6522, ppl=38.56
+2026-04-09 20:29:21 - INFO - parrotllm.training -   ** New best validation loss! **
+2026-04-09 20:29:22 - INFO - parrotllm.training -   Saved best checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_3p6522_epoch_0000_step_0002700.pt
+2026-04-09 20:29:22 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:29:25 - INFO - parrotllm.training - step   2710 | epoch 0 | loss 3.7036 | lr 5.85e-04 | grad 0.2198
+2026-04-09 20:29:28 - INFO - parrotllm.training - step   2720 | epoch 0 | loss 3.8031 | lr 5.70e-04 | grad 0.2188
+2026-04-09 20:29:31 - INFO - parrotllm.training - step   2730 | epoch 0 | loss 3.7135 | lr 5.56e-04 | grad 0.2157
+2026-04-09 20:29:35 - INFO - parrotllm.training - step   2740 | epoch 0 | loss 3.6840 | lr 5.41e-04 | grad 0.2102
+2026-04-09 20:29:38 - INFO - parrotllm.training - step   2750 | epoch 0 | loss 3.5823 | lr 5.26e-04 | grad 0.2191
+2026-04-09 20:29:39 - INFO - parrotllm.training - Saved checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/last_epoch_0000_step_0002750.pt
+2026-04-09 20:29:42 - INFO - parrotllm.training - step   2760 | epoch 0 | loss 3.7461 | lr 5.12e-04 | grad 0.2181
+2026-04-09 20:29:45 - INFO - parrotllm.training - step   2770 | epoch 0 | loss 3.7561 | lr 4.97e-04 | grad 0.2149
+2026-04-09 20:29:49 - INFO - parrotllm.training - step   2780 | epoch 0 | loss 3.7118 | lr 4.83e-04 | grad 0.2103
+2026-04-09 20:29:52 - INFO - parrotllm.training - step   2790 | epoch 0 | loss 3.7092 | lr 4.68e-04 | grad 0.2018
+2026-04-09 20:29:55 - INFO - parrotllm.training - step   2800 | epoch 0 | loss 3.7002 | lr 4.54e-04 | grad 0.1974
+2026-04-09 20:29:55 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:29:55 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:30:00 - INFO - parrotllm.training -   Train: loss=3.7002, ppl=40.46
+2026-04-09 20:30:00 - INFO - parrotllm.training -   Val:   loss=3.6293, ppl=37.69
+2026-04-09 20:30:00 - INFO - parrotllm.training -   ** New best validation loss! **
+2026-04-09 20:30:01 - INFO - parrotllm.training -   Saved best checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_3p6293_epoch_0000_step_0002800.pt
+2026-04-09 20:30:02 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:30:05 - INFO - parrotllm.training - step   2810 | epoch 0 | loss 3.6960 | lr 4.39e-04 | grad 0.2256
+2026-04-09 20:30:08 - INFO - parrotllm.training - step   2820 | epoch 0 | loss 3.6442 | lr 4.25e-04 | grad 0.2039
+2026-04-09 20:30:11 - INFO - parrotllm.training - step   2830 | epoch 0 | loss 3.6760 | lr 4.10e-04 | grad 0.1911
+2026-04-09 20:30:14 - INFO - parrotllm.training - step   2840 | epoch 0 | loss 3.7377 | lr 3.95e-04 | grad 0.2111
+2026-04-09 20:30:17 - INFO - parrotllm.training - step   2850 | epoch 0 | loss 3.8213 | lr 3.81e-04 | grad 0.1981
+2026-04-09 20:30:20 - INFO - parrotllm.training - step   2860 | epoch 0 | loss 3.6535 | lr 3.66e-04 | grad 0.2097
+2026-04-09 20:30:23 - INFO - parrotllm.training - step   2870 | epoch 0 | loss 3.6860 | lr 3.52e-04 | grad 0.1989
+2026-04-09 20:30:26 - INFO - parrotllm.training - step   2880 | epoch 0 | loss 3.7217 | lr 3.37e-04 | grad 0.2058
+2026-04-09 20:30:29 - INFO - parrotllm.training - step   2890 | epoch 0 | loss 3.6958 | lr 3.23e-04 | grad 0.1999
+2026-04-09 20:30:32 - INFO - parrotllm.training - step   2900 | epoch 0 | loss 3.6893 | lr 3.08e-04 | grad 0.1950
+2026-04-09 20:30:32 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:30:32 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:30:38 - INFO - parrotllm.training -   Train: loss=3.6893, ppl=40.02
+2026-04-09 20:30:38 - INFO - parrotllm.training -   Val:   loss=3.6109, ppl=37.00
+2026-04-09 20:30:38 - INFO - parrotllm.training -   ** New best validation loss! **
+2026-04-09 20:30:39 - INFO - parrotllm.training -   Saved best checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_3p6109_epoch_0000_step_0002900.pt
+2026-04-09 20:30:39 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:30:42 - INFO - parrotllm.training - step   2910 | epoch 0 | loss 3.7027 | lr 2.93e-04 | grad 0.1933
+2026-04-09 20:30:46 - INFO - parrotllm.training - step   2920 | epoch 0 | loss 3.6981 | lr 2.79e-04 | grad 0.1996
+2026-04-09 20:30:49 - INFO - parrotllm.training - step   2930 | epoch 0 | loss 3.7253 | lr 2.64e-04 | grad 0.1880
+2026-04-09 20:30:52 - INFO - parrotllm.training - step   2940 | epoch 0 | loss 3.6731 | lr 2.50e-04 | grad 0.2053
+2026-04-09 20:30:55 - INFO - parrotllm.training - step   2950 | epoch 0 | loss 3.8306 | lr 2.35e-04 | grad 0.1794
+2026-04-09 20:30:58 - INFO - parrotllm.training - step   2960 | epoch 0 | loss 3.7582 | lr 2.21e-04 | grad 0.1914
+2026-04-09 20:31:01 - INFO - parrotllm.training - step   2970 | epoch 0 | loss 3.7368 | lr 2.06e-04 | grad 0.1796
+2026-04-09 20:31:04 - INFO - parrotllm.training - step   2980 | epoch 0 | loss 3.6227 | lr 1.92e-04 | grad 0.1845
+2026-04-09 20:31:07 - INFO - parrotllm.training - step   2990 | epoch 0 | loss 3.7574 | lr 1.77e-04 | grad 0.1824
+2026-04-09 20:31:10 - INFO - parrotllm.training - step   3000 | epoch 0 | loss 3.7157 | lr 1.62e-04 | grad 0.1855
+2026-04-09 20:31:10 - INFO - parrotllm.training - Starting evaluation...
+2026-04-09 20:31:10 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:31:15 - INFO - parrotllm.training -   Train: loss=3.7157, ppl=41.09
+2026-04-09 20:31:15 - INFO - parrotllm.training -   Val:   loss=3.5917, ppl=36.29
+2026-04-09 20:31:15 - INFO - parrotllm.training -   ** New best validation loss! **
+2026-04-09 20:31:16 - INFO - parrotllm.training -   Saved best checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/best_loss_3p5917_epoch_0000_step_0003000.pt
+2026-04-09 20:31:17 - INFO - parrotllm.training - ------------------------------------------------------------
+2026-04-09 20:31:18 - INFO - parrotllm.training - Saved checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/last_epoch_0000_step_0003000.pt
+2026-04-09 20:31:22 - INFO - parrotllm.training - step   3010 | epoch 0 | loss 3.6207 | lr 1.48e-04 | grad 0.1843
+2026-04-09 20:31:25 - INFO - parrotllm.training - step   3020 | epoch 0 | loss 3.7382 | lr 1.33e-04 | grad 0.1680
+2026-04-09 20:31:28 - INFO - parrotllm.training - step   3030 | epoch 0 | loss 3.6651 | lr 1.19e-04 | grad 0.1678
+2026-04-09 20:31:31 - INFO - parrotllm.training - step   3040 | epoch 0 | loss 3.7017 | lr 1.04e-04 | grad 0.1664
+2026-04-09 20:31:34 - INFO - parrotllm.training - step   3050 | epoch 0 | loss 3.6611 | lr 8.96e-05 | grad 0.1665
+2026-04-09 20:31:36 - INFO - parrotllm.training - Saved final checkpoint: /home/user/code/ParrotLLM/runs/big_run/exp_a/run_20260409_201114/checkpoints/last_epoch_0000_step_0003052.pt
+2026-04-09 20:31:36 - INFO - parrotllm.training - 
+Loss Curve:
+  Loss
+  ^
+   10.94 |*                                                 
+         |*                                                 
+         |*                                                 
+         |**                                                
+         | **                                               
+         |  ****                                            
+         |    ********                                      
+    3.51 |         *****************************************
+         +--------------------------------------------------> step
+         1                                             3052
+2026-04-09 20:31:36 - INFO - parrotllm.training - 
+============================================================
+TRAINING COMPLETE
+============================================================
+  Epochs: 1
+  Total steps: 3052
+  Total time: 0.34 hours
+  Best validation loss: 3.5917
+  Run directory: runs/big_run/exp_a/run_20260409_201114
+============================================================
+2026-04-09 20:31:36 - INFO - parrotllm.training - Uploading completed run directory to Hugging Face: repo=ParrotLabs/Preprocessed

runs/big_run/exp_a/run_20260409_201114/training_plots.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4fc36e72b0a56decf32186c440219f39bd210e2bac72d5b1f3c323bc5d916c96
+size 112397