Hugging Face's logo

Datasets:
ParrotLabs
/
Preprocessed

Dataset card Files
xet
 Community
thaferbe commited on
Commit
a12ee50
·
verified ·
1 Parent(s): ee91277

Upload training run run_20260408_121550

Browse files

ParrotLLM end-of-training sync for local run runs/smoketest/exp_c/run_20260408_121550.

Files changed (9) hide show
  1. runs/smoketest/exp_c/run_20260408_121550/checkpoints/best_loss_7p7412_epoch_0004_step_0000012.pt +3 -0
  2. runs/smoketest/exp_c/run_20260408_121550/checkpoints/best_loss_7p9581_epoch_0003_step_0000009.pt +3 -0
  3. runs/smoketest/exp_c/run_20260408_121550/checkpoints/best_loss_8p7229_epoch_0002_step_0000006.pt +3 -0
  4. runs/smoketest/exp_c/run_20260408_121550/checkpoints/best_loss_9p9140_epoch_0001_step_0000003.pt +3 -0
  5. runs/smoketest/exp_c/run_20260408_121550/checkpoints/last_epoch_0004_step_0000012.pt +3 -0
  6. runs/smoketest/exp_c/run_20260408_121550/config.json +68 -0
  7. runs/smoketest/exp_c/run_20260408_121550/metrics.jsonl +26 -0
  8. runs/smoketest/exp_c/run_20260408_121550/train.log +227 -0
  9. runs/smoketest/exp_c/run_20260408_121550/training_plots.pdf +0 -0
runs/smoketest/exp_c/run_20260408_121550/checkpoints/best_loss_7p7412_epoch_0004_step_0000012.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0a7a63529a9ff9e2291d0b26770ce488cd33dec4815eae468d8725b478f80933
3
+ size 479869116
runs/smoketest/exp_c/run_20260408_121550/checkpoints/best_loss_7p9581_epoch_0003_step_0000009.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:afbf834eb6e2d5a54d18c898aec662a353803d12f6b37cdac777e1d91602cd52
3
+ size 479869116
runs/smoketest/exp_c/run_20260408_121550/checkpoints/best_loss_8p7229_epoch_0002_step_0000006.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d254b04fee60ede895c36bc25fb919061f06a264128ee9aae793e7f4b54f7fca
3
+ size 479869116
runs/smoketest/exp_c/run_20260408_121550/checkpoints/best_loss_9p9140_epoch_0001_step_0000003.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2253ec0640058812c4320568694405db68458308945d1ae63b0ac87cb677d929
3
+ size 479869116
runs/smoketest/exp_c/run_20260408_121550/checkpoints/last_epoch_0004_step_0000012.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dc2e68478836aabcb5c4611ef3b2b0a5b3763071f9f6d061cf6c57160f8b7184
3
+ size 479860104
runs/smoketest/exp_c/run_20260408_121550/config.json ADDED
@@ -0,0 +1,68 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "preprocess": null,
3
+ "model": {
4
+ "vocab_size": 50258,
5
+ "pad_token_id": 50257,
6
+ "bos_token_id": 50256,
7
+ "eos_token_id": 50256,
8
+ "d_model": 384,
9
+ "n_layers": 14,
10
+ "n_heads": 6,
11
+ "d_ff": 768,
12
+ "context_length": 1024,
13
+ "bias": false,
14
+ "dropout": 0.005,
15
+ "rope_theta": 10000.0,
16
+ "gradient_checkpointing": false
17
+ },
18
+ "training": {
19
+ "device": "auto",
20
+ "train_bin": "data/SmokeTest/train.bin",
21
+ "val_bin": "data/SmokeTest/val.bin",
22
+ "num_workers": 4,
23
+ "pin_memory": true,
24
+ "batch_size": 32,
25
+ "gradient_accumulation_steps": 1,
26
+ "learning_rate": 0.001305,
27
+ "min_lr": 0.0001305,
28
+ "weight_decay": 0.063,
29
+ "beta1": 0.9,
30
+ "beta2": 0.9978,
31
+ "grad_clip": 1.786,
32
+ "warmup_steps": 4,
33
+ "max_steps": 12,
34
+ "lr_schedule": "wsd",
35
+ "lr_decay_ratio": 0.179,
36
+ "z_loss_coeff": 0.000275,
37
+ "save_every": 100,
38
+ "eval_every": 100,
39
+ "checkpoint_dir": "checkpoints",
40
+ "keep_last_checkpoints": 10,
41
+ "keep_best_checkpoints": 10,
42
+ "early_stopping_patience": 0,
43
+ "early_stopping_min_delta": 0.001,
44
+ "runs_dir": "runs/smoketest/exp_c",
45
+ "log_every": 1,
46
+ "hf_upload": {
47
+ "repo_id": "ParrotLabs/Preprocessed",
48
+ "repo_type": "dataset",
49
+ "path_in_repo": "",
50
+ "private": null
51
+ },
52
+ "compile": true
53
+ },
54
+ "tune": null,
55
+ "logging": {
56
+ "console_level": "INFO",
57
+ "file_level": "DEBUG",
58
+ "components": {
59
+ "data_preprocessing": "DEBUG",
60
+ "model_initialization": "INFO",
61
+ "training": "INFO"
62
+ },
63
+ "profiler": null
64
+ },
65
+ "eval": null,
66
+ "inference": null,
67
+ "chat": null
68
+ }
runs/smoketest/exp_c/run_20260408_121550/metrics.jsonl ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"stage": "pretraining", "type": "config", "timestamp": "2026-04-08T12:15:50.802435+00:00", "device": "auto", "train_bin": "data/SmokeTest/train.bin", "val_bin": "data/SmokeTest/val.bin", "num_workers": 4, "pin_memory": true, "batch_size": 32, "gradient_accumulation_steps": 1, "learning_rate": 0.001305, "min_lr": 0.0001305, "weight_decay": 0.063, "beta1": 0.9, "beta2": 0.9978, "grad_clip": 1.786, "warmup_steps": 4, "max_steps": 12, "lr_schedule": "wsd", "lr_decay_ratio": 0.179, "z_loss_coeff": 0.000275, "save_every": 100, "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/smoketest/exp_c", "log_every": 1, "hf_upload": {"repo_id": "ParrotLabs/Preprocessed", "repo_type": "dataset", "path_in_repo": "", "private": null}, "compile": true}
2
+ {"stage": "pretraining", "type": "checkpoint_policy", "timestamp": "2026-04-08T12:15:50.802528+00:00", "checkpoint_dir": "/home/user/code/ParrotLLM/runs/smoketest/exp_c/run_20260408_121550/checkpoints", "keep_last": 10, "keep_best": 10}
3
+ {"stage": "pretraining", "type": "model_architecture", "timestamp": "2026-04-08T12:15:52.504195+00:00", "vocab_size": 50258, "context_length": 1024, "n_layers": 14, "n_heads": 6, "d_model": 384, "d_ff": 768, "dropout": 0.005, "bias": false, "total_params": 39966592, "total_params_non_embedding": 39966592, "trainable_params": 39966592, "non_trainable_params": 0, "params_size_mb": 159.87}
4
+ {"stage": "pretraining", "type": "initial_validation", "timestamp": "2026-04-08T12:15:56.333692+00:00", "val_loss": 10.906056880950928, "val_ppl": 54505.499343765834}
5
+ {"stage": "pretraining", "type": "step", "timestamp": "2026-04-08T12:16:21.011074+00:00", "epoch": 0, "step": 1, "train_loss": 10.925907135009766, "perplexity": 55598.25723258657, "lr": 0.00032625, "grad_norm": 6.790669, "tokens_per_sec": 1328}
6
+ {"stage": "pretraining", "type": "step", "timestamp": "2026-04-08T12:16:21.358661+00:00", "epoch": 0, "step": 2, "train_loss": 10.627500534057617, "perplexity": 41253.883552460495, "lr": 0.0006525, "grad_norm": 17.098476, "tokens_per_sec": 94240}
7
+ {"stage": "pretraining", "type": "step", "timestamp": "2026-04-08T12:16:21.661981+00:00", "epoch": 0, "step": 3, "train_loss": 10.267926216125488, "perplexity": 28794.112441917918, "lr": 0.00097875, "grad_norm": 1.812592, "tokens_per_sec": 108026}
8
+ {"stage": "pretraining", "type": "eval", "timestamp": "2026-04-08T12:16:24.210773+00:00", "step": 3, "epoch": 1, "val_loss": 9.914009094238281, "val_ppl": 20211.541755223578, "eval_train_loss": 10.267926216125488, "eval_train_ppl": 28794.112441917918}
9
+ {"stage": "pretraining", "type": "best_checkpoint", "timestamp": "2026-04-08T12:16:25.115092+00:00", "step": 3, "epoch": 1, "path": "/home/user/code/ParrotLLM/runs/smoketest/exp_c/run_20260408_121550/checkpoints/best_loss_9p9140_epoch_0001_step_0000003.pt", "val_loss": 9.914009094238281}
10
+ {"stage": "pretraining", "type": "step", "timestamp": "2026-04-08T12:16:26.661349+00:00", "epoch": 1, "step": 4, "train_loss": 9.929620742797852, "perplexity": 20529.55312393707, "lr": 0.001305, "grad_norm": 1.660996, "tokens_per_sec": 6554}
11
+ {"stage": "pretraining", "type": "step", "timestamp": "2026-04-08T12:16:26.966312+00:00", "epoch": 1, "step": 5, "train_loss": 9.469917297363281, "perplexity": 12963.815045251664, "lr": 0.001305, "grad_norm": 1.596051, "tokens_per_sec": 107454}
12
+ {"stage": "pretraining", "type": "step", "timestamp": "2026-04-08T12:16:27.270674+00:00", "epoch": 1, "step": 6, "train_loss": 9.081244468688965, "perplexity": 8788.896730949346, "lr": 0.001305, "grad_norm": 1.58809, "tokens_per_sec": 107658}
13
+ {"stage": "pretraining", "type": "eval", "timestamp": "2026-04-08T12:16:29.843613+00:00", "step": 6, "epoch": 2, "val_loss": 8.722934246063232, "val_ppl": 6142.175326275997, "eval_train_loss": 9.081244468688965, "eval_train_ppl": 8788.896730949346}
14
+ {"stage": "pretraining", "type": "best_checkpoint", "timestamp": "2026-04-08T12:16:30.754211+00:00", "step": 6, "epoch": 2, "path": "/home/user/code/ParrotLLM/runs/smoketest/exp_c/run_20260408_121550/checkpoints/best_loss_8p7229_epoch_0002_step_0000006.pt", "val_loss": 8.722934246063232}
15
+ {"stage": "pretraining", "type": "step", "timestamp": "2026-04-08T12:16:31.645387+00:00", "epoch": 2, "step": 7, "train_loss": 8.700838088989258, "perplexity": 6007.945300685089, "lr": 0.001305, "grad_norm": 1.531264, "tokens_per_sec": 7490}
16
+ {"stage": "pretraining", "type": "step", "timestamp": "2026-04-08T12:16:31.948671+00:00", "epoch": 2, "step": 8, "train_loss": 8.399039268493652, "perplexity": 4442.796362239903, "lr": 0.001305, "grad_norm": 1.477197, "tokens_per_sec": 108037}
17
+ {"stage": "pretraining", "type": "step", "timestamp": "2026-04-08T12:16:32.251874+00:00", "epoch": 2, "step": 9, "train_loss": 8.159897804260254, "perplexity": 3497.829122262445, "lr": 0.001305, "grad_norm": 1.317899, "tokens_per_sec": 108074}
18
+ {"stage": "pretraining", "type": "eval", "timestamp": "2026-04-08T12:16:34.809847+00:00", "step": 9, "epoch": 3, "val_loss": 7.9581074714660645, "val_ppl": 2858.657738560314, "eval_train_loss": 8.159897804260254, "eval_train_ppl": 3497.829122262445}
19
+ {"stage": "pretraining", "type": "best_checkpoint", "timestamp": "2026-04-08T12:16:35.835105+00:00", "step": 9, "epoch": 3, "path": "/home/user/code/ParrotLLM/runs/smoketest/exp_c/run_20260408_121550/checkpoints/best_loss_7p9581_epoch_0003_step_0000009.pt", "val_loss": 7.9581074714660645}
20
+ {"stage": "pretraining", "type": "step", "timestamp": "2026-04-08T12:16:36.810067+00:00", "epoch": 3, "step": 10, "train_loss": 7.925987720489502, "perplexity": 2768.2973095511697, "lr": 0.001305, "grad_norm": 1.134153, "tokens_per_sec": 7189}
21
+ {"stage": "pretraining", "type": "step", "timestamp": "2026-04-08T12:16:37.113550+00:00", "epoch": 3, "step": 11, "train_loss": 7.736440658569336, "perplexity": 2290.3058774545566, "lr": 0.00071775, "grad_norm": 0.919109, "tokens_per_sec": 107985}
22
+ {"stage": "pretraining", "type": "step", "timestamp": "2026-04-08T12:16:37.415788+00:00", "epoch": 3, "step": 12, "train_loss": 7.815825939178467, "perplexity": 2479.534051825406, "lr": 0.00013049999999999997, "grad_norm": 0.729313, "tokens_per_sec": 108400}
23
+ {"stage": "pretraining", "type": "eval", "timestamp": "2026-04-08T12:16:39.966284+00:00", "step": 12, "epoch": 4, "val_loss": 7.7412261962890625, "val_ppl": 2301.2924900780213, "eval_train_loss": 7.815825939178467, "eval_train_ppl": 2479.534051825406}
24
+ {"stage": "pretraining", "type": "best_checkpoint", "timestamp": "2026-04-08T12:16:42.108939+00:00", "step": 12, "epoch": 4, "path": "/home/user/code/ParrotLLM/runs/smoketest/exp_c/run_20260408_121550/checkpoints/best_loss_7p7412_epoch_0004_step_0000012.pt", "val_loss": 7.7412261962890625}
25
+ {"stage": "pretraining", "type": "checkpoint", "timestamp": "2026-04-08T12:16:45.152272+00:00", "step": 12, "epoch": 4, "path": "/home/user/code/ParrotLLM/runs/smoketest/exp_c/run_20260408_121550/checkpoints/last_epoch_0004_step_0000012.pt", "category": "last"}
26
+ {"stage": "pretraining", "type": "training_complete", "timestamp": "2026-04-08T12:16:45.152543+00:00", "epochs": 4, "total_steps": 12, "total_time_hours": 0.01, "best_val_loss": 7.7412261962890625, "best_checkpoint_path": "/home/user/code/ParrotLLM/runs/smoketest/exp_c/run_20260408_121550/checkpoints/best_loss_7p7412_epoch_0004_step_0000012.pt", "checkpoint_dir": "/home/user/code/ParrotLLM/runs/smoketest/exp_c/run_20260408_121550/checkpoints", "stopped_early": false, "stop_reason": null, "run_dir": "runs/smoketest/exp_c/run_20260408_121550"}
runs/smoketest/exp_c/run_20260408_121550/train.log ADDED
@@ -0,0 +1,227 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ 2026-04-08 12:15:50 - INFO - parrotllm - Logging initialised -> runs/smoketest/exp_c/run_20260408_121550/train.log
2
+ 2026-04-08 12:15:50 - INFO - parrotllm.training - device=cuda:0 | rank=0 | world_size=8 | distributed=yes
3
+ 2026-04-08 12:15:50 - INFO - parrotllm.training - checkpoint directory=/home/user/code/ParrotLLM/runs/smoketest/exp_c/run_20260408_121550/checkpoints | keep_last=10 | keep_best=10
4
+ 2026-04-08 12:15:51 - INFO - parrotllm.training - compiling model with torch.compile...
5
+ 2026-04-08 12:15:52 - INFO - parrotllm.training -
6
+ ============================================================
7
+ MODEL ARCHITECTURE SUMMARY
8
+ ============================================================
9
+
10
+ Configuration:
11
+ Vocab size: 50258
12
+ Block size (context): 1024
13
+ Layers: 14
14
+ Heads: 6
15
+ Embedding dim: 384
16
+ FFN hidden dim: 768
17
+ Dropout: 0.005
18
+ Bias: False
19
+
20
+ Parameters (unique, weight-tied layers counted once):
21
+ Total: 39,966,592 (39.97M)
22
+ Trainable: 39,966,592 (39.97M)
23
+ Non-trainable: 0
24
+ Non-embedding: 39,966,592 (39.97M)
25
+ Position embeddings: 0 (RoPE: 0 learned params)
26
+ Size (MB): 159.87
27
+
28
+ Layer-wise breakdown (torchinfo):
29
+ Note: torchinfo double-counts weight-tied layers (tok_emb/lm_head).
30
+ ============================================================================================================================================
31
+ Layer (type:depth-idx) Input Shape Output Shape Param # Trainable
32
+ ============================================================================================================================================
33
+ ParrotLLM [32, 1024] [32, 1024, 50258] -- True
34
+ ├─Embedding: 1-1 [32, 1024] [32, 1024, 384] 19,299,072 True
35
+ ├─Dropout: 1-2 [32, 1024, 384] [32, 1024, 384] -- --
36
+ ├─ModuleList: 1-3 -- -- -- True
37
+ │ └─TransformerBlock: 2-1 [32, 1024, 384] [32, 1024, 384] -- True
38
+ │ │ └─RMSNorm: 3-1 [32, 1024, 384] [32, 1024, 384] 384 True
39
+ │ │ └─MultiHeadAttention: 3-2 [32, 1024, 384] [32, 1024, 384] 589,952 True
40
+ │ │ └─RMSNorm: 3-3 [32, 1024, 384] [32, 1024, 384] 384 True
41
+ │ │ └─RMSNorm: 3-4 [32, 1024, 384] [32, 1024, 384] 384 True
42
+ │ │ └─SwiGLUMLP: 3-5 [32, 1024, 384] [32, 1024, 384] 884,736 True
43
+ │ │ └─RMSNorm: 3-6 [32, 1024, 384] [32, 1024, 384] 384 True
44
+ │ └─TransformerBlock: 2-2 [32, 1024, 384] [32, 1024, 384] -- True
45
+ │ │ └─RMSNorm: 3-7 [32, 1024, 384] [32, 1024, 384] 384 True
46
+ │ │ └─MultiHeadAttention: 3-8 [32, 1024, 384] [32, 1024, 384] 589,952 True
47
+ │ │ └─RMSNorm: 3-9 [32, 1024, 384] [32, 1024, 384] 384 True
48
+ │ │ └─RMSNorm: 3-10 [32, 1024, 384] [32, 1024, 384] 384 True
49
+ │ │ └─SwiGLUMLP: 3-11 [32, 1024, 384] [32, 1024, 384] 884,736 True
50
+ │ │ └─RMSNorm: 3-12 [32, 1024, 384] [32, 1024, 384] 384 True
51
+ │ └─TransformerBlock: 2-3 [32, 1024, 384] [32, 1024, 384] -- True
52
+ │ │ └─RMSNorm: 3-13 [32, 1024, 384] [32, 1024, 384] 384 True
53
+ │ │ └─MultiHeadAttention: 3-14 [32, 1024, 384] [32, 1024, 384] 589,952 True
54
+ │ │ └─RMSNorm: 3-15 [32, 1024, 384] [32, 1024, 384] 384 True
55
+ │ │ └─RMSNorm: 3-16 [32, 1024, 384] [32, 1024, 384] 384 True
56
+ │ │ └─SwiGLUMLP: 3-17 [32, 1024, 384] [32, 1024, 384] 884,736 True
57
+ │ │ └─RMSNorm: 3-18 [32, 1024, 384] [32, 1024, 384] 384 True
58
+ │ └─TransformerBlock: 2-4 [32, 1024, 384] [32, 1024, 384] -- True
59
+ │ │ └─RMSNorm: 3-19 [32, 1024, 384] [32, 1024, 384] 384 True
60
+ │ │ └─MultiHeadAttention: 3-20 [32, 1024, 384] [32, 1024, 384] 589,952 True
61
+ │ │ └─RMSNorm: 3-21 [32, 1024, 384] [32, 1024, 384] 384 True
62
+ │ │ └─RMSNorm: 3-22 [32, 1024, 384] [32, 1024, 384] 384 True
63
+ │ │ └─SwiGLUMLP: 3-23 [32, 1024, 384] [32, 1024, 384] 884,736 True
64
+ │ │ └─RMSNorm: 3-24 [32, 1024, 384] [32, 1024, 384] 384 True
65
+ │ └─TransformerBlock: 2-5 [32, 1024, 384] [32, 1024, 384] -- True
66
+ │ │ └─RMSNorm: 3-25 [32, 1024, 384] [32, 1024, 384] 384 True
67
+ │ │ └─MultiHeadAttention: 3-26 [32, 1024, 384] [32, 1024, 384] 589,952 True
68
+ │ │ └─RMSNorm: 3-27 [32, 1024, 384] [32, 1024, 384] 384 True
69
+ │ │ └─RMSNorm: 3-28 [32, 1024, 384] [32, 1024, 384] 384 True
70
+ │ │ └─SwiGLUMLP: 3-29 [32, 1024, 384] [32, 1024, 384] 884,736 True
71
+ │ │ └─RMSNorm: 3-30 [32, 1024, 384] [32, 1024, 384] 384 True
72
+ │ └─TransformerBlock: 2-6 [32, 1024, 384] [32, 1024, 384] -- True
73
+ │ │ └─RMSNorm: 3-31 [32, 1024, 384] [32, 1024, 384] 384 True
74
+ │ │ └─MultiHeadAttention: 3-32 [32, 1024, 384] [32, 1024, 384] 589,952 True
75
+ │ │ └─RMSNorm: 3-33 [32, 1024, 384] [32, 1024, 384] 384 True
76
+ │ │ └─RMSNorm: 3-34 [32, 1024, 384] [32, 1024, 384] 384 True
77
+ │ │ └─SwiGLUMLP: 3-35 [32, 1024, 384] [32, 1024, 384] 884,736 True
78
+ │ │ └─RMSNorm: 3-36 [32, 1024, 384] [32, 1024, 384] 384 True
79
+ │ └─TransformerBlock: 2-7 [32, 1024, 384] [32, 1024, 384] -- True
80
+ │ │ └─RMSNorm: 3-37 [32, 1024, 384] [32, 1024, 384] 384 True
81
+ │ │ └─MultiHeadAttention: 3-38 [32, 1024, 384] [32, 1024, 384] 589,952 True
82
+ │ │ └─RMSNorm: 3-39 [32, 1024, 384] [32, 1024, 384] 384 True
83
+ │ │ └─RMSNorm: 3-40 [32, 1024, 384] [32, 1024, 384] 384 True
84
+ │ │ └─SwiGLUMLP: 3-41 [32, 1024, 384] [32, 1024, 384] 884,736 True
85
+ │ │ └─RMSNorm: 3-42 [32, 1024, 384] [32, 1024, 384] 384 True
86
+ │ └─TransformerBlock: 2-8 [32, 1024, 384] [32, 1024, 384] -- True
87
+ │ │ └─RMSNorm: 3-43 [32, 1024, 384] [32, 1024, 384] 384 True
88
+ │ │ └─MultiHeadAttention: 3-44 [32, 1024, 384] [32, 1024, 384] 589,952 True
89
+ │ │ └─RMSNorm: 3-45 [32, 1024, 384] [32, 1024, 384] 384 True
90
+ │ │ └─RMSNorm: 3-46 [32, 1024, 384] [32, 1024, 384] 384 True
91
+ │ │ └─SwiGLUMLP: 3-47 [32, 1024, 384] [32, 1024, 384] 884,736 True
92
+ │ │ └─RMSNorm: 3-48 [32, 1024, 384] [32, 1024, 384] 384 True
93
+ │ └─TransformerBlock: 2-9 [32, 1024, 384] [32, 1024, 384] -- True
94
+ │ │ └─RMSNorm: 3-49 [32, 1024, 384] [32, 1024, 384] 384 True
95
+ │ │ └─MultiHeadAttention: 3-50 [32, 1024, 384] [32, 1024, 384] 589,952 True
96
+ │ │ └─RMSNorm: 3-51 [32, 1024, 384] [32, 1024, 384] 384 True
97
+ │ │ └─RMSNorm: 3-52 [32, 1024, 384] [32, 1024, 384] 384 True
98
+ │ │ └─SwiGLUMLP: 3-53 [32, 1024, 384] [32, 1024, 384] 884,736 True
99
+ │ │ └─RMSNorm: 3-54 [32, 1024, 384] [32, 1024, 384] 384 True
100
+ │ └─TransformerBlock: 2-10 [32, 1024, 384] [32, 1024, 384] -- True
101
+ │ │ └─RMSNorm: 3-55 [32, 1024, 384] [32, 1024, 384] 384 True
102
+ │ │ └─MultiHeadAttention: 3-56 [32, 1024, 384] [32, 1024, 384] 589,952 True
103
+ │ │ └─RMSNorm: 3-57 [32, 1024, 384] [32, 1024, 384] 384 True
104
+ │ │ └─RMSNorm: 3-58 [32, 1024, 384] [32, 1024, 384] 384 True
105
+ │ │ └─SwiGLUMLP: 3-59 [32, 1024, 384] [32, 1024, 384] 884,736 True
106
+ │ │ └─RMSNorm: 3-60 [32, 1024, 384] [32, 1024, 384] 384 True
107
+ │ └─TransformerBlock: 2-11 [32, 1024, 384] [32, 1024, 384] -- True
108
+ │ │ └─RMSNorm: 3-61 [32, 1024, 384] [32, 1024, 384] 384 True
109
+ │ │ └─MultiHeadAttention: 3-62 [32, 1024, 384] [32, 1024, 384] 589,952 True
110
+ │ │ └─RMSNorm: 3-63 [32, 1024, 384] [32, 1024, 384] 384 True
111
+ │ │ └─RMSNorm: 3-64 [32, 1024, 384] [32, 1024, 384] 384 True
112
+ │ │ └─SwiGLUMLP: 3-65 [32, 1024, 384] [32, 1024, 384] 884,736 True
113
+ │ │ └─RMSNorm: 3-66 [32, 1024, 384] [32, 1024, 384] 384 True
114
+ │ └─TransformerBlock: 2-12 [32, 1024, 384] [32, 1024, 384] -- True
115
+ │ │ └─RMSNorm: 3-67 [32, 1024, 384] [32, 1024, 384] 384 True
116
+ │ │ └─MultiHeadAttention: 3-68 [32, 1024, 384] [32, 1024, 384] 589,952 True
117
+ │ │ └─RMSNorm: 3-69 [32, 1024, 384] [32, 1024, 384] 384 True
118
+ │ │ └─RMSNorm: 3-70 [32, 1024, 384] [32, 1024, 384] 384 True
119
+ │ │ └─SwiGLUMLP: 3-71 [32, 1024, 384] [32, 1024, 384] 884,736 True
120
+ │ │ └─RMSNorm: 3-72 [32, 1024, 384] [32, 1024, 384] 384 True
121
+ │ └─TransformerBlock: 2-13 [32, 1024, 384] [32, 1024, 384] -- True
122
+ │ │ └─RMSNorm: 3-73 [32, 1024, 384] [32, 1024, 384] 384 True
123
+ │ │ └─MultiHeadAttention: 3-74 [32, 1024, 384] [32, 1024, 384] 589,952 True
124
+ │ │ └─RMSNorm: 3-75 [32, 1024, 384] [32, 1024, 384] 384 True
125
+ │ │ └─RMSNorm: 3-76 [32, 1024, 384] [32, 1024, 384] 384 True
126
+ │ │ └─SwiGLUMLP: 3-77 [32, 1024, 384] [32, 1024, 384] 884,736 True
127
+ │ │ └─RMSNorm: 3-78 [32, 1024, 384] [32, 1024, 384] 384 True
128
+ │ └─TransformerBlock: 2-14 [32, 1024, 384] [32, 1024, 384] -- True
129
+ │ │ └─RMSNorm: 3-79 [32, 1024, 384] [32, 1024, 384] 384 True
130
+ │ │ └─MultiHeadAttention: 3-80 [32, 1024, 384] [32, 1024, 384] 589,952 True
131
+ │ │ └─RMSNorm: 3-81 [32, 1024, 384] [32, 1024, 384] 384 True
132
+ │ │ └─RMSNorm: 3-82 [32, 1024, 384] [32, 1024, 384] 384 True
133
+ │ │ └─SwiGLUMLP: 3-83 [32, 1024, 384] [32, 1024, 384] 884,736 True
134
+ │ │ └─RMSNorm: 3-84 [32, 1024, 384] [32, 1024, 384] 384 True
135
+ ├─RMSNorm: 1-4 [32, 1024, 384] [32, 1024, 384] 384 True
136
+ ├─Linear: 1-5 [32, 1024, 384] [32, 1024, 50258] 19,299,072 True
137
+ ============================================================================================================================================
138
+ Total params: 59,265,664
139
+ Trainable params: 59,265,664
140
+ Non-trainable params: 0
141
+ Total mult-adds (Units.GIGABYTES): 1.90
142
+ ============================================================================================================================================
143
+ Input size (MB): 0.26
144
+ Forward/backward pass size (MB): 34515.45
145
+ Params size (MB): 237.06
146
+ Estimated Total Size (MB): 34752.78
147
+ ============================================================================================================================================
148
+ ============================================================
149
+ 2026-04-08 12:15:52 - INFO - parrotllm.training - Starting evaluation...
150
+ 2026-04-08 12:15:56 - INFO - parrotllm.training - Initial val: loss=10.9061, ppl=54505.50
151
+ 2026-04-08 12:15:58 - INFO - parrotllm.training -
152
+ ============================================================
153
+ Starting training...
154
+ Steps per epoch (approx): 3
155
+ Max steps: 12
156
+ ============================================================
157
+ 2026-04-08 12:15:58 - INFO - parrotllm.training - LR schedule=wsd | decay_ratio=0.179 | z_loss_coeff=0.000275
158
+ 2026-04-08 12:16:21 - INFO - parrotllm.training - step 1 | epoch 0 | loss 10.9259 | lr 3.26e-04 | grad 6.7907
159
+ 2026-04-08 12:16:21 - INFO - parrotllm.training - step 2 | epoch 0 | loss 10.6275 | lr 6.52e-04 | grad 17.0985
160
+ 2026-04-08 12:16:21 - INFO - parrotllm.training - step 3 | epoch 0 | loss 10.2679 | lr 9.79e-04 | grad 1.8126
161
+ 2026-04-08 12:16:21 - INFO - parrotllm.training - Starting evaluation...
162
+ 2026-04-08 12:16:21 - INFO - parrotllm.training - ------------------------------------------------------------
163
+ 2026-04-08 12:16:24 - INFO - parrotllm.training - Epoch 1 complete:
164
+ 2026-04-08 12:16:24 - INFO - parrotllm.training - Train: loss=10.2679, ppl=28794.11
165
+ 2026-04-08 12:16:24 - INFO - parrotllm.training - Val: loss=9.9140, ppl=20211.54
166
+ 2026-04-08 12:16:24 - INFO - parrotllm.training - ** New best validation loss! **
167
+ 2026-04-08 12:16:25 - INFO - parrotllm.training - Saved best checkpoint: /home/user/code/ParrotLLM/runs/smoketest/exp_c/run_20260408_121550/checkpoints/best_loss_9p9140_epoch_0001_step_0000003.pt
168
+ 2026-04-08 12:16:26 - INFO - parrotllm.training - ------------------------------------------------------------
169
+ 2026-04-08 12:16:26 - INFO - parrotllm.training - step 4 | epoch 1 | loss 9.9296 | lr 1.30e-03 | grad 1.6610
170
+ 2026-04-08 12:16:26 - INFO - parrotllm.training - step 5 | epoch 1 | loss 9.4699 | lr 1.30e-03 | grad 1.5961
171
+ 2026-04-08 12:16:27 - INFO - parrotllm.training - step 6 | epoch 1 | loss 9.0812 | lr 1.30e-03 | grad 1.5881
172
+ 2026-04-08 12:16:27 - INFO - parrotllm.training - Starting evaluation...
173
+ 2026-04-08 12:16:27 - INFO - parrotllm.training - ------------------------------------------------------------
174
+ 2026-04-08 12:16:29 - INFO - parrotllm.training - Epoch 2 complete:
175
+ 2026-04-08 12:16:29 - INFO - parrotllm.training - Train: loss=9.0812, ppl=8788.90
176
+ 2026-04-08 12:16:29 - INFO - parrotllm.training - Val: loss=8.7229, ppl=6142.18
177
+ 2026-04-08 12:16:29 - INFO - parrotllm.training - ** New best validation loss! **
178
+ 2026-04-08 12:16:30 - INFO - parrotllm.training - Saved best checkpoint: /home/user/code/ParrotLLM/runs/smoketest/exp_c/run_20260408_121550/checkpoints/best_loss_8p7229_epoch_0002_step_0000006.pt
179
+ 2026-04-08 12:16:31 - INFO - parrotllm.training - ------------------------------------------------------------
180
+ 2026-04-08 12:16:31 - INFO - parrotllm.training - step 7 | epoch 2 | loss 8.7008 | lr 1.30e-03 | grad 1.5313
181
+ 2026-04-08 12:16:31 - INFO - parrotllm.training - step 8 | epoch 2 | loss 8.3990 | lr 1.30e-03 | grad 1.4772
182
+ 2026-04-08 12:16:32 - INFO - parrotllm.training - step 9 | epoch 2 | loss 8.1599 | lr 1.30e-03 | grad 1.3179
183
+ 2026-04-08 12:16:32 - INFO - parrotllm.training - Starting evaluation...
184
+ 2026-04-08 12:16:32 - INFO - parrotllm.training - ------------------------------------------------------------
185
+ 2026-04-08 12:16:34 - INFO - parrotllm.training - Epoch 3 complete:
186
+ 2026-04-08 12:16:34 - INFO - parrotllm.training - Train: loss=8.1599, ppl=3497.83
187
+ 2026-04-08 12:16:34 - INFO - parrotllm.training - Val: loss=7.9581, ppl=2858.66
188
+ 2026-04-08 12:16:34 - INFO - parrotllm.training - ** New best validation loss! **
189
+ 2026-04-08 12:16:35 - INFO - parrotllm.training - Saved best checkpoint: /home/user/code/ParrotLLM/runs/smoketest/exp_c/run_20260408_121550/checkpoints/best_loss_7p9581_epoch_0003_step_0000009.pt
190
+ 2026-04-08 12:16:36 - INFO - parrotllm.training - ------------------------------------------------------------
191
+ 2026-04-08 12:16:36 - INFO - parrotllm.training - step 10 | epoch 3 | loss 7.9260 | lr 1.30e-03 | grad 1.1342
192
+ 2026-04-08 12:16:37 - INFO - parrotllm.training - step 11 | epoch 3 | loss 7.7364 | lr 7.18e-04 | grad 0.9191
193
+ 2026-04-08 12:16:37 - INFO - parrotllm.training - step 12 | epoch 3 | loss 7.8158 | lr 1.30e-04 | grad 0.7293
194
+ 2026-04-08 12:16:37 - INFO - parrotllm.training - Starting evaluation...
195
+ 2026-04-08 12:16:37 - INFO - parrotllm.training - ------------------------------------------------------------
196
+ 2026-04-08 12:16:39 - INFO - parrotllm.training - Epoch 4 complete:
197
+ 2026-04-08 12:16:39 - INFO - parrotllm.training - Train: loss=7.8158, ppl=2479.53
198
+ 2026-04-08 12:16:39 - INFO - parrotllm.training - Val: loss=7.7412, ppl=2301.29
199
+ 2026-04-08 12:16:39 - INFO - parrotllm.training - ** New best validation loss! **
200
+ 2026-04-08 12:16:42 - INFO - parrotllm.training - Saved best checkpoint: /home/user/code/ParrotLLM/runs/smoketest/exp_c/run_20260408_121550/checkpoints/best_loss_7p7412_epoch_0004_step_0000012.pt
201
+ 2026-04-08 12:16:42 - INFO - parrotllm.training - ------------------------------------------------------------
202
+ 2026-04-08 12:16:45 - INFO - parrotllm.training - Saved final checkpoint: /home/user/code/ParrotLLM/runs/smoketest/exp_c/run_20260408_121550/checkpoints/last_epoch_0004_step_0000012.pt
203
+ 2026-04-08 12:16:45 - INFO - parrotllm.training -
204
+ Loss Curve:
205
+ Loss
206
+ ^
207
+ 10.93 |*
208
+ | *
209
+ | *
210
+ | *
211
+ | *
212
+ | * *
213
+ | *
214
+ 7.74 | * * * *
215
+ +--------------------------------------------------> step
216
+ 1 12
217
+ 2026-04-08 12:16:45 - INFO - parrotllm.training -
218
+ ============================================================
219
+ TRAINING COMPLETE
220
+ ============================================================
221
+ Epochs: 4
222
+ Total steps: 12
223
+ Total time: 0.01 hours
224
+ Best validation loss: 7.7412
225
+ Run directory: runs/smoketest/exp_c/run_20260408_121550
226
+ ============================================================
227
+ 2026-04-08 12:16:45 - INFO - parrotllm.training - Uploading completed run directory to Hugging Face: repo=ParrotLabs/Preprocessed
runs/smoketest/exp_c/run_20260408_121550/training_plots.pdf ADDED
Binary file (21.1 kB). View file