winterdewdev commited on
Commit
904aed4
·
verified ·
1 Parent(s): 1b3f321

C2: best epoch 100, val loss 0.0993

Browse files
Files changed (6) hide show
  1. C2_best.pt +3 -0
  2. README.md +86 -0
  3. config.json +113 -0
  4. summary.json +1964 -0
  5. tokenizer_cipher.json +0 -0
  6. tokenizer_plain.json +0 -0
C2_best.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fde9061a389146483bea353f589f1d293c484b3ccf0acb46d2f9761442968db5
3
+ size 33282087
README.md ADDED
@@ -0,0 +1,86 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: pytorch
3
+ pipeline_tag: translation
4
+ tags:
5
+ - transformer
6
+ - seq2seq
7
+ - cryptanalysis
8
+ - from-scratch
9
+ - anlp
10
+ ---
11
+
12
+ # C2 -- RoPE instead of sinusoidal absolute
13
+
14
+ Encoder-decoder transformer written from scratch -- no `nn.Transformer`, no
15
+ pretrained weights, tokenizer fitted in-repo -- trained to decipher a binary
16
+ cipher into English plaintext. One row of a five-configuration ablation in
17
+ which exactly one architectural choice moves at a time.
18
+
19
+ | | |
20
+ | --- | --- |
21
+ | configuration | `C2` -- RoPE instead of sinusoidal absolute |
22
+ | positional encoding | rope |
23
+ | attention | mha |
24
+ | normalisation | layernorm |
25
+ | tokenisation | unigram LM subword, fitted on the training split |
26
+ | depth | 2 encoder / 2 decoder layers |
27
+ | width | dim_model 128, 4 heads, dim_ff 512 |
28
+ | parameters | 2,763,264 |
29
+ | windowing | 1024-character source windows at stride 768 (25,654 training windows) |
30
+ | best epoch | 100 (val loss 0.0993) |
31
+
32
+ ## Test metrics
33
+
34
+ | metric | test | test_full | gen_test_win | gen_test | gen_test_full |
35
+ | --- | --- | --- | --- | --- | --- |
36
+ | loss | 0.1198 | 5.9206 | -- | -- | -- |
37
+ | perplexity | 1.1273 | 372.6371 | -- | -- | -- |
38
+ | token acc | 0.9747 | 0.3038 | 0.8211 | 0.4405 | 0.0425 |
39
+ | bit acc | -- | -- | 0.9561 | 0.8182 | 0.3518 |
40
+ | seq acc | 0.5632 | 0.0528 | 0.6598 | 0.2040 | 0.0540 |
41
+ | levenshtein | -- | -- | 1.5377 | 22.9500 | 505.6000 |
42
+ | lev / len | -- | -- | 0.0103 | 0.0342 | 0.6313 |
43
+ | BLEU | -- | -- | 0.9453 | 0.9246 | 0.1027 |
44
+ | ROUGE-1 | -- | -- | 0.9753 | 0.9620 | 0.3624 |
45
+ | ROUGE-2 | -- | -- | 0.9555 | 0.9412 | 0.2560 |
46
+ | ROUGE-L | -- | -- | 0.9753 | 0.9619 | 0.3113 |
47
+
48
+ - `test` -- teacher-forced, per window
49
+ - `test_full` -- teacher-forced, per whole line
50
+ - `gen_test_win` -- greedy, per window
51
+ - `gen_test` -- greedy, windowed and stitched
52
+ - `gen_test_full` -- greedy, whole line in one pass
53
+
54
+ ## Files
55
+
56
+ - `C2_best.pt` -- weights, optimiser state, config and vocabulary sizes,
57
+ from the best epoch.
58
+ - `config.json` -- the full ExperimentConfig this was trained under.
59
+ - `summary.json` -- the run record: the metrics above plus per-epoch history.
60
+ - `tokenizer_cipher.json`, `tokenizer_plain.json` -- the fitted unigram
61
+ vocabularies, when the config uses them. Byte-level configurations have none
62
+ by construction.
63
+
64
+ ## Loading it
65
+
66
+ The checkpoint is a plain `torch.save` dict, not a `transformers` model. The
67
+ vocabulary sizes travel with it because they are not recoverable from the
68
+ config -- the tokenizer lands wherever pruning leaves it:
69
+
70
+ ```python
71
+ import torch
72
+ from huggingface_hub import hf_hub_download
73
+
74
+ from src.config import get_config
75
+ from src.models.base import build_model
76
+
77
+ path = hf_hub_download("winterdewdev/anlp-a1-transformers-C2", "C2_best.pt")
78
+ state = torch.load(path, map_location="cpu", weights_only=False)
79
+
80
+ cfg = get_config(state["config_name"])
81
+ model = build_model(cfg, state["src_vocab_size"], state["tgt_vocab_size"])
82
+ model.load_state_dict(state["model"])
83
+ model.eval()
84
+ ```
85
+
86
+ Trained with `python -m src.train --config C2`.
config.json ADDED
@@ -0,0 +1,113 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "name": "C2",
3
+ "description": "RoPE instead of sinusoidal absolute",
4
+ "model": {
5
+ "dim_model": 128,
6
+ "num_heads": 4,
7
+ "num_encoder_layers": 2,
8
+ "num_decoder_layers": 2,
9
+ "dim_ff": 512,
10
+ "dropout": 0.1,
11
+ "norm_eps": 1e-06,
12
+ "num_kv_heads": 2,
13
+ "rope_base": 10000.0,
14
+ "pos_encoding": "rope",
15
+ "attention": "mha",
16
+ "norm": "layernorm",
17
+ "tokenization": "subword",
18
+ "tie_output_embedding": false
19
+ },
20
+ "data": {
21
+ "cipher_path": "/home2/winterdew/ANLP_A1/ANLP-M26-Assignment-1/dataset/brown_cipher.txt",
22
+ "plain_path": "/home2/winterdew/ANLP_A1/ANLP-M26-Assignment-1/dataset/brown_plain.txt",
23
+ "split_ratios": [
24
+ 0.8,
25
+ 0.1,
26
+ 0.1
27
+ ],
28
+ "split_seed": 1234,
29
+ "max_src_len_subword": 768,
30
+ "max_tgt_len_subword": 256,
31
+ "max_src_len_byte": 3072,
32
+ "max_tgt_len_byte": 1152,
33
+ "length_policy": "drop",
34
+ "effective_batch_size": 64,
35
+ "micro_batch_size_subword": 64,
36
+ "micro_batch_size_byte": 64,
37
+ "num_workers": 0,
38
+ "verbose": true,
39
+ "tokenizer": {
40
+ "src_vocab_size": 4096,
41
+ "tgt_vocab_size": 4096,
42
+ "src_seed_size": 20000,
43
+ "tgt_seed_size": 50000,
44
+ "src_max_segment": 16,
45
+ "tgt_max_segment": 32,
46
+ "max_piece_len": 16,
47
+ "shrink_factor": 0.75,
48
+ "em_iters": 2,
49
+ "min_freq": 2,
50
+ "sample_seed": 1234,
51
+ "cache_dir": "/home2/winterdew/ANLP_A1/ANLP-M26-Assignment-1/outputs/tokenizers",
52
+ "retrain": false
53
+ },
54
+ "window": {
55
+ "enabled": true,
56
+ "src_window": 1024,
57
+ "src_stride": 768,
58
+ "min_tgt_chars": 16,
59
+ "eval_full_sentences": true
60
+ }
61
+ },
62
+ "train": {
63
+ "epochs": 100,
64
+ "lr": 0.0007,
65
+ "betas": [
66
+ 0.9,
67
+ 0.98
68
+ ],
69
+ "adam_eps": 1e-09,
70
+ "weight_decay": 0.01,
71
+ "grad_clip": 1.0,
72
+ "label_smoothing": 0.05,
73
+ "scheduler": "cosine",
74
+ "warmup_steps": 400,
75
+ "min_lr": 1e-05,
76
+ "seed": 26,
77
+ "device": "auto",
78
+ "amp": true,
79
+ "eval_every_epochs": 1,
80
+ "generate_every_epochs": 5,
81
+ "max_eval_batches": null,
82
+ "length_slack": 16,
83
+ "gen_eval_examples": 128,
84
+ "gen_batch_size": 32,
85
+ "test_gen_examples": null,
86
+ "log_every_steps": 50,
87
+ "checkpoint_dir": "/home2/winterdew/ANLP_A1/ANLP-M26-Assignment-1/outputs/checkpoints",
88
+ "keep_best_only": true,
89
+ "early_stopping_patience": null,
90
+ "wandb_project": "anlp-a1-transformers",
91
+ "wandb_entity": null,
92
+ "wandb_mode": "online"
93
+ },
94
+ "blt": {
95
+ "patch_size": 8,
96
+ "local_dim": 128,
97
+ "local_heads": 4,
98
+ "local_dim_ff": 512,
99
+ "local_encoder_layers": 1,
100
+ "local_decoder_layers": 1,
101
+ "local_dropout": 0.1
102
+ },
103
+ "derived": {
104
+ "token_free": false,
105
+ "max_src_len": 1026,
106
+ "max_tgt_len": 1026,
107
+ "micro_batch_size": 64,
108
+ "windowed": true,
109
+ "grad_accum_steps": 1,
110
+ "pe_max_len": 1042,
111
+ "head_dim": 32
112
+ }
113
+ }
summary.json ADDED
@@ -0,0 +1,1964 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "config": "C2",
3
+ "description": "RoPE instead of sinusoidal absolute",
4
+ "parameters": 2763264,
5
+ "windowed": true,
6
+ "window": {
7
+ "src_window": 1024,
8
+ "src_stride": 768,
9
+ "train_examples": 25654
10
+ },
11
+ "best_epoch": 100,
12
+ "best_val_loss": 0.09927769922586938,
13
+ "test": {
14
+ "loss": 0.11979607583332225,
15
+ "ppl": 1.1272669511654074,
16
+ "token_acc": 0.9746887539870357,
17
+ "seq_acc": 0.56318347509113
18
+ },
19
+ "test_full": {
20
+ "loss": 5.920605090370729,
21
+ "ppl": 372.63712480831,
22
+ "token_acc": 0.30378654455794296,
23
+ "seq_acc": 0.052752293577981654
24
+ },
25
+ "gen_test_win": {
26
+ "n": 3292.0,
27
+ "seq_acc": 0.6597812879708383,
28
+ "bit_acc": 0.9560890886286575,
29
+ "lev": 1.537667071688943,
30
+ "lev_norm": 0.010323837570599004,
31
+ "token_acc": 0.8211193850588517,
32
+ "bleu": 0.9453322550492058,
33
+ "rouge1": 0.9753460409870339,
34
+ "rouge2": 0.9555024454872828,
35
+ "rougeL": 0.9753460409870339
36
+ },
37
+ "gen_test": {
38
+ "n": 500.0,
39
+ "seq_acc": 0.204,
40
+ "bit_acc": 0.8181579795401267,
41
+ "lev": 22.95,
42
+ "lev_norm": 0.034200148953254766,
43
+ "token_acc": 0.44049529344873334,
44
+ "bleu": 0.9245620567121133,
45
+ "rouge1": 0.9619928822649874,
46
+ "rouge2": 0.9411761597516598,
47
+ "rougeL": 0.9619324395189165
48
+ },
49
+ "gen_test_full": {
50
+ "n": 500.0,
51
+ "seq_acc": 0.054,
52
+ "bit_acc": 0.3517906454336038,
53
+ "lev": 505.6,
54
+ "lev_norm": 0.6312979292199293,
55
+ "token_acc": 0.04248673746235677,
56
+ "bleu": 0.1026796740588983,
57
+ "rouge1": 0.36243435714957306,
58
+ "rouge2": 0.25601572873722805,
59
+ "rougeL": 0.3112700877732138
60
+ },
61
+ "checkpoint": "/home2/winterdew/ANLP_A1/ANLP-M26-Assignment-1/outputs/checkpoints/C2_best.pt",
62
+ "history": [
63
+ {
64
+ "train/epoch_loss": 6.995370723362873,
65
+ "epoch": 1,
66
+ "val/loss": 5.975532108766493,
67
+ "val/ppl": 393.67752463013096,
68
+ "val/token_acc": 0.12048348974279702,
69
+ "val/seq_acc": 0.0,
70
+ "val_full/loss": 6.270644691050796,
71
+ "val_full/ppl": 528.818192361973,
72
+ "val_full/token_acc": 0.07332962884899116,
73
+ "val_full/seq_acc": 0.0,
74
+ "system/peak_gpu_mem_gb": 1.0650625228881836
75
+ },
76
+ {
77
+ "train/epoch_loss": 5.833912618736971,
78
+ "epoch": 2,
79
+ "val/loss": 5.281581922754959,
80
+ "val/ppl": 196.68076316142339,
81
+ "val/token_acc": 0.18133857246861007,
82
+ "val/seq_acc": 0.0,
83
+ "val_full/loss": 5.8494414531797885,
84
+ "val_full/ppl": 347.04048797371195,
85
+ "val_full/token_acc": 0.10122822817068731,
86
+ "val_full/seq_acc": 0.0,
87
+ "system/peak_gpu_mem_gb": 1.0650625228881836
88
+ },
89
+ {
90
+ "train/epoch_loss": 5.267378161375659,
91
+ "epoch": 3,
92
+ "val/loss": 4.599296502182211,
93
+ "val/ppl": 99.41435325505417,
94
+ "val/token_acc": 0.26174068274974815,
95
+ "val/seq_acc": 0.0,
96
+ "val_full/loss": 5.996897170021461,
97
+ "val_full/ppl": 402.17896254727714,
98
+ "val_full/token_acc": 0.0968594914637184,
99
+ "val_full/seq_acc": 0.0,
100
+ "system/peak_gpu_mem_gb": 1.0650625228881836
101
+ },
102
+ {
103
+ "train/epoch_loss": 4.551063683859428,
104
+ "epoch": 4,
105
+ "val/loss": 3.5084439771734224,
106
+ "val/ppl": 33.39626198735998,
107
+ "val/token_acc": 0.38315165900409887,
108
+ "val/seq_acc": 0.0,
109
+ "val_full/loss": 6.7181524872339,
110
+ "val_full/ppl": 827.2876741803625,
111
+ "val_full/token_acc": 0.1017839008220123,
112
+ "val_full/seq_acc": 0.0,
113
+ "system/peak_gpu_mem_gb": 1.0650625228881836
114
+ },
115
+ {
116
+ "train/epoch_loss": 3.580359074242989,
117
+ "epoch": 5,
118
+ "val/loss": 2.288402971404525,
119
+ "val/ppl": 9.859179709632619,
120
+ "val/token_acc": 0.5378125280489873,
121
+ "val/seq_acc": 0.00032123353678124,
122
+ "val_full/loss": 7.171605172426181,
123
+ "val_full/ppl": 1301.9327529909626,
124
+ "val_full/token_acc": 0.1069382436911993,
125
+ "val_full/seq_acc": 0.0,
126
+ "system/peak_gpu_mem_gb": 1.0650625228881836,
127
+ "gen_val_win/n": 128.0,
128
+ "gen_val_win/seq_acc": 0.0,
129
+ "gen_val_win/bit_acc": 0.6191782853752595,
130
+ "gen_val_win/lev": 62.53125,
131
+ "gen_val_win/lev_norm": 0.45235598570141544,
132
+ "gen_val_win/token_acc": 0.07652274479568234,
133
+ "gen_val_win/bleu": 0.15222595619365087,
134
+ "gen_val_win/rouge1": 0.42920903000545196,
135
+ "gen_val_win/rouge2": 0.21950845824242435,
136
+ "gen_val_win/rougeL": 0.4219266487165362,
137
+ "gen_val/n": 128.0,
138
+ "gen_val/seq_acc": 0.0,
139
+ "gen_val/bit_acc": 0.5144970288021381,
140
+ "gen_val/lev": 425.484375,
141
+ "gen_val/lev_norm": 0.5156729614081624,
142
+ "gen_val/token_acc": 0.015991277485008178,
143
+ "gen_val/bleu": 0.12413074730544832,
144
+ "gen_val/rouge1": 0.40704193764823327,
145
+ "gen_val/rouge2": 0.19375547017279462,
146
+ "gen_val/rougeL": 0.37537546462183163,
147
+ "gen_val_full/n": 128.0,
148
+ "gen_val_full/seq_acc": 0.0,
149
+ "gen_val_full/bit_acc": 0.1293192520602159,
150
+ "gen_val_full/lev": 533.875,
151
+ "gen_val_full/lev_norm": 0.8060229206813219,
152
+ "gen_val_full/token_acc": 0.011334698157487392,
153
+ "gen_val_full/bleu": 0.0005790934932958215,
154
+ "gen_val_full/rouge1": 0.1511024572918868,
155
+ "gen_val_full/rouge2": 0.047324442374536424,
156
+ "gen_val_full/rougeL": 0.13051932962132223
157
+ },
158
+ {
159
+ "train/epoch_loss": 2.7552905706990685,
160
+ "epoch": 6,
161
+ "val/loss": 1.5545950089515226,
162
+ "val/ppl": 4.733169245769669,
163
+ "val/token_acc": 0.6527709906154322,
164
+ "val/seq_acc": 0.00128493414712496,
165
+ "val_full/loss": 7.066416803588879,
166
+ "val_full/ppl": 1171.941206280403,
167
+ "val_full/token_acc": 0.1304489451800188,
168
+ "val_full/seq_acc": 0.0,
169
+ "system/peak_gpu_mem_gb": 1.0650625228881836
170
+ },
171
+ {
172
+ "train/epoch_loss": 2.2788533207187034,
173
+ "epoch": 7,
174
+ "val/loss": 1.151090940280356,
175
+ "val/ppl": 3.1616401895944373,
176
+ "val/token_acc": 0.7287251548304096,
177
+ "val/seq_acc": 0.004497269514937359,
178
+ "val_full/loss": 7.7966491831372515,
179
+ "val_full/ppl": 2432.4376535790266,
180
+ "val_full/token_acc": 0.12631014198394297,
181
+ "val_full/seq_acc": 0.0,
182
+ "system/peak_gpu_mem_gb": 1.0650625228881836
183
+ },
184
+ {
185
+ "train/epoch_loss": 1.9741628886458285,
186
+ "epoch": 8,
187
+ "val/loss": 0.9012647076520286,
188
+ "val/ppl": 2.4627157579117815,
189
+ "val/token_acc": 0.7783406967119108,
190
+ "val/seq_acc": 0.008352071956312239,
191
+ "val_full/loss": 7.011426917070647,
192
+ "val_full/ppl": 1109.2361642020871,
193
+ "val_full/token_acc": 0.16438329916265879,
194
+ "val_full/seq_acc": 0.0023094688221709007,
195
+ "system/peak_gpu_mem_gb": 1.0650625228881836
196
+ },
197
+ {
198
+ "train/epoch_loss": 1.7607509602335027,
199
+ "epoch": 9,
200
+ "val/loss": 0.7322117839463648,
201
+ "val/ppl": 2.0796753170286784,
202
+ "val/token_acc": 0.8156196706924236,
203
+ "val/seq_acc": 0.01574044330228076,
204
+ "val_full/loss": 6.722441211270574,
205
+ "val_full/ppl": 830.8433018167867,
206
+ "val_full/token_acc": 0.1844833202398973,
207
+ "val_full/seq_acc": 0.0023094688221709007,
208
+ "system/peak_gpu_mem_gb": 1.0650625228881836
209
+ },
210
+ {
211
+ "train/epoch_loss": 1.6072707785632545,
212
+ "epoch": 10,
213
+ "val/loss": 0.638098754085413,
214
+ "val/ppl": 1.892878628248731,
215
+ "val/token_acc": 0.8388168064545083,
216
+ "val/seq_acc": 0.0273048506264054,
217
+ "val_full/loss": 6.620422362111747,
218
+ "val_full/ppl": 750.261912407472,
219
+ "val_full/token_acc": 0.20368276839947116,
220
+ "val_full/seq_acc": 0.009237875288683603,
221
+ "system/peak_gpu_mem_gb": 1.0650625228881836,
222
+ "gen_val_win/n": 128.0,
223
+ "gen_val_win/seq_acc": 0.0625,
224
+ "gen_val_win/bit_acc": 0.7596008799497171,
225
+ "gen_val_win/lev": 16.0078125,
226
+ "gen_val_win/lev_norm": 0.12610161520236987,
227
+ "gen_val_win/token_acc": 0.3641123274631128,
228
+ "gen_val_win/bleu": 0.6404382598407361,
229
+ "gen_val_win/rouge1": 0.8049537970832047,
230
+ "gen_val_win/rouge2": 0.68604201758791,
231
+ "gen_val_win/rougeL": 0.8042237353216579,
232
+ "gen_val/n": 128.0,
233
+ "gen_val/seq_acc": 0.0078125,
234
+ "gen_val/bit_acc": 0.5627391010305949,
235
+ "gen_val/lev": 197.3828125,
236
+ "gen_val/lev_norm": 0.2425218587382373,
237
+ "gen_val/token_acc": 0.06832864333557925,
238
+ "gen_val/bleu": 0.527264293628675,
239
+ "gen_val/rouge1": 0.7405585252014839,
240
+ "gen_val/rouge2": 0.6223908061616639,
241
+ "gen_val/rougeL": 0.7340760854405636,
242
+ "gen_val_full/n": 128.0,
243
+ "gen_val_full/seq_acc": 0.0078125,
244
+ "gen_val_full/bit_acc": 0.24735600187574205,
245
+ "gen_val_full/lev": 610.1328125,
246
+ "gen_val_full/lev_norm": 0.714567196137067,
247
+ "gen_val_full/token_acc": 0.018877166541070082,
248
+ "gen_val_full/bleu": 0.04987182029735536,
249
+ "gen_val_full/rouge1": 0.26309011076183286,
250
+ "gen_val_full/rouge2": 0.13794812321551939,
251
+ "gen_val_full/rougeL": 0.22134718759848468
252
+ },
253
+ {
254
+ "train/epoch_loss": 1.4831541544778686,
255
+ "epoch": 11,
256
+ "val/loss": 0.53435421331434,
257
+ "val/ppl": 1.706345950873767,
258
+ "val/token_acc": 0.8636794287480927,
259
+ "val/seq_acc": 0.043366527465467396,
260
+ "val_full/loss": 6.250100071973979,
261
+ "val_full/ppl": 518.0646658281336,
262
+ "val_full/token_acc": 0.24012722987602753,
263
+ "val_full/seq_acc": 0.006928406466512702,
264
+ "system/peak_gpu_mem_gb": 1.0650625228881836
265
+ },
266
+ {
267
+ "train/epoch_loss": 1.3880311767953888,
268
+ "epoch": 12,
269
+ "val/loss": 0.46726363072187654,
270
+ "val/ppl": 1.5956220029535655,
271
+ "val/token_acc": 0.8795663751234155,
272
+ "val/seq_acc": 0.06488917442981047,
273
+ "val_full/loss": 6.244773578364215,
274
+ "val_full/ppl": 515.3125338098071,
275
+ "val_full/token_acc": 0.24543486175247659,
276
+ "val_full/seq_acc": 0.009237875288683603,
277
+ "system/peak_gpu_mem_gb": 1.0650625228881836
278
+ },
279
+ {
280
+ "train/epoch_loss": 1.3117436664062843,
281
+ "epoch": 13,
282
+ "val/loss": 0.4200845250387076,
283
+ "val/ppl": 1.5220902049150007,
284
+ "val/token_acc": 0.8917234295060387,
285
+ "val/seq_acc": 0.07805974943784132,
286
+ "val_full/loss": 6.244714148934642,
287
+ "val_full/ppl": 515.281909989858,
288
+ "val_full/token_acc": 0.2527352507233325,
289
+ "val_full/seq_acc": 0.006928406466512702,
290
+ "system/peak_gpu_mem_gb": 1.0650625228881836
291
+ },
292
+ {
293
+ "train/epoch_loss": 1.2502200576729905,
294
+ "epoch": 14,
295
+ "val/loss": 0.3869152949997694,
296
+ "val/ppl": 1.4724317636578035,
297
+ "val/token_acc": 0.899023645919558,
298
+ "val/seq_acc": 0.09251525859299711,
299
+ "val_full/loss": 6.005919665183276,
300
+ "val_full/ppl": 405.8240394078294,
301
+ "val_full/token_acc": 0.2704976144398245,
302
+ "val_full/seq_acc": 0.011547344110854504,
303
+ "system/peak_gpu_mem_gb": 1.0650625228881836
304
+ },
305
+ {
306
+ "train/epoch_loss": 1.199270762112967,
307
+ "epoch": 15,
308
+ "val/loss": 0.35389138099325046,
309
+ "val/ppl": 1.4246004393910614,
310
+ "val/token_acc": 0.9086774840183104,
311
+ "val/seq_acc": 0.10986186957918406,
312
+ "val_full/loss": 6.192432492958286,
313
+ "val_full/ppl": 489.03423290158656,
314
+ "val_full/token_acc": 0.2578895935925195,
315
+ "val_full/seq_acc": 0.016166281755196306,
316
+ "system/peak_gpu_mem_gb": 1.0650625228881836,
317
+ "gen_val_win/n": 128.0,
318
+ "gen_val_win/seq_acc": 0.2265625,
319
+ "gen_val_win/bit_acc": 0.8369911956105652,
320
+ "gen_val_win/lev": 8.1875,
321
+ "gen_val_win/lev_norm": 0.06252630647566886,
322
+ "gen_val_win/token_acc": 0.5716736094094584,
323
+ "gen_val_win/bleu": 0.789540025710193,
324
+ "gen_val_win/rouge1": 0.8967285636361726,
325
+ "gen_val_win/rouge2": 0.8273286565962761,
326
+ "gen_val_win/rougeL": 0.8967285636361726,
327
+ "gen_val/n": 128.0,
328
+ "gen_val/seq_acc": 0.0078125,
329
+ "gen_val/bit_acc": 0.6282546518508778,
330
+ "gen_val/lev": 113.0859375,
331
+ "gen_val/lev_norm": 0.14689947429206585,
332
+ "gen_val/token_acc": 0.12761000483537738,
333
+ "gen_val/bleu": 0.6973874922009002,
334
+ "gen_val/rouge1": 0.8463161177245291,
335
+ "gen_val/rouge2": 0.7678106345672289,
336
+ "gen_val/rougeL": 0.843933897528292,
337
+ "gen_val_full/n": 128.0,
338
+ "gen_val_full/seq_acc": 0.0078125,
339
+ "gen_val_full/bit_acc": 0.2395406291805118,
340
+ "gen_val_full/lev": 706.4765625,
341
+ "gen_val_full/lev_norm": 0.7012449348833671,
342
+ "gen_val_full/token_acc": 0.02245118616040196,
343
+ "gen_val_full/bleu": 0.07419303060286554,
344
+ "gen_val_full/rouge1": 0.2679363814726357,
345
+ "gen_val_full/rouge2": 0.14769219870334008,
346
+ "gen_val_full/rougeL": 0.22277268388857696
347
+ },
348
+ {
349
+ "train/epoch_loss": 1.1549483582265954,
350
+ "epoch": 16,
351
+ "val/loss": 0.31556516386491046,
352
+ "val/ppl": 1.371033950883735,
353
+ "val/token_acc": 0.9177030248027844,
354
+ "val/seq_acc": 0.13363315130099582,
355
+ "val_full/loss": 5.822308026475886,
356
+ "val_full/ppl": 337.75069226512585,
357
+ "val_full/token_acc": 0.29075092452432505,
358
+ "val_full/seq_acc": 0.013856812933025405,
359
+ "system/peak_gpu_mem_gb": 1.0650625228881836
360
+ },
361
+ {
362
+ "train/epoch_loss": 1.118490819621859,
363
+ "epoch": 17,
364
+ "val/loss": 0.2969314056733401,
365
+ "val/ppl": 1.3457229873199037,
366
+ "val/token_acc": 0.9234674033369569,
367
+ "val/seq_acc": 0.1625441696113074,
368
+ "val_full/loss": 6.052475362983579,
369
+ "val_full/ppl": 425.1641641727175,
370
+ "val_full/token_acc": 0.27586272969399683,
371
+ "val_full/seq_acc": 0.023094688221709007,
372
+ "system/peak_gpu_mem_gb": 1.0650625228881836
373
+ },
374
+ {
375
+ "train/epoch_loss": 1.0856515222654082,
376
+ "epoch": 18,
377
+ "val/loss": 0.28207586462902334,
378
+ "val/ppl": 1.3258793035128686,
379
+ "val/token_acc": 0.9284538899582132,
380
+ "val/seq_acc": 0.17860584645036942,
381
+ "val_full/loss": 5.706772050863209,
382
+ "val_full/ppl": 300.89821483396275,
383
+ "val_full/token_acc": 0.29916265879783094,
384
+ "val_full/seq_acc": 0.03233256351039261,
385
+ "system/peak_gpu_mem_gb": 1.0650625228881836
386
+ },
387
+ {
388
+ "train/epoch_loss": 1.0577344556996353,
389
+ "epoch": 19,
390
+ "val/loss": 0.26301759980696454,
391
+ "val/ppl": 1.3008496134523408,
392
+ "val/token_acc": 0.9329118089976165,
393
+ "val/seq_acc": 0.19241888853196273,
394
+ "val_full/loss": 5.9044769222202,
395
+ "val_full/ppl": 366.67537587410794,
396
+ "val_full/token_acc": 0.2831247964130372,
397
+ "val_full/seq_acc": 0.025404157043879907,
398
+ "system/peak_gpu_mem_gb": 1.0650625228881836
399
+ },
400
+ {
401
+ "train/epoch_loss": 1.0343907209405876,
402
+ "epoch": 20,
403
+ "val/loss": 0.25522526245135446,
404
+ "val/ppl": 1.2907523461645032,
405
+ "val/token_acc": 0.9350161063517867,
406
+ "val/seq_acc": 0.20558946353999358,
407
+ "val_full/loss": 5.448175741056544,
408
+ "val_full/ppl": 232.3339418067963,
409
+ "val_full/token_acc": 0.31219222441510663,
410
+ "val_full/seq_acc": 0.02771362586605081,
411
+ "system/peak_gpu_mem_gb": 1.0650625228881836,
412
+ "gen_val_win/n": 128.0,
413
+ "gen_val_win/seq_acc": 0.234375,
414
+ "gen_val_win/bit_acc": 0.8554692484369019,
415
+ "gen_val_win/lev": 5.328125,
416
+ "gen_val_win/lev_norm": 0.03916613653801227,
417
+ "gen_val_win/token_acc": 0.6049562682215743,
418
+ "gen_val_win/bleu": 0.8381115916312833,
419
+ "gen_val_win/rouge1": 0.9267279458070664,
420
+ "gen_val_win/rouge2": 0.8700091054588754,
421
+ "gen_val_win/rougeL": 0.9267279458070664,
422
+ "gen_val/n": 128.0,
423
+ "gen_val/seq_acc": 0.0234375,
424
+ "gen_val/bit_acc": 0.6670638690052131,
425
+ "gen_val/lev": 81.796875,
426
+ "gen_val/lev_norm": 0.1102067239985438,
427
+ "gen_val/token_acc": 0.1839252506009888,
428
+ "gen_val/bleu": 0.7737386319407429,
429
+ "gen_val/rouge1": 0.8910038846170638,
430
+ "gen_val/rouge2": 0.8324292306391359,
431
+ "gen_val/rougeL": 0.8898227137651546,
432
+ "gen_val_full/n": 128.0,
433
+ "gen_val_full/seq_acc": 0.0234375,
434
+ "gen_val_full/bit_acc": 0.32297812137955617,
435
+ "gen_val_full/lev": 644.3828125,
436
+ "gen_val_full/lev_norm": 0.6681950287002262,
437
+ "gen_val_full/token_acc": 0.022691022901682265,
438
+ "gen_val_full/bleu": 0.09839284557174968,
439
+ "gen_val_full/rouge1": 0.3288217650557424,
440
+ "gen_val_full/rouge2": 0.19653596049434877,
441
+ "gen_val_full/rougeL": 0.2578958906244661
442
+ },
443
+ {
444
+ "train/epoch_loss": 1.0122020176223983,
445
+ "epoch": 21,
446
+ "val/loss": 0.23777997387577712,
447
+ "val/ppl": 1.26843007436625,
448
+ "val/token_acc": 0.940042484866013,
449
+ "val/seq_acc": 0.22518470928364923,
450
+ "val_full/loss": 5.419640183994711,
451
+ "val_full/ppl": 225.79786219875206,
452
+ "val_full/token_acc": 0.3242062503592711,
453
+ "val_full/seq_acc": 0.03695150115473441,
454
+ "system/peak_gpu_mem_gb": 1.0650625228881836
455
+ },
456
+ {
457
+ "train/epoch_loss": 0.9939897338351109,
458
+ "epoch": 22,
459
+ "val/loss": 0.22978879153996576,
460
+ "val/ppl": 1.2583342110300577,
461
+ "val/token_acc": 0.9416481335580577,
462
+ "val/seq_acc": 0.23514294892386764,
463
+ "val_full/loss": 5.366829052218859,
464
+ "val_full/ppl": 214.18262784325685,
465
+ "val_full/token_acc": 0.32307574393071337,
466
+ "val_full/seq_acc": 0.03233256351039261,
467
+ "system/peak_gpu_mem_gb": 1.0650625228881836
468
+ },
469
+ {
470
+ "train/epoch_loss": 0.9776245095486058,
471
+ "epoch": 23,
472
+ "val/loss": 0.22080388952579136,
473
+ "val/ppl": 1.2470788413586418,
474
+ "val/token_acc": 0.9448694039153893,
475
+ "val/seq_acc": 0.254416961130742,
476
+ "val_full/loss": 5.202963702240893,
477
+ "val_full/ppl": 181.8102757168398,
478
+ "val_full/token_acc": 0.33005039376113743,
479
+ "val_full/seq_acc": 0.03926096997690531,
480
+ "system/peak_gpu_mem_gb": 1.0650625228881836
481
+ },
482
+ {
483
+ "train/epoch_loss": 0.961638304955347,
484
+ "epoch": 24,
485
+ "val/loss": 0.21207630866494379,
486
+ "val/ppl": 1.2362422174724308,
487
+ "val/token_acc": 0.9466445931525566,
488
+ "val/seq_acc": 0.26019916479280436,
489
+ "val_full/loss": 5.282296377110119,
490
+ "val_full/ppl": 196.82133279855557,
491
+ "val_full/token_acc": 0.330242005020215,
492
+ "val_full/seq_acc": 0.046189376443418015,
493
+ "system/peak_gpu_mem_gb": 1.0650625228881836
494
+ },
495
+ {
496
+ "train/epoch_loss": 0.9482644060603401,
497
+ "epoch": 25,
498
+ "val/loss": 0.20377230890322381,
499
+ "val/ppl": 1.2260189680698952,
500
+ "val/token_acc": 0.9485494310418765,
501
+ "val/seq_acc": 0.28172181175714744,
502
+ "val_full/loss": 5.015769382077641,
503
+ "val_full/ppl": 150.77209349922734,
504
+ "val_full/token_acc": 0.3471037958190423,
505
+ "val_full/seq_acc": 0.03695150115473441,
506
+ "system/peak_gpu_mem_gb": 1.0650625228881836,
507
+ "gen_val_win/n": 128.0,
508
+ "gen_val_win/seq_acc": 0.390625,
509
+ "gen_val_win/bit_acc": 0.9124491640307275,
510
+ "gen_val_win/lev": 3.171875,
511
+ "gen_val_win/lev_norm": 0.024217467751750245,
512
+ "gen_val_win/token_acc": 0.7240009928021842,
513
+ "gen_val_win/bleu": 0.886857168388725,
514
+ "gen_val_win/rouge1": 0.9444426307964542,
515
+ "gen_val_win/rouge2": 0.905085973956554,
516
+ "gen_val_win/rougeL": 0.9444426307964542,
517
+ "gen_val/n": 128.0,
518
+ "gen_val/seq_acc": 0.0546875,
519
+ "gen_val/bit_acc": 0.715185808048957,
520
+ "gen_val/lev": 56.625,
521
+ "gen_val/lev_norm": 0.07757158657628029,
522
+ "gen_val/token_acc": 0.2344697860333882,
523
+ "gen_val/bleu": 0.8272041845035049,
524
+ "gen_val/rouge1": 0.9173450261938174,
525
+ "gen_val/rouge2": 0.8721517697627691,
526
+ "gen_val/rougeL": 0.9165902105502579,
527
+ "gen_val_full/n": 128.0,
528
+ "gen_val_full/seq_acc": 0.0390625,
529
+ "gen_val_full/bit_acc": 0.2619298668434054,
530
+ "gen_val_full/lev": 594.4765625,
531
+ "gen_val_full/lev_norm": 0.6808404487174158,
532
+ "gen_val_full/token_acc": 0.025328769591064674,
533
+ "gen_val_full/bleu": 0.09182151966783342,
534
+ "gen_val_full/rouge1": 0.33321025973798307,
535
+ "gen_val_full/rouge2": 0.21227233105110935,
536
+ "gen_val_full/rougeL": 0.2732308977848592
537
+ },
538
+ {
539
+ "train/epoch_loss": 0.9360308318661336,
540
+ "epoch": 26,
541
+ "val/loss": 0.19933261227064888,
542
+ "val/ppl": 1.2205878808965673,
543
+ "val/token_acc": 0.9503844581184989,
544
+ "val/seq_acc": 0.2894314166398972,
545
+ "val_full/loss": 5.190170132834505,
546
+ "val_full/ppl": 179.49908902296949,
547
+ "val_full/token_acc": 0.33873038379735193,
548
+ "val_full/seq_acc": 0.04157043879907621,
549
+ "system/peak_gpu_mem_gb": 1.0650625228881836
550
+ },
551
+ {
552
+ "train/epoch_loss": 0.9254991490941987,
553
+ "epoch": 27,
554
+ "val/loss": 0.19405014994204572,
555
+ "val/ppl": 1.214157171341223,
556
+ "val/token_acc": 0.9513518365230226,
557
+ "val/seq_acc": 0.2942499196916158,
558
+ "val_full/loss": 5.019456327003775,
559
+ "val_full/ppl": 151.32900792990333,
560
+ "val_full/token_acc": 0.36066987296173525,
561
+ "val_full/seq_acc": 0.04849884526558892,
562
+ "system/peak_gpu_mem_gb": 1.0650625228881836
563
+ },
564
+ {
565
+ "train/epoch_loss": 0.9159192450325983,
566
+ "epoch": 28,
567
+ "val/loss": 0.1910588487145636,
568
+ "val/ppl": 1.2105306881601412,
569
+ "val/token_acc": 0.9533264852250402,
570
+ "val/seq_acc": 0.3135239318984902,
571
+ "val_full/loss": 5.137287221684646,
572
+ "val_full/ppl": 170.25328188214968,
573
+ "val_full/token_acc": 0.3437505987851846,
574
+ "val_full/seq_acc": 0.03695150115473441,
575
+ "system/peak_gpu_mem_gb": 1.0650625228881836
576
+ },
577
+ {
578
+ "train/epoch_loss": 0.9065227007628082,
579
+ "epoch": 29,
580
+ "val/loss": 0.18516592114836566,
581
+ "val/ppl": 1.2034180960761303,
582
+ "val/token_acc": 0.9541143501111986,
583
+ "val/seq_acc": 0.32123353678123995,
584
+ "val_full/loss": 5.164095957122191,
585
+ "val_full/ppl": 174.87928871096668,
586
+ "val_full/token_acc": 0.3512042767633026,
587
+ "val_full/seq_acc": 0.04849884526558892,
588
+ "system/peak_gpu_mem_gb": 1.0650625228881836
589
+ },
590
+ {
591
+ "train/epoch_loss": 0.8977805069854432,
592
+ "epoch": 30,
593
+ "val/loss": 0.1796077594319005,
594
+ "val/ppl": 1.1967478579886603,
595
+ "val/token_acc": 0.9550418366227523,
596
+ "val/seq_acc": 0.32605203983295855,
597
+ "val_full/loss": 5.207041279651842,
598
+ "val_full/ppl": 182.55313469340206,
599
+ "val_full/token_acc": 0.34631818965682426,
600
+ "val_full/seq_acc": 0.04849884526558892,
601
+ "system/peak_gpu_mem_gb": 1.0650625228881836,
602
+ "gen_val_win/n": 128.0,
603
+ "gen_val_win/seq_acc": 0.4765625,
604
+ "gen_val_win/bit_acc": 0.9290884067357513,
605
+ "gen_val_win/lev": 2.0859375,
606
+ "gen_val_win/lev_norm": 0.017199462753933442,
607
+ "gen_val_win/token_acc": 0.775118424333084,
608
+ "gen_val_win/bleu": 0.9065473022474428,
609
+ "gen_val_win/rouge1": 0.9534948095035979,
610
+ "gen_val_win/rouge2": 0.9190306638019985,
611
+ "gen_val_win/rougeL": 0.9534948095035979,
612
+ "gen_val/n": 128.0,
613
+ "gen_val/seq_acc": 0.0703125,
614
+ "gen_val/bit_acc": 0.7390455845506547,
615
+ "gen_val/lev": 46.875,
616
+ "gen_val/lev_norm": 0.06591076993169548,
617
+ "gen_val/token_acc": 0.27367025055652916,
618
+ "gen_val/bleu": 0.8555323033741441,
619
+ "gen_val/rouge1": 0.9286462069303709,
620
+ "gen_val/rouge2": 0.8880547235922424,
621
+ "gen_val/rougeL": 0.9281386055939489,
622
+ "gen_val_full/n": 128.0,
623
+ "gen_val_full/seq_acc": 0.03125,
624
+ "gen_val_full/bit_acc": 0.31932685358138485,
625
+ "gen_val_full/lev": 614.6484375,
626
+ "gen_val_full/lev_norm": 0.656036086833709,
627
+ "gen_val_full/token_acc": 0.02236951118475559,
628
+ "gen_val_full/bleu": 0.10616910255511497,
629
+ "gen_val_full/rouge1": 0.34030767142198753,
630
+ "gen_val_full/rouge2": 0.22211420901161488,
631
+ "gen_val_full/rougeL": 0.28067522370149545
632
+ },
633
+ {
634
+ "train/epoch_loss": 0.8909414646928745,
635
+ "epoch": 31,
636
+ "val/loss": 0.17644704366060554,
637
+ "val/ppl": 1.1929712496964127,
638
+ "val/token_acc": 0.9560491069202461,
639
+ "val/seq_acc": 0.3369739800835207,
640
+ "val_full/loss": 4.972597494203759,
641
+ "val_full/ppl": 144.40148266999464,
642
+ "val_full/token_acc": 0.3619153461457395,
643
+ "val_full/seq_acc": 0.04157043879907621,
644
+ "system/peak_gpu_mem_gb": 1.0650625228881836
645
+ },
646
+ {
647
+ "train/epoch_loss": 0.8828385904840104,
648
+ "epoch": 32,
649
+ "val/loss": 0.17426898212655229,
650
+ "val/ppl": 1.1903757125521166,
651
+ "val/token_acc": 0.9571461339769225,
652
+ "val/seq_acc": 0.3472534532605204,
653
+ "val_full/loss": 5.272345390671406,
654
+ "val_full/ppl": 194.87247897562096,
655
+ "val_full/token_acc": 0.33982256797409416,
656
+ "val_full/seq_acc": 0.04849884526558892,
657
+ "system/peak_gpu_mem_gb": 1.0650625228881836
658
+ },
659
+ {
660
+ "train/epoch_loss": 0.8764852108503518,
661
+ "epoch": 33,
662
+ "val/loss": 0.1680034250126493,
663
+ "val/ppl": 1.182940662227604,
664
+ "val/token_acc": 0.9589512421338173,
665
+ "val/seq_acc": 0.36684869900417605,
666
+ "val_full/loss": 5.243562909569066,
667
+ "val_full/ppl": 189.3435158621706,
668
+ "val_full/token_acc": 0.342294353216195,
669
+ "val_full/seq_acc": 0.053117782909930716,
670
+ "system/peak_gpu_mem_gb": 1.0650625228881836
671
+ },
672
+ {
673
+ "train/epoch_loss": 0.8692677410463443,
674
+ "epoch": 34,
675
+ "val/loss": 0.1662353731105501,
676
+ "val/ppl": 1.180851009591197,
677
+ "val/token_acc": 0.9597391070199759,
678
+ "val/seq_acc": 0.3710247349823322,
679
+ "val_full/loss": 5.125430601395888,
680
+ "val_full/ppl": 168.24657326545085,
681
+ "val_full/token_acc": 0.35971181666634733,
682
+ "val_full/seq_acc": 0.050808314087759814,
683
+ "system/peak_gpu_mem_gb": 1.0650625228881836
684
+ },
685
+ {
686
+ "train/epoch_loss": 0.8641646281442142,
687
+ "epoch": 35,
688
+ "val/loss": 0.16242443123179154,
689
+ "val/ppl": 1.1763594190574553,
690
+ "val/token_acc": 0.9608361340766522,
691
+ "val/seq_acc": 0.3787343398650819,
692
+ "val_full/loss": 5.187735621970147,
693
+ "val_full/ppl": 179.06262804067197,
694
+ "val_full/token_acc": 0.3523731054436759,
695
+ "val_full/seq_acc": 0.050808314087759814,
696
+ "system/peak_gpu_mem_gb": 1.0650625228881836,
697
+ "gen_val_win/n": 128.0,
698
+ "gen_val_win/seq_acc": 0.5078125,
699
+ "gen_val_win/bit_acc": 0.934458933437135,
700
+ "gen_val_win/lev": 1.515625,
701
+ "gen_val_win/lev_norm": 0.012608255240775584,
702
+ "gen_val_win/token_acc": 0.7862842892768079,
703
+ "gen_val_win/bleu": 0.9190175899136278,
704
+ "gen_val_win/rouge1": 0.9629300041650583,
705
+ "gen_val_win/rouge2": 0.9325652026415067,
706
+ "gen_val_win/rougeL": 0.9629300041650583,
707
+ "gen_val/n": 128.0,
708
+ "gen_val/seq_acc": 0.09375,
709
+ "gen_val/bit_acc": 0.758615883558934,
710
+ "gen_val/lev": 36.1484375,
711
+ "gen_val/lev_norm": 0.04977548676594661,
712
+ "gen_val/token_acc": 0.2996242050491424,
713
+ "gen_val/bleu": 0.8759583484058917,
714
+ "gen_val/rouge1": 0.9427869829855836,
715
+ "gen_val/rouge2": 0.9081305703292875,
716
+ "gen_val/rougeL": 0.9425731720874744,
717
+ "gen_val_full/n": 128.0,
718
+ "gen_val_full/seq_acc": 0.0546875,
719
+ "gen_val_full/bit_acc": 0.2445423834667575,
720
+ "gen_val_full/lev": 529.265625,
721
+ "gen_val_full/lev_norm": 0.6818199514553388,
722
+ "gen_val_full/token_acc": 0.033334754252099405,
723
+ "gen_val_full/bleu": 0.06715355758530663,
724
+ "gen_val_full/rouge1": 0.33110716118594996,
725
+ "gen_val_full/rouge2": 0.23193709834381152,
726
+ "gen_val_full/rougeL": 0.28028347532296566
727
+ },
728
+ {
729
+ "train/epoch_loss": 0.8560918325795199,
730
+ "epoch": 36,
731
+ "val/loss": 0.1609244988561467,
732
+ "val/ppl": 1.174596282103075,
733
+ "val/token_acc": 0.9611752151668977,
734
+ "val/seq_acc": 0.384195309990363,
735
+ "val_full/loss": 5.36468143430608,
736
+ "val_full/ppl": 213.72313897496375,
737
+ "val_full/token_acc": 0.34240931997164153,
738
+ "val_full/seq_acc": 0.050808314087759814,
739
+ "system/peak_gpu_mem_gb": 1.0650625228881836
740
+ },
741
+ {
742
+ "train/epoch_loss": 0.8503464565907334,
743
+ "epoch": 37,
744
+ "val/loss": 0.15771310727542617,
745
+ "val/ppl": 1.170830243840375,
746
+ "val/token_acc": 0.9625714314208494,
747
+ "val/seq_acc": 0.39029874718920654,
748
+ "val_full/loss": 5.132414195580007,
749
+ "val_full/ppl": 169.42565136546253,
750
+ "val_full/token_acc": 0.35858131023778955,
751
+ "val_full/seq_acc": 0.053117782909930716,
752
+ "system/peak_gpu_mem_gb": 1.0650625228881836
753
+ },
754
+ {
755
+ "train/epoch_loss": 0.8460360304673117,
756
+ "epoch": 38,
757
+ "val/loss": 0.15577538608773545,
758
+ "val/ppl": 1.168563697945893,
759
+ "val/token_acc": 0.9631997287351278,
760
+ "val/seq_acc": 0.4050754898811436,
761
+ "val_full/loss": 5.050719949366725,
762
+ "val_full/ppl": 156.1348332059374,
763
+ "val_full/token_acc": 0.3668589166299412,
764
+ "val_full/seq_acc": 0.053117782909930716,
765
+ "system/peak_gpu_mem_gb": 1.0650625228881836
766
+ },
767
+ {
768
+ "train/epoch_loss": 0.8416729933603149,
769
+ "epoch": 39,
770
+ "val/loss": 0.15192259107822684,
771
+ "val/ppl": 1.1640701235312552,
772
+ "val/token_acc": 0.964137188219924,
773
+ "val/seq_acc": 0.4124638612271121,
774
+ "val_full/loss": 5.31732957675468,
775
+ "val_full/ppl": 203.838818622384,
776
+ "val_full/token_acc": 0.3524114276954914,
777
+ "val_full/seq_acc": 0.057736720554272515,
778
+ "system/peak_gpu_mem_gb": 1.0650625228881836
779
+ },
780
+ {
781
+ "train/epoch_loss": 0.8355102124356867,
782
+ "epoch": 40,
783
+ "val/loss": 0.14684852000297943,
784
+ "val/ppl": 1.1581785088642205,
785
+ "val/token_acc": 0.965712917992241,
786
+ "val/seq_acc": 0.42788307099261164,
787
+ "val_full/loss": 5.166461495418096,
788
+ "val_full/ppl": 175.29346204387824,
789
+ "val_full/token_acc": 0.3586196324896051,
790
+ "val_full/seq_acc": 0.053117782909930716,
791
+ "system/peak_gpu_mem_gb": 1.0650625228881836,
792
+ "gen_val_win/n": 128.0,
793
+ "gen_val_win/seq_acc": 0.5703125,
794
+ "gen_val_win/bit_acc": 0.9493728472086826,
795
+ "gen_val_win/lev": 1.4453125,
796
+ "gen_val_win/lev_norm": 0.0121548993457164,
797
+ "gen_val_win/token_acc": 0.8024937655860349,
798
+ "gen_val_win/bleu": 0.9298298073616039,
799
+ "gen_val_win/rouge1": 0.9655077197660651,
800
+ "gen_val_win/rouge2": 0.9393315916832448,
801
+ "gen_val_win/rougeL": 0.9655077197660651,
802
+ "gen_val/n": 128.0,
803
+ "gen_val/seq_acc": 0.078125,
804
+ "gen_val/bit_acc": 0.7715689040421624,
805
+ "gen_val/lev": 33.25,
806
+ "gen_val/lev_norm": 0.04768892438456196,
807
+ "gen_val/token_acc": 0.3463888755232642,
808
+ "gen_val/bleu": 0.8883294517665103,
809
+ "gen_val/rouge1": 0.9483271124692857,
810
+ "gen_val/rouge2": 0.9168530717913945,
811
+ "gen_val/rougeL": 0.9479645382085561,
812
+ "gen_val_full/n": 128.0,
813
+ "gen_val_full/seq_acc": 0.03125,
814
+ "gen_val_full/bit_acc": 0.32233132791388186,
815
+ "gen_val_full/lev": 518.3203125,
816
+ "gen_val_full/lev_norm": 0.6488054957973685,
817
+ "gen_val_full/token_acc": 0.030329453690715397,
818
+ "gen_val_full/bleu": 0.0938472377889516,
819
+ "gen_val_full/rouge1": 0.3601287541890134,
820
+ "gen_val_full/rouge2": 0.24646846885896284,
821
+ "gen_val_full/rougeL": 0.3034602996508524
822
+ },
823
+ {
824
+ "train/epoch_loss": 0.8304888334060251,
825
+ "epoch": 41,
826
+ "val/loss": 0.14268469852096327,
827
+ "val/ppl": 1.1533660862961466,
828
+ "val/token_acc": 0.9666304315305522,
829
+ "val/seq_acc": 0.4416961130742049,
830
+ "val_full/loss": 5.358802112274617,
831
+ "val_full/ppl": 212.47027840962429,
832
+ "val_full/token_acc": 0.35145337140010346,
833
+ "val_full/seq_acc": 0.06004618937644342,
834
+ "system/peak_gpu_mem_gb": 1.0650625228881836
835
+ },
836
+ {
837
+ "train/epoch_loss": 0.826247526523181,
838
+ "epoch": 42,
839
+ "val/loss": 0.13970463727989532,
840
+ "val/ppl": 1.1499341010286386,
841
+ "val/token_acc": 0.9673385126307706,
842
+ "val/seq_acc": 0.4500481850305172,
843
+ "val_full/loss": 5.450080551576961,
844
+ "val_full/ppl": 232.77691570012897,
845
+ "val_full/token_acc": 0.3342466803349365,
846
+ "val_full/seq_acc": 0.057736720554272515,
847
+ "system/peak_gpu_mem_gb": 1.0650625228881836
848
+ },
849
+ {
850
+ "train/epoch_loss": 0.8227115215506042,
851
+ "epoch": 43,
852
+ "val/loss": 0.1390314311509267,
853
+ "val/ppl": 1.1491602188642516,
854
+ "val/token_acc": 0.9682261072493542,
855
+ "val/seq_acc": 0.4465146161259235,
856
+ "val_full/loss": 5.193236661461228,
857
+ "val_full/ppl": 180.0503729498964,
858
+ "val_full/token_acc": 0.36227940753798693,
859
+ "val_full/seq_acc": 0.06466512702078522,
860
+ "system/peak_gpu_mem_gb": 1.0650625228881836
861
+ },
862
+ {
863
+ "train/epoch_loss": 0.8173399352076047,
864
+ "epoch": 44,
865
+ "val/loss": 0.1356666351058748,
866
+ "val/ppl": 1.1453000271256764,
867
+ "val/token_acc": 0.9691137018679379,
868
+ "val/seq_acc": 0.46514616125923547,
869
+ "val_full/loss": 5.500843089539941,
870
+ "val_full/ppl": 244.89831646077423,
871
+ "val_full/token_acc": 0.3399375347295407,
872
+ "val_full/seq_acc": 0.05542725173210162,
873
+ "system/peak_gpu_mem_gb": 1.0650625228881836
874
+ },
875
+ {
876
+ "train/epoch_loss": 0.8136944240168146,
877
+ "epoch": 45,
878
+ "val/loss": 0.1337494416315866,
879
+ "val/ppl": 1.1431063688931162,
880
+ "val/token_acc": 0.969562485663851,
881
+ "val/seq_acc": 0.46900096370061034,
882
+ "val_full/loss": 5.126671359146563,
883
+ "val_full/ppl": 168.4554560649331,
884
+ "val_full/token_acc": 0.3674145892812662,
885
+ "val_full/seq_acc": 0.06235565819861432,
886
+ "system/peak_gpu_mem_gb": 1.0650625228881836,
887
+ "gen_val_win/n": 128.0,
888
+ "gen_val_win/seq_acc": 0.5859375,
889
+ "gen_val_win/bit_acc": 0.9471821841319309,
890
+ "gen_val_win/lev": 1.25,
891
+ "gen_val_win/lev_norm": 0.010280795885099246,
892
+ "gen_val_win/token_acc": 0.8131539611360239,
893
+ "gen_val_win/bleu": 0.9391197762238938,
894
+ "gen_val_win/rouge1": 0.9711827774040873,
895
+ "gen_val_win/rouge2": 0.9486937066297777,
896
+ "gen_val_win/rougeL": 0.9711827774040873,
897
+ "gen_val/n": 128.0,
898
+ "gen_val/seq_acc": 0.1015625,
899
+ "gen_val/bit_acc": 0.7821110627283063,
900
+ "gen_val/lev": 28.25,
901
+ "gen_val/lev_norm": 0.03832984563677537,
902
+ "gen_val/token_acc": 0.351591737545565,
903
+ "gen_val/bleu": 0.9043745164371645,
904
+ "gen_val/rouge1": 0.9555825978731163,
905
+ "gen_val/rouge2": 0.9295113795421714,
906
+ "gen_val/rougeL": 0.9555155377872794,
907
+ "gen_val_full/n": 128.0,
908
+ "gen_val_full/seq_acc": 0.0546875,
909
+ "gen_val_full/bit_acc": 0.3742634223616899,
910
+ "gen_val_full/lev": 552.1796875,
911
+ "gen_val_full/lev_norm": 0.6319002319393148,
912
+ "gen_val_full/token_acc": 0.033000545728869055,
913
+ "gen_val_full/bleu": 0.12726190921975344,
914
+ "gen_val_full/rouge1": 0.3730387399774349,
915
+ "gen_val_full/rouge2": 0.25131445966374627,
916
+ "gen_val_full/rougeL": 0.30975626539946866
917
+ },
918
+ {
919
+ "train/epoch_loss": 0.8100172247078056,
920
+ "epoch": 46,
921
+ "val/loss": 0.13167286623952215,
922
+ "val/ppl": 1.1407350852636906,
923
+ "val/token_acc": 0.9703702964964945,
924
+ "val/seq_acc": 0.4779955027304851,
925
+ "val_full/loss": 5.344028135718255,
926
+ "val_full/ppl": 209.35432170964398,
927
+ "val_full/token_acc": 0.3525263944509379,
928
+ "val_full/seq_acc": 0.06928406466512702,
929
+ "system/peak_gpu_mem_gb": 1.0650625228881836
930
+ },
931
+ {
932
+ "train/epoch_loss": 0.8063960959786489,
933
+ "epoch": 47,
934
+ "val/loss": 0.1294364190767629,
935
+ "val/ppl": 1.1381867421982717,
936
+ "val/token_acc": 0.9712479181418356,
937
+ "val/seq_acc": 0.4815290716350787,
938
+ "val_full/loss": 5.178985873459925,
939
+ "val_full/ppl": 177.50270947732005,
940
+ "val_full/token_acc": 0.36779781179942134,
941
+ "val_full/seq_acc": 0.06928406466512702,
942
+ "system/peak_gpu_mem_gb": 1.0650625228881836
943
+ },
944
+ {
945
+ "train/epoch_loss": 0.8027049875318856,
946
+ "epoch": 48,
947
+ "val/loss": 0.12705491853648876,
948
+ "val/ppl": 1.1354793749346737,
949
+ "val/token_acc": 0.9717066749109912,
950
+ "val/seq_acc": 0.48827497590748475,
951
+ "val_full/loss": 5.290455498883865,
952
+ "val_full/ppl": 198.43379119747243,
953
+ "val_full/token_acc": 0.36042077832493435,
954
+ "val_full/seq_acc": 0.06466512702078522,
955
+ "system/peak_gpu_mem_gb": 1.0650625228881836
956
+ },
957
+ {
958
+ "train/epoch_loss": 0.7997443438170854,
959
+ "epoch": 49,
960
+ "val/loss": 0.12609530869098648,
961
+ "val/ppl": 1.134390280383623,
962
+ "val/token_acc": 0.9716368640982936,
963
+ "val/seq_acc": 0.49020237712817216,
964
+ "val_full/loss": 5.219981027491426,
965
+ "val_full/ppl": 184.93067543857435,
966
+ "val_full/token_acc": 0.36523022092778173,
967
+ "val_full/seq_acc": 0.06928406466512702,
968
+ "system/peak_gpu_mem_gb": 1.0650625228881836
969
+ },
970
+ {
971
+ "train/epoch_loss": 0.7962049421229564,
972
+ "epoch": 50,
973
+ "val/loss": 0.12423698263302155,
974
+ "val/ppl": 1.1322841708908584,
975
+ "val/token_acc": 0.9724347019576947,
976
+ "val/seq_acc": 0.5024092515258592,
977
+ "val_full/loss": 5.256799355467627,
978
+ "val_full/ppl": 191.8664113158115,
979
+ "val_full/token_acc": 0.36300753032248173,
980
+ "val_full/seq_acc": 0.06928406466512702,
981
+ "system/peak_gpu_mem_gb": 1.0650625228881836,
982
+ "gen_val_win/n": 128.0,
983
+ "gen_val_win/seq_acc": 0.546875,
984
+ "gen_val_win/bit_acc": 0.9418925947597685,
985
+ "gen_val_win/lev": 1.2734375,
986
+ "gen_val_win/lev_norm": 0.010597328446190442,
987
+ "gen_val_win/token_acc": 0.7919680718383637,
988
+ "gen_val_win/bleu": 0.9346683056433839,
989
+ "gen_val_win/rouge1": 0.9676755114021661,
990
+ "gen_val_win/rouge2": 0.9431690249745048,
991
+ "gen_val_win/rougeL": 0.9676755114021661,
992
+ "gen_val/n": 128.0,
993
+ "gen_val/seq_acc": 0.1015625,
994
+ "gen_val/bit_acc": 0.7954056396765551,
995
+ "gen_val/lev": 25.1796875,
996
+ "gen_val/lev_norm": 0.03469096867452959,
997
+ "gen_val/token_acc": 0.4104618971295352,
998
+ "gen_val/bleu": 0.9148643251858689,
999
+ "gen_val/rouge1": 0.9612759444173142,
1000
+ "gen_val/rouge2": 0.9383258717336574,
1001
+ "gen_val/rougeL": 0.9611566745210505,
1002
+ "gen_val_full/n": 128.0,
1003
+ "gen_val_full/seq_acc": 0.0546875,
1004
+ "gen_val_full/bit_acc": 0.3792326461507818,
1005
+ "gen_val_full/lev": 568.140625,
1006
+ "gen_val_full/lev_norm": 0.6416667424147317,
1007
+ "gen_val_full/token_acc": 0.03232423696152172,
1008
+ "gen_val_full/bleu": 0.12455500585887742,
1009
+ "gen_val_full/rouge1": 0.36369442228980675,
1010
+ "gen_val_full/rouge2": 0.2485710655072301,
1011
+ "gen_val_full/rougeL": 0.29943750735353
1012
+ },
1013
+ {
1014
+ "train/epoch_loss": 0.7935096948521393,
1015
+ "epoch": 51,
1016
+ "val/loss": 0.12211543709954227,
1017
+ "val/ppl": 1.1298845248447096,
1018
+ "val/token_acc": 0.973102891164943,
1019
+ "val/seq_acc": 0.5101188564086091,
1020
+ "val_full/loss": 5.14476754560348,
1021
+ "val_full/ppl": 171.53160676021028,
1022
+ "val_full/token_acc": 0.37293299354270054,
1023
+ "val_full/seq_acc": 0.07390300230946882,
1024
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1025
+ },
1026
+ {
1027
+ "train/epoch_loss": 0.7909734631119821,
1028
+ "epoch": 52,
1029
+ "val/loss": 0.11956716591250602,
1030
+ "val/ppl": 1.1270089381088781,
1031
+ "val/token_acc": 0.9738109722651614,
1032
+ "val/seq_acc": 0.5165435271442339,
1033
+ "val_full/loss": 5.223838476498879,
1034
+ "val_full/ppl": 185.6454137359905,
1035
+ "val_full/token_acc": 0.3640039088696852,
1036
+ "val_full/seq_acc": 0.07159353348729793,
1037
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1038
+ },
1039
+ {
1040
+ "train/epoch_loss": 0.7875587506187229,
1041
+ "epoch": 53,
1042
+ "val/loss": 0.12078814995346984,
1043
+ "val/ppl": 1.128385838451948,
1044
+ "val/token_acc": 0.9736613776665237,
1045
+ "val/seq_acc": 0.512688724702859,
1046
+ "val_full/loss": 5.316853243140317,
1047
+ "val_full/ppl": 203.7417464623638,
1048
+ "val_full/token_acc": 0.3535419341240491,
1049
+ "val_full/seq_acc": 0.07390300230946882,
1050
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1051
+ },
1052
+ {
1053
+ "train/epoch_loss": 0.7846282487200977,
1054
+ "epoch": 54,
1055
+ "val/loss": 0.11707850749736769,
1056
+ "val/ppl": 1.124207684958009,
1057
+ "val/token_acc": 0.9744691884991673,
1058
+ "val/seq_acc": 0.52071956312239,
1059
+ "val_full/loss": 5.237640699548756,
1060
+ "val_full/ppl": 188.22549763227875,
1061
+ "val_full/token_acc": 0.36473203165418,
1062
+ "val_full/seq_acc": 0.07159353348729793,
1063
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1064
+ },
1065
+ {
1066
+ "train/epoch_loss": 0.7828222343749239,
1067
+ "epoch": 55,
1068
+ "val/loss": 0.11614146314732193,
1069
+ "val/ppl": 1.1231547459008828,
1070
+ "val/token_acc": 0.9748381885091403,
1071
+ "val/seq_acc": 0.5300353356890459,
1072
+ "val_full/loss": 5.197993672936347,
1073
+ "val_full/ppl": 180.908915068467,
1074
+ "val_full/token_acc": 0.3668972388817567,
1075
+ "val_full/seq_acc": 0.07852193995381063,
1076
+ "system/peak_gpu_mem_gb": 1.0650625228881836,
1077
+ "gen_val_win/n": 128.0,
1078
+ "gen_val_win/seq_acc": 0.5234375,
1079
+ "gen_val_win/bit_acc": 0.9366707856399584,
1080
+ "gen_val_win/lev": 1.4296875,
1081
+ "gen_val_win/lev_norm": 0.01176483948693388,
1082
+ "gen_val_win/token_acc": 0.799800299550674,
1083
+ "gen_val_win/bleu": 0.9290983820929968,
1084
+ "gen_val_win/rouge1": 0.965755473669198,
1085
+ "gen_val_win/rouge2": 0.9411171707758922,
1086
+ "gen_val_win/rougeL": 0.965755473669198,
1087
+ "gen_val/n": 128.0,
1088
+ "gen_val/seq_acc": 0.1640625,
1089
+ "gen_val/bit_acc": 0.8011712959587859,
1090
+ "gen_val/lev": 24.7265625,
1091
+ "gen_val/lev_norm": 0.03567827127089045,
1092
+ "gen_val/token_acc": 0.39545210559703314,
1093
+ "gen_val/bleu": 0.9164283567075726,
1094
+ "gen_val/rouge1": 0.9624503069182171,
1095
+ "gen_val/rouge2": 0.9409586148036581,
1096
+ "gen_val/rougeL": 0.9624026697230952,
1097
+ "gen_val_full/n": 128.0,
1098
+ "gen_val_full/seq_acc": 0.0625,
1099
+ "gen_val_full/bit_acc": 0.3757417117538352,
1100
+ "gen_val_full/lev": 500.8515625,
1101
+ "gen_val_full/lev_norm": 0.6181358328513049,
1102
+ "gen_val_full/token_acc": 0.03886472224203095,
1103
+ "gen_val_full/bleu": 0.11915294477701922,
1104
+ "gen_val_full/rouge1": 0.39043593908446095,
1105
+ "gen_val_full/rouge2": 0.2787649988337111,
1106
+ "gen_val_full/rougeL": 0.3261082252920391
1107
+ },
1108
+ {
1109
+ "train/epoch_loss": 0.7798933871369113,
1110
+ "epoch": 56,
1111
+ "val/loss": 0.11468604560401079,
1112
+ "val/ppl": 1.1215212757587365,
1113
+ "val/token_acc": 0.9749079993218378,
1114
+ "val/seq_acc": 0.5326052039832958,
1115
+ "val_full/loss": 5.27014859752534,
1116
+ "val_full/ppl": 194.4448543228764,
1117
+ "val_full/token_acc": 0.3668589166299412,
1118
+ "val_full/seq_acc": 0.07852193995381063,
1119
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1120
+ },
1121
+ {
1122
+ "train/epoch_loss": 0.7770069360435752,
1123
+ "epoch": 57,
1124
+ "val/loss": 0.11484508718129241,
1125
+ "val/ppl": 1.1216996584561447,
1126
+ "val/token_acc": 0.9753468101445084,
1127
+ "val/seq_acc": 0.5374237070350144,
1128
+ "val_full/loss": 5.390086617270881,
1129
+ "val_full/ppl": 219.22237317529513,
1130
+ "val_full/token_acc": 0.3521431719327828,
1131
+ "val_full/seq_acc": 0.08083140877598152,
1132
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1133
+ },
1134
+ {
1135
+ "train/epoch_loss": 0.7758665425224495,
1136
+ "epoch": 58,
1137
+ "val/loss": 0.11302239058068163,
1138
+ "val/ppl": 1.1196570024383736,
1139
+ "val/token_acc": 0.975516350689631,
1140
+ "val/seq_acc": 0.5415997430131706,
1141
+ "val_full/loss": 5.211792827212152,
1142
+ "val_full/ppl": 183.42260863271198,
1143
+ "val_full/token_acc": 0.36733794477763515,
1144
+ "val_full/seq_acc": 0.07852193995381063,
1145
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1146
+ },
1147
+ {
1148
+ "train/epoch_loss": 0.7734360559622843,
1149
+ "epoch": 59,
1150
+ "val/loss": 0.11200202885361335,
1151
+ "val/ppl": 1.1185151299462062,
1152
+ "val/token_acc": 0.9760449182714843,
1153
+ "val/seq_acc": 0.5470607131384516,
1154
+ "val_full/loss": 5.517750911949836,
1155
+ "val_full/ppl": 249.0742169141425,
1156
+ "val_full/token_acc": 0.3381938722719347,
1157
+ "val_full/seq_acc": 0.06928406466512702,
1158
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1159
+ },
1160
+ {
1161
+ "train/epoch_loss": 0.7710748243510277,
1162
+ "epoch": 60,
1163
+ "val/loss": 0.11054647731977883,
1164
+ "val/ppl": 1.1168882578185346,
1165
+ "val/token_acc": 0.9764238912546997,
1166
+ "val/seq_acc": 0.5566977192418888,
1167
+ "val_full/loss": 5.328467205733009,
1168
+ "val_full/ppl": 206.12178959787667,
1169
+ "val_full/token_acc": 0.3525838778286612,
1170
+ "val_full/seq_acc": 0.06928406466512702,
1171
+ "system/peak_gpu_mem_gb": 1.0650625228881836,
1172
+ "gen_val_win/n": 128.0,
1173
+ "gen_val_win/seq_acc": 0.6015625,
1174
+ "gen_val_win/bit_acc": 0.9388405147576669,
1175
+ "gen_val_win/lev": 3.1015625,
1176
+ "gen_val_win/lev_norm": 0.015023159623174206,
1177
+ "gen_val_win/token_acc": 0.8217016029593095,
1178
+ "gen_val_win/bleu": 0.9252322601744094,
1179
+ "gen_val_win/rouge1": 0.9687495089006884,
1180
+ "gen_val_win/rouge2": 0.9471590340864586,
1181
+ "gen_val_win/rougeL": 0.9687495089006884,
1182
+ "gen_val/n": 128.0,
1183
+ "gen_val/seq_acc": 0.171875,
1184
+ "gen_val/bit_acc": 0.787561823285657,
1185
+ "gen_val/lev": 24.1015625,
1186
+ "gen_val/lev_norm": 0.0344065360000764,
1187
+ "gen_val/token_acc": 0.4139342265529842,
1188
+ "gen_val/bleu": 0.9224372884442231,
1189
+ "gen_val/rouge1": 0.9641586755754376,
1190
+ "gen_val/rouge2": 0.9443289078504787,
1191
+ "gen_val/rougeL": 0.9641255014777731,
1192
+ "gen_val_full/n": 128.0,
1193
+ "gen_val_full/seq_acc": 0.0546875,
1194
+ "gen_val_full/bit_acc": 0.3918829288835079,
1195
+ "gen_val_full/lev": 525.8671875,
1196
+ "gen_val_full/lev_norm": 0.6262776687851552,
1197
+ "gen_val_full/token_acc": 0.03767860638089646,
1198
+ "gen_val_full/bleu": 0.12005813143076255,
1199
+ "gen_val_full/rouge1": 0.3709799318307528,
1200
+ "gen_val_full/rouge2": 0.25285862530257186,
1201
+ "gen_val_full/rougeL": 0.3057391487110338
1202
+ },
1203
+ {
1204
+ "train/epoch_loss": 0.7692888557465,
1205
+ "epoch": 61,
1206
+ "val/loss": 0.10926879314589201,
1207
+ "val/ppl": 1.1154621386264896,
1208
+ "val/token_acc": 0.9764937020673974,
1209
+ "val/seq_acc": 0.5544490844844202,
1210
+ "val_full/loss": 5.5136764781611065,
1211
+ "val_full/ppl": 248.06144514568888,
1212
+ "val_full/token_acc": 0.34595412826457683,
1213
+ "val_full/seq_acc": 0.08083140877598152,
1214
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1215
+ },
1216
+ {
1217
+ "train/epoch_loss": 0.767349373967273,
1218
+ "epoch": 62,
1219
+ "val/loss": 0.10924034142619984,
1220
+ "val/ppl": 1.1154304022618733,
1221
+ "val/token_acc": 0.9767330534252177,
1222
+ "val/seq_acc": 0.5560552521683264,
1223
+ "val_full/loss": 5.578898181369637,
1224
+ "val_full/ppl": 264.7797057925477,
1225
+ "val_full/token_acc": 0.3364502098143287,
1226
+ "val_full/seq_acc": 0.07852193995381063,
1227
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1228
+ },
1229
+ {
1230
+ "train/epoch_loss": 0.7658214072634157,
1231
+ "epoch": 63,
1232
+ "val/loss": 0.10810739653960619,
1233
+ "val/ppl": 1.114167396684231,
1234
+ "val/token_acc": 0.9771020534351906,
1235
+ "val/seq_acc": 0.5650497911982011,
1236
+ "val_full/loss": 5.402696284777443,
1237
+ "val_full/ppl": 222.00419648460908,
1238
+ "val_full/token_acc": 0.34903906953572594,
1239
+ "val_full/seq_acc": 0.08314087759815242,
1240
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1241
+ },
1242
+ {
1243
+ "train/epoch_loss": 0.7634172452952797,
1244
+ "epoch": 64,
1245
+ "val/loss": 0.10817990131509235,
1246
+ "val/ppl": 1.1142481820698091,
1247
+ "val/token_acc": 0.9772516480338284,
1248
+ "val/seq_acc": 0.5611949887568262,
1249
+ "val_full/loss": 5.360503036439671,
1250
+ "val_full/ppl": 212.83198176831377,
1251
+ "val_full/token_acc": 0.35151085477782673,
1252
+ "val_full/seq_acc": 0.08083140877598152,
1253
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1254
+ },
1255
+ {
1256
+ "train/epoch_loss": 0.7615973996997177,
1257
+ "epoch": 65,
1258
+ "val/loss": 0.1061087566860069,
1259
+ "val/ppl": 1.111942801144812,
1260
+ "val/token_acc": 0.9775907291240737,
1261
+ "val/seq_acc": 0.5660134918085448,
1262
+ "val_full/loss": 5.4703817644762305,
1263
+ "val_full/ppl": 237.55086393369695,
1264
+ "val_full/token_acc": 0.3476594684703673,
1265
+ "val_full/seq_acc": 0.08545034642032333,
1266
+ "system/peak_gpu_mem_gb": 1.0650625228881836,
1267
+ "gen_val_win/n": 128.0,
1268
+ "gen_val_win/seq_acc": 0.625,
1269
+ "gen_val_win/bit_acc": 0.9240180500826236,
1270
+ "gen_val_win/lev": 3.984375,
1271
+ "gen_val_win/lev_norm": 0.016169124204950366,
1272
+ "gen_val_win/token_acc": 0.8010792249202845,
1273
+ "gen_val_win/bleu": 0.9230816637610787,
1274
+ "gen_val_win/rouge1": 0.9682264551272864,
1275
+ "gen_val_win/rouge2": 0.9492211007789696,
1276
+ "gen_val_win/rougeL": 0.9682264551272864,
1277
+ "gen_val/n": 128.0,
1278
+ "gen_val/seq_acc": 0.203125,
1279
+ "gen_val/bit_acc": 0.8057363574600247,
1280
+ "gen_val/lev": 24.1953125,
1281
+ "gen_val/lev_norm": 0.03162261251077031,
1282
+ "gen_val/token_acc": 0.4415028226591396,
1283
+ "gen_val/bleu": 0.9247399313901736,
1284
+ "gen_val/rouge1": 0.9666540727733139,
1285
+ "gen_val/rouge2": 0.9480247759222942,
1286
+ "gen_val/rougeL": 0.9666056981603108,
1287
+ "gen_val_full/n": 128.0,
1288
+ "gen_val_full/seq_acc": 0.0625,
1289
+ "gen_val_full/bit_acc": 0.34558169266172334,
1290
+ "gen_val_full/lev": 556.265625,
1291
+ "gen_val_full/lev_norm": 0.6461050540429291,
1292
+ "gen_val_full/token_acc": 0.035525197725341524,
1293
+ "gen_val_full/bleu": 0.1176912809591565,
1294
+ "gen_val_full/rouge1": 0.35404522179498393,
1295
+ "gen_val_full/rouge2": 0.24455123647055177,
1296
+ "gen_val_full/rougeL": 0.29453265957843894
1297
+ },
1298
+ {
1299
+ "train/epoch_loss": 0.7608585213485204,
1300
+ "epoch": 66,
1301
+ "val/loss": 0.1095111308717146,
1302
+ "val/ppl": 1.1157324899412437,
1303
+ "val/token_acc": 0.9770920804619482,
1304
+ "val/seq_acc": 0.5634436235142949,
1305
+ "val_full/loss": 5.420797560908429,
1306
+ "val_full/ppl": 226.0593467205135,
1307
+ "val_full/token_acc": 0.347448696085382,
1308
+ "val_full/seq_acc": 0.07390300230946882,
1309
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1310
+ },
1311
+ {
1312
+ "train/epoch_loss": 0.7589125273530918,
1313
+ "epoch": 67,
1314
+ "val/loss": 0.10670304391846144,
1315
+ "val/ppl": 1.1126038109501415,
1316
+ "val/token_acc": 0.9779497561608043,
1317
+ "val/seq_acc": 0.5737230966912946,
1318
+ "val_full/loss": 5.493937000613156,
1319
+ "val_full/ppl": 243.21285358639852,
1320
+ "val_full/token_acc": 0.3474870183371975,
1321
+ "val_full/seq_acc": 0.07852193995381063,
1322
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1323
+ },
1324
+ {
1325
+ "train/epoch_loss": 0.7570422626195703,
1326
+ "epoch": 68,
1327
+ "val/loss": 0.10556327852618037,
1328
+ "val/ppl": 1.1113364260290377,
1329
+ "val/token_acc": 0.9778899183213492,
1330
+ "val/seq_acc": 0.5692258271763572,
1331
+ "val_full/loss": 5.432119765419916,
1332
+ "val_full/ppl": 228.6333812509411,
1333
+ "val_full/token_acc": 0.34719960144858114,
1334
+ "val_full/seq_acc": 0.07159353348729793,
1335
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1336
+ },
1337
+ {
1338
+ "train/epoch_loss": 0.7554686881062991,
1339
+ "epoch": 69,
1340
+ "val/loss": 0.10390940998502457,
1341
+ "val/ppl": 1.1094999407466137,
1342
+ "val/token_acc": 0.9779397831875617,
1343
+ "val/seq_acc": 0.5753292643752008,
1344
+ "val_full/loss": 5.518696318908199,
1345
+ "val_full/ppl": 249.3098047576079,
1346
+ "val_full/token_acc": 0.34212190308302515,
1347
+ "val_full/seq_acc": 0.07621247113163972,
1348
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1349
+ },
1350
+ {
1351
+ "train/epoch_loss": 0.7545096016584192,
1352
+ "epoch": 70,
1353
+ "val/loss": 0.10571780794785876,
1354
+ "val/ppl": 1.111508173473917,
1355
+ "val/token_acc": 0.977850026428379,
1356
+ "val/seq_acc": 0.5685833601027948,
1357
+ "val_full/loss": 5.426737060850946,
1358
+ "val_full/ppl": 227.40602152587084,
1359
+ "val_full/token_acc": 0.3519707217996129,
1360
+ "val_full/seq_acc": 0.07159353348729793,
1361
+ "system/peak_gpu_mem_gb": 1.0650625228881836,
1362
+ "gen_val_win/n": 128.0,
1363
+ "gen_val_win/seq_acc": 0.6171875,
1364
+ "gen_val_win/bit_acc": 0.9549847074900761,
1365
+ "gen_val_win/lev": 1.109375,
1366
+ "gen_val_win/lev_norm": 0.009229312465825404,
1367
+ "gen_val_win/token_acc": 0.8390316945345645,
1368
+ "gen_val_win/bleu": 0.9452998286784585,
1369
+ "gen_val_win/rouge1": 0.9735241637088105,
1370
+ "gen_val_win/rouge2": 0.954144793721871,
1371
+ "gen_val_win/rougeL": 0.9735241637088105,
1372
+ "gen_val/n": 128.0,
1373
+ "gen_val/seq_acc": 0.1953125,
1374
+ "gen_val/bit_acc": 0.8208642782107346,
1375
+ "gen_val/lev": 19.8359375,
1376
+ "gen_val/lev_norm": 0.029189174862170115,
1377
+ "gen_val/token_acc": 0.4288859533245266,
1378
+ "gen_val/bleu": 0.9331933491874678,
1379
+ "gen_val/rouge1": 0.9686683826556418,
1380
+ "gen_val/rouge2": 0.9502815413471055,
1381
+ "gen_val/rougeL": 0.9685862607424227,
1382
+ "gen_val_full/n": 128.0,
1383
+ "gen_val_full/seq_acc": 0.0546875,
1384
+ "gen_val_full/bit_acc": 0.3671029687313845,
1385
+ "gen_val_full/lev": 521.71875,
1386
+ "gen_val_full/lev_norm": 0.6435382618336629,
1387
+ "gen_val_full/token_acc": 0.03712480252764613,
1388
+ "gen_val_full/bleu": 0.10995822767720215,
1389
+ "gen_val_full/rouge1": 0.36238189831126444,
1390
+ "gen_val_full/rouge2": 0.2479118388065199,
1391
+ "gen_val_full/rougeL": 0.300873198260952
1392
+ },
1393
+ {
1394
+ "train/epoch_loss": 0.7535619166426528,
1395
+ "epoch": 71,
1396
+ "val/loss": 0.10508038048431718,
1397
+ "val/ppl": 1.110799893400761,
1398
+ "val/token_acc": 0.978199080491867,
1399
+ "val/seq_acc": 0.5820751686476068,
1400
+ "val_full/loss": 5.591022234091475,
1401
+ "val_full/ppl": 268.00944812614483,
1402
+ "val_full/token_acc": 0.3339401023204123,
1403
+ "val_full/seq_acc": 0.07159353348729793,
1404
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1405
+ },
1406
+ {
1407
+ "train/epoch_loss": 0.7523766571744125,
1408
+ "epoch": 72,
1409
+ "val/loss": 0.10527595871316661,
1410
+ "val/ppl": 1.11101716292242,
1411
+ "val/token_acc": 0.9781492156256545,
1412
+ "val/seq_acc": 0.5788628332797944,
1413
+ "val_full/loss": 5.439379586694514,
1414
+ "val_full/ppl": 230.2992584047655,
1415
+ "val_full/token_acc": 0.34662476767134837,
1416
+ "val_full/seq_acc": 0.07159353348729793,
1417
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1418
+ },
1419
+ {
1420
+ "train/epoch_loss": 0.7505472156769617,
1421
+ "epoch": 73,
1422
+ "val/loss": 0.10373660443856816,
1423
+ "val/ppl": 1.1093082295679142,
1424
+ "val/token_acc": 0.9783087831975347,
1425
+ "val/seq_acc": 0.5804690009637006,
1426
+ "val_full/loss": 5.596636443982448,
1427
+ "val_full/ppl": 269.51834107828176,
1428
+ "val_full/token_acc": 0.3318515395964667,
1429
+ "val_full/seq_acc": 0.07852193995381063,
1430
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1431
+ },
1432
+ {
1433
+ "train/epoch_loss": 0.7498734516098612,
1434
+ "epoch": 74,
1435
+ "val/loss": 0.10334283763851525,
1436
+ "val/ppl": 1.1088715068051902,
1437
+ "val/token_acc": 0.9786678102342652,
1438
+ "val/seq_acc": 0.5833601027947317,
1439
+ "val_full/loss": 5.616988029685374,
1440
+ "val_full/ppl": 275.05966277632007,
1441
+ "val_full/token_acc": 0.3316982505892046,
1442
+ "val_full/seq_acc": 0.07621247113163972,
1443
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1444
+ },
1445
+ {
1446
+ "train/epoch_loss": 0.7482658022656999,
1447
+ "epoch": 75,
1448
+ "val/loss": 0.10210829080438107,
1449
+ "val/ppl": 1.107503397668166,
1450
+ "val/token_acc": 0.9787176751004777,
1451
+ "val/seq_acc": 0.583681336331513,
1452
+ "val_full/loss": 5.644142339022591,
1453
+ "val_full/ppl": 282.63105062072907,
1454
+ "val_full/token_acc": 0.3300120715093219,
1455
+ "val_full/seq_acc": 0.07159353348729793,
1456
+ "system/peak_gpu_mem_gb": 1.0650625228881836,
1457
+ "gen_val_win/n": 128.0,
1458
+ "gen_val_win/seq_acc": 0.6484375,
1459
+ "gen_val_win/bit_acc": 0.9476811781423589,
1460
+ "gen_val_win/lev": 1.90625,
1461
+ "gen_val_win/lev_norm": 0.01187550800905095,
1462
+ "gen_val_win/token_acc": 0.8533498759305211,
1463
+ "gen_val_win/bleu": 0.9433507117683827,
1464
+ "gen_val_win/rouge1": 0.9736264156205103,
1465
+ "gen_val_win/rouge2": 0.9563676193284467,
1466
+ "gen_val_win/rougeL": 0.9736264156205103,
1467
+ "gen_val/n": 128.0,
1468
+ "gen_val/seq_acc": 0.2421875,
1469
+ "gen_val/bit_acc": 0.8271393643031785,
1470
+ "gen_val/lev": 18.8125,
1471
+ "gen_val/lev_norm": 0.026790460131742057,
1472
+ "gen_val/token_acc": 0.4637106641217638,
1473
+ "gen_val/bleu": 0.935424432617851,
1474
+ "gen_val/rouge1": 0.9709846665014976,
1475
+ "gen_val/rouge2": 0.95441035265913,
1476
+ "gen_val/rougeL": 0.9709517717646555,
1477
+ "gen_val_full/n": 128.0,
1478
+ "gen_val_full/seq_acc": 0.0625,
1479
+ "gen_val_full/bit_acc": 0.34337851148687143,
1480
+ "gen_val_full/lev": 600.1015625,
1481
+ "gen_val_full/lev_norm": 0.6594099125553171,
1482
+ "gen_val_full/token_acc": 0.032464076636508785,
1483
+ "gen_val_full/bleu": 0.11365944691942924,
1484
+ "gen_val_full/rouge1": 0.34628273502360013,
1485
+ "gen_val_full/rouge2": 0.2298203567705291,
1486
+ "gen_val_full/rougeL": 0.2840028360407299
1487
+ },
1488
+ {
1489
+ "train/epoch_loss": 0.7471561143523142,
1490
+ "epoch": 76,
1491
+ "val/loss": 0.10176311059638243,
1492
+ "val/ppl": 1.1071211753865784,
1493
+ "val/token_acc": 0.978588026448325,
1494
+ "val/seq_acc": 0.5843238034050755,
1495
+ "val_full/loss": 5.645573211069382,
1496
+ "val_full/ppl": 283.035748957338,
1497
+ "val_full/token_acc": 0.32731035275632797,
1498
+ "val_full/seq_acc": 0.07621247113163972,
1499
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1500
+ },
1501
+ {
1502
+ "train/epoch_loss": 0.7469619139471554,
1503
+ "epoch": 77,
1504
+ "val/loss": 0.10257696416575826,
1505
+ "val/ppl": 1.1080225766614666,
1506
+ "val/token_acc": 0.9790866751104507,
1507
+ "val/seq_acc": 0.5878573723096692,
1508
+ "val_full/loss": 5.633106204133055,
1509
+ "val_full/ppl": 279.5290447838613,
1510
+ "val_full/token_acc": 0.3300120715093219,
1511
+ "val_full/seq_acc": 0.08314087759815242,
1512
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1513
+ },
1514
+ {
1515
+ "train/epoch_loss": 0.7454144519166161,
1516
+ "epoch": 78,
1517
+ "val/loss": 0.10266572934941523,
1518
+ "val/ppl": 1.1081209348543068,
1519
+ "val/token_acc": 0.9785979994215676,
1520
+ "val/seq_acc": 0.5843238034050755,
1521
+ "val_full/loss": 5.641723022332292,
1522
+ "val_full/ppl": 281.9481030690838,
1523
+ "val_full/token_acc": 0.32993542700569084,
1524
+ "val_full/seq_acc": 0.07852193995381063,
1525
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1526
+ },
1527
+ {
1528
+ "train/epoch_loss": 0.7446135588714904,
1529
+ "epoch": 79,
1530
+ "val/loss": 0.10225602011188839,
1531
+ "val/ppl": 1.1076670204638093,
1532
+ "val/token_acc": 0.9788473237526304,
1533
+ "val/seq_acc": 0.5881786058464503,
1534
+ "val_full/loss": 5.644910430718159,
1535
+ "val_full/ppl": 282.8482205761633,
1536
+ "val_full/token_acc": 0.331429994826496,
1537
+ "val_full/seq_acc": 0.08083140877598152,
1538
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1539
+ },
1540
+ {
1541
+ "train/epoch_loss": 0.7434256564351983,
1542
+ "epoch": 80,
1543
+ "val/loss": 0.10209512184139186,
1544
+ "val/ppl": 1.107488813092944,
1545
+ "val/token_acc": 0.9790368102442382,
1546
+ "val/seq_acc": 0.5913909412142627,
1547
+ "val_full/loss": 5.628657005666903,
1548
+ "val_full/ppl": 278.28812717795273,
1549
+ "val_full/token_acc": 0.33104677230834084,
1550
+ "val_full/seq_acc": 0.08083140877598152,
1551
+ "system/peak_gpu_mem_gb": 1.0650625228881836,
1552
+ "gen_val_win/n": 128.0,
1553
+ "gen_val_win/seq_acc": 0.6328125,
1554
+ "gen_val_win/bit_acc": 0.9589264820592823,
1555
+ "gen_val_win/lev": 1.46875,
1556
+ "gen_val_win/lev_norm": 0.011870788211873987,
1557
+ "gen_val_win/token_acc": 0.837447020693094,
1558
+ "gen_val_win/bleu": 0.9458252409859191,
1559
+ "gen_val_win/rouge1": 0.9736477876341096,
1560
+ "gen_val_win/rouge2": 0.9547537997486909,
1561
+ "gen_val_win/rougeL": 0.9736477876341096,
1562
+ "gen_val/n": 128.0,
1563
+ "gen_val/seq_acc": 0.2265625,
1564
+ "gen_val/bit_acc": 0.8308444020449598,
1565
+ "gen_val/lev": 19.421875,
1566
+ "gen_val/lev_norm": 0.027883458066460365,
1567
+ "gen_val/token_acc": 0.48211485691885536,
1568
+ "gen_val/bleu": 0.9346281821713329,
1569
+ "gen_val/rouge1": 0.970167401334752,
1570
+ "gen_val/rouge2": 0.952445849287562,
1571
+ "gen_val/rougeL": 0.9700579496696512,
1572
+ "gen_val_full/n": 128.0,
1573
+ "gen_val_full/seq_acc": 0.0625,
1574
+ "gen_val_full/bit_acc": 0.33647078636998906,
1575
+ "gen_val_full/lev": 574.203125,
1576
+ "gen_val_full/lev_norm": 0.6577566522728574,
1577
+ "gen_val_full/token_acc": 0.03564499687613035,
1578
+ "gen_val_full/bleu": 0.10642088306875847,
1579
+ "gen_val_full/rouge1": 0.3347771511495598,
1580
+ "gen_val_full/rouge2": 0.2283691153378196,
1581
+ "gen_val_full/rougeL": 0.2830156498412585
1582
+ },
1583
+ {
1584
+ "train/epoch_loss": 0.7427948204953772,
1585
+ "epoch": 81,
1586
+ "val/loss": 0.1017492253433381,
1587
+ "val/ppl": 1.1071058028356335,
1588
+ "val/token_acc": 0.9790368102442382,
1589
+ "val/seq_acc": 0.5920334082878252,
1590
+ "val_full/loss": 5.572770887423595,
1591
+ "val_full/ppl": 263.1622829770605,
1592
+ "val_full/token_acc": 0.33472570848263045,
1593
+ "val_full/seq_acc": 0.07852193995381063,
1594
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1595
+ },
1596
+ {
1597
+ "train/epoch_loss": 0.7426253491803595,
1598
+ "epoch": 82,
1599
+ "val/loss": 0.1011202050432829,
1600
+ "val/ppl": 1.106409629787785,
1601
+ "val/token_acc": 0.9790966480836932,
1602
+ "val/seq_acc": 0.5920334082878252,
1603
+ "val_full/loss": 5.594441147799345,
1604
+ "val_full/ppl": 268.9273174661375,
1605
+ "val_full/token_acc": 0.33342275192090287,
1606
+ "val_full/seq_acc": 0.07852193995381063,
1607
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1608
+ },
1609
+ {
1610
+ "train/epoch_loss": 0.7415810908165359,
1611
+ "epoch": 83,
1612
+ "val/loss": 0.10094672155499612,
1613
+ "val/ppl": 1.1062177026343127,
1614
+ "val/token_acc": 0.9789869453780256,
1615
+ "val/seq_acc": 0.5907484741407003,
1616
+ "val_full/loss": 5.5852166374858685,
1617
+ "val_full/ppl": 266.4580012787265,
1618
+ "val_full/token_acc": 0.33643104868842094,
1619
+ "val_full/seq_acc": 0.07852193995381063,
1620
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1621
+ },
1622
+ {
1623
+ "train/epoch_loss": 0.7408571424626947,
1624
+ "epoch": 84,
1625
+ "val/loss": 0.1009505429860274,
1626
+ "val/ppl": 1.1062219299770462,
1627
+ "val/token_acc": 0.9792362697090884,
1628
+ "val/seq_acc": 0.5929971088981689,
1629
+ "val_full/loss": 5.707580410862442,
1630
+ "val_full/ppl": 301.1415472515213,
1631
+ "val_full/token_acc": 0.32464695625514955,
1632
+ "val_full/seq_acc": 0.07621247113163972,
1633
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1634
+ },
1635
+ {
1636
+ "train/epoch_loss": 0.7399543473250848,
1637
+ "epoch": 85,
1638
+ "val/loss": 0.10107410515099295,
1639
+ "val/ppl": 1.1063586255986761,
1640
+ "val/token_acc": 0.9791963778161183,
1641
+ "val/seq_acc": 0.5929971088981689,
1642
+ "val_full/loss": 5.687350004311253,
1643
+ "val_full/ppl": 295.11054185742114,
1644
+ "val_full/token_acc": 0.3250493398992125,
1645
+ "val_full/seq_acc": 0.07852193995381063,
1646
+ "system/peak_gpu_mem_gb": 1.0650625228881836,
1647
+ "gen_val_win/n": 128.0,
1648
+ "gen_val_win/seq_acc": 0.65625,
1649
+ "gen_val_win/bit_acc": 0.936570142391929,
1650
+ "gen_val_win/lev": 3.5703125,
1651
+ "gen_val_win/lev_norm": 0.015941331271142705,
1652
+ "gen_val_win/token_acc": 0.8228346456692913,
1653
+ "gen_val_win/bleu": 0.9309919214200336,
1654
+ "gen_val_win/rouge1": 0.9705177049034177,
1655
+ "gen_val_win/rouge2": 0.9530451958188538,
1656
+ "gen_val_win/rougeL": 0.9705177049034177,
1657
+ "gen_val/n": 128.0,
1658
+ "gen_val/seq_acc": 0.265625,
1659
+ "gen_val/bit_acc": 0.8383745107632093,
1660
+ "gen_val/lev": 19.0625,
1661
+ "gen_val/lev_norm": 0.025524067608020805,
1662
+ "gen_val/token_acc": 0.4900353426271353,
1663
+ "gen_val/bleu": 0.936846118862419,
1664
+ "gen_val/rouge1": 0.971977426375064,
1665
+ "gen_val/rouge2": 0.9559564493688095,
1666
+ "gen_val/rougeL": 0.9719442522773994,
1667
+ "gen_val_full/n": 128.0,
1668
+ "gen_val_full/seq_acc": 0.0703125,
1669
+ "gen_val_full/bit_acc": 0.33686501520985346,
1670
+ "gen_val_full/lev": 549.171875,
1671
+ "gen_val_full/lev_norm": 0.6643301571195462,
1672
+ "gen_val_full/token_acc": 0.03650568672866181,
1673
+ "gen_val_full/bleu": 0.0961379279523444,
1674
+ "gen_val_full/rouge1": 0.33109183900682865,
1675
+ "gen_val_full/rouge2": 0.2272072947724632,
1676
+ "gen_val_full/rougeL": 0.2799981833360931
1677
+ },
1678
+ {
1679
+ "train/epoch_loss": 0.7401719008597947,
1680
+ "epoch": 86,
1681
+ "val/loss": 0.10011620043191252,
1682
+ "val/ppl": 1.1052993468752617,
1683
+ "val/token_acc": 0.9791963778161183,
1684
+ "val/seq_acc": 0.5920334082878252,
1685
+ "val_full/loss": 5.726109444159689,
1686
+ "val_full/ppl": 306.77342449878154,
1687
+ "val_full/token_acc": 0.3233248385675142,
1688
+ "val_full/seq_acc": 0.08314087759815242,
1689
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1690
+ },
1691
+ {
1692
+ "train/epoch_loss": 0.7391969173626413,
1693
+ "epoch": 87,
1694
+ "val/loss": 0.1004848159704973,
1695
+ "val/ppl": 1.1057068524911442,
1696
+ "val/token_acc": 0.9795254859331213,
1697
+ "val/seq_acc": 0.5974943784131064,
1698
+ "val_full/loss": 5.680714416591619,
1699
+ "val_full/ppl": 293.1587926329848,
1700
+ "val_full/token_acc": 0.3283833758071624,
1701
+ "val_full/seq_acc": 0.08314087759815242,
1702
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1703
+ },
1704
+ {
1705
+ "train/epoch_loss": 0.7387118480746585,
1706
+ "epoch": 88,
1707
+ "val/loss": 0.10021054421704194,
1708
+ "val/ppl": 1.1054036299184982,
1709
+ "val/token_acc": 0.9794157832274536,
1710
+ "val/seq_acc": 0.5907484741407003,
1711
+ "val_full/loss": 5.7281915699429,
1712
+ "val_full/ppl": 307.4128307866947,
1713
+ "val_full/token_acc": 0.3224625879016651,
1714
+ "val_full/seq_acc": 0.08083140877598152,
1715
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1716
+ },
1717
+ {
1718
+ "train/epoch_loss": 0.7382475876451431,
1719
+ "epoch": 89,
1720
+ "val/loss": 0.10013674062370566,
1721
+ "val/ppl": 1.1053220501689995,
1722
+ "val/token_acc": 0.979365918361241,
1723
+ "val/seq_acc": 0.5926758753613878,
1724
+ "val_full/loss": 5.713398656805074,
1725
+ "val_full/ppl": 302.89876985650113,
1726
+ "val_full/token_acc": 0.3234781275747763,
1727
+ "val_full/seq_acc": 0.08314087759815242,
1728
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1729
+ },
1730
+ {
1731
+ "train/epoch_loss": 0.7378488964868009,
1732
+ "epoch": 90,
1733
+ "val/loss": 0.10072225372146756,
1734
+ "val/ppl": 1.105969420209943,
1735
+ "val/token_acc": 0.9793758913344836,
1736
+ "val/seq_acc": 0.5936395759717314,
1737
+ "val_full/loss": 5.65790436861695,
1738
+ "val_full/ppl": 286.54751494005836,
1739
+ "val_full/token_acc": 0.32836421468125465,
1740
+ "val_full/seq_acc": 0.08545034642032333,
1741
+ "system/peak_gpu_mem_gb": 1.0650625228881836,
1742
+ "gen_val_win/n": 128.0,
1743
+ "gen_val_win/seq_acc": 0.65625,
1744
+ "gen_val_win/bit_acc": 0.9397923322683706,
1745
+ "gen_val_win/lev": 3.453125,
1746
+ "gen_val_win/lev_norm": 0.015042173523238098,
1747
+ "gen_val_win/token_acc": 0.8247600295348265,
1748
+ "gen_val_win/bleu": 0.9334896953527388,
1749
+ "gen_val_win/rouge1": 0.971912387630821,
1750
+ "gen_val_win/rouge2": 0.9553194334608761,
1751
+ "gen_val_win/rougeL": 0.971912387630821,
1752
+ "gen_val/n": 128.0,
1753
+ "gen_val/seq_acc": 0.25,
1754
+ "gen_val/bit_acc": 0.8328767206872987,
1755
+ "gen_val/lev": 20.34375,
1756
+ "gen_val/lev_norm": 0.028285833316708944,
1757
+ "gen_val/token_acc": 0.4728242258206546,
1758
+ "gen_val/bleu": 0.9344087206825136,
1759
+ "gen_val/rouge1": 0.9704404877921282,
1760
+ "gen_val/rouge2": 0.9538362591267303,
1761
+ "gen_val/rougeL": 0.9704068856415906,
1762
+ "gen_val_full/n": 128.0,
1763
+ "gen_val_full/seq_acc": 0.0625,
1764
+ "gen_val_full/bit_acc": 0.355097238973122,
1765
+ "gen_val_full/lev": 522.0703125,
1766
+ "gen_val_full/lev_norm": 0.639539040440194,
1767
+ "gen_val_full/token_acc": 0.038817536055105115,
1768
+ "gen_val_full/bleu": 0.10566389750921357,
1769
+ "gen_val_full/rouge1": 0.34989792471680825,
1770
+ "gen_val_full/rouge2": 0.237879814249418,
1771
+ "gen_val_full/rougeL": 0.2906843767576277
1772
+ },
1773
+ {
1774
+ "train/epoch_loss": 0.737777391425392,
1775
+ "epoch": 91,
1776
+ "val/loss": 0.09986305337632671,
1777
+ "val/ppl": 1.1050195790127586,
1778
+ "val/token_acc": 0.9795055399866363,
1779
+ "val/seq_acc": 0.5962094442659813,
1780
+ "val_full/loss": 5.693944874638333,
1781
+ "val_full/ppl": 297.0631892642554,
1782
+ "val_full/token_acc": 0.3261032018241392,
1783
+ "val_full/seq_acc": 0.08545034642032333,
1784
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1785
+ },
1786
+ {
1787
+ "train/epoch_loss": 0.7368834512192115,
1788
+ "epoch": 92,
1789
+ "val/loss": 0.09936335234456517,
1790
+ "val/ppl": 1.104467537528346,
1791
+ "val/token_acc": 0.9794756210669087,
1792
+ "val/seq_acc": 0.5962094442659813,
1793
+ "val_full/loss": 5.677080015065435,
1794
+ "val_full/ppl": 292.09526967501756,
1795
+ "val_full/token_acc": 0.3263331353350323,
1796
+ "val_full/seq_acc": 0.08314087759815242,
1797
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1798
+ },
1799
+ {
1800
+ "train/epoch_loss": 0.737032663643806,
1801
+ "epoch": 93,
1802
+ "val/loss": 0.09950750781057163,
1803
+ "val/ppl": 1.1046267640373175,
1804
+ "val/token_acc": 0.9795753507993338,
1805
+ "val/seq_acc": 0.5965306778027626,
1806
+ "val_full/loss": 5.6828042517340815,
1807
+ "val_full/ppl": 293.7720867987923,
1808
+ "val_full/token_acc": 0.32495353426967366,
1809
+ "val_full/seq_acc": 0.08545034642032333,
1810
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1811
+ },
1812
+ {
1813
+ "train/epoch_loss": 0.7366967462839331,
1814
+ "epoch": 94,
1815
+ "val/loss": 0.09931489304731249,
1816
+ "val/ppl": 1.1044140171044305,
1817
+ "val/token_acc": 0.9795952967458188,
1818
+ "val/seq_acc": 0.5978156119498875,
1819
+ "val_full/loss": 5.678785280423078,
1820
+ "val_full/ppl": 292.5937945573253,
1821
+ "val_full/token_acc": 0.3254134012914599,
1822
+ "val_full/seq_acc": 0.08545034642032333,
1823
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1824
+ },
1825
+ {
1826
+ "train/epoch_loss": 0.7365873708689302,
1827
+ "epoch": 95,
1828
+ "val/loss": 0.09952979725968017,
1829
+ "val/ppl": 1.1046513858337605,
1830
+ "val/token_acc": 0.9796651075585164,
1831
+ "val/seq_acc": 0.5949245101188564,
1832
+ "val_full/loss": 5.713141852809021,
1833
+ "val_full/ppl": 302.82099422897585,
1834
+ "val_full/token_acc": 0.3227500047902815,
1835
+ "val_full/seq_acc": 0.08314087759815242,
1836
+ "system/peak_gpu_mem_gb": 1.0650625228881836,
1837
+ "gen_val_win/n": 128.0,
1838
+ "gen_val_win/seq_acc": 0.6328125,
1839
+ "gen_val_win/bit_acc": 0.9246040544821033,
1840
+ "gen_val_win/lev": 4.5625,
1841
+ "gen_val_win/lev_norm": 0.016696424812685837,
1842
+ "gen_val_win/token_acc": 0.8137542829172785,
1843
+ "gen_val_win/bleu": 0.9220869414589093,
1844
+ "gen_val_win/rouge1": 0.9701212523269171,
1845
+ "gen_val_win/rouge2": 0.9529250186396473,
1846
+ "gen_val_win/rougeL": 0.9701212523269171,
1847
+ "gen_val/n": 128.0,
1848
+ "gen_val/seq_acc": 0.2421875,
1849
+ "gen_val/bit_acc": 0.8262547146854848,
1850
+ "gen_val/lev": 22.0546875,
1851
+ "gen_val/lev_norm": 0.03003813391315662,
1852
+ "gen_val/token_acc": 0.47081484371190324,
1853
+ "gen_val/bleu": 0.931260298878397,
1854
+ "gen_val/rouge1": 0.9696616974165042,
1855
+ "gen_val/rouge2": 0.9529502680523851,
1856
+ "gen_val/rougeL": 0.9696281673735857,
1857
+ "gen_val_full/n": 128.0,
1858
+ "gen_val_full/seq_acc": 0.0625,
1859
+ "gen_val_full/bit_acc": 0.3551788052891724,
1860
+ "gen_val_full/lev": 540.96875,
1861
+ "gen_val_full/lev_norm": 0.6521694306109027,
1862
+ "gen_val_full/token_acc": 0.03675133059354578,
1863
+ "gen_val_full/bleu": 0.1015755239554731,
1864
+ "gen_val_full/rouge1": 0.34066941377132964,
1865
+ "gen_val_full/rouge2": 0.2313868540644178,
1866
+ "gen_val_full/rougeL": 0.2839247701539138
1867
+ },
1868
+ {
1869
+ "train/epoch_loss": 0.7363973328300247,
1870
+ "epoch": 96,
1871
+ "val/loss": 0.09948764023446655,
1872
+ "val/ppl": 1.1046048179990233,
1873
+ "val/token_acc": 0.9797049994514865,
1874
+ "val/seq_acc": 0.5971731448763251,
1875
+ "val_full/loss": 5.77722736112974,
1876
+ "val_full/ppl": 322.86276640559106,
1877
+ "val_full/token_acc": 0.31815133457241945,
1878
+ "val_full/seq_acc": 0.08083140877598152,
1879
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1880
+ },
1881
+ {
1882
+ "train/epoch_loss": 0.7360071862427671,
1883
+ "epoch": 97,
1884
+ "val/loss": 0.10000829233185782,
1885
+ "val/ppl": 1.1051800825576574,
1886
+ "val/token_acc": 0.9797448913444565,
1887
+ "val/seq_acc": 0.59845807902345,
1888
+ "val_full/loss": 5.743435518021039,
1889
+ "val_full/ppl": 312.1349161880764,
1890
+ "val_full/token_acc": 0.3214662093544617,
1891
+ "val_full/seq_acc": 0.08314087759815242,
1892
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1893
+ },
1894
+ {
1895
+ "train/epoch_loss": 0.7357251083761677,
1896
+ "epoch": 98,
1897
+ "val/loss": 0.09932624503489099,
1898
+ "val/ppl": 1.1044265544697962,
1899
+ "val/token_acc": 0.9797249453979715,
1900
+ "val/seq_acc": 0.5974943784131064,
1901
+ "val_full/loss": 5.735954670165169,
1902
+ "val_full/ppl": 309.80859464732436,
1903
+ "val_full/token_acc": 0.32181110962080134,
1904
+ "val_full/seq_acc": 0.08083140877598152,
1905
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1906
+ },
1907
+ {
1908
+ "train/epoch_loss": 0.7358884735594962,
1909
+ "epoch": 99,
1910
+ "val/loss": 0.09961430160556396,
1911
+ "val/ppl": 1.1047447376208104,
1912
+ "val/token_acc": 0.9796252156655464,
1913
+ "val/seq_acc": 0.5974943784131064,
1914
+ "val_full/loss": 5.736716100375558,
1915
+ "val_full/ppl": 310.04458210331444,
1916
+ "val_full/token_acc": 0.32100634233267544,
1917
+ "val_full/seq_acc": 0.08314087759815242,
1918
+ "system/peak_gpu_mem_gb": 1.0650625228881836
1919
+ },
1920
+ {
1921
+ "train/epoch_loss": 0.7359878973770617,
1922
+ "epoch": 100,
1923
+ "val/loss": 0.09927769922586938,
1924
+ "val/ppl": 1.104372940490582,
1925
+ "val/token_acc": 0.9795753507993338,
1926
+ "val/seq_acc": 0.5978156119498875,
1927
+ "val_full/loss": 5.748038529431489,
1928
+ "val_full/ppl": 313.5749885610196,
1929
+ "val_full/token_acc": 0.3209488589549522,
1930
+ "val_full/seq_acc": 0.08083140877598152,
1931
+ "system/peak_gpu_mem_gb": 1.0650625228881836,
1932
+ "gen_val_win/n": 128.0,
1933
+ "gen_val_win/seq_acc": 0.6328125,
1934
+ "gen_val_win/bit_acc": 0.9441492797622892,
1935
+ "gen_val_win/lev": 2.1796875,
1936
+ "gen_val_win/lev_norm": 0.014352797667106544,
1937
+ "gen_val_win/token_acc": 0.8238946845504223,
1938
+ "gen_val_win/bleu": 0.9441925726890976,
1939
+ "gen_val_win/rouge1": 0.9727711842923557,
1940
+ "gen_val_win/rouge2": 0.9555740634354313,
1941
+ "gen_val_win/rougeL": 0.9727711842923557,
1942
+ "gen_val/n": 128.0,
1943
+ "gen_val/seq_acc": 0.25,
1944
+ "gen_val/bit_acc": 0.8362129070137241,
1945
+ "gen_val/lev": 18.828125,
1946
+ "gen_val/lev_norm": 0.027976049053401193,
1947
+ "gen_val/token_acc": 0.48612336961851527,
1948
+ "gen_val/bleu": 0.9366272176056055,
1949
+ "gen_val/rouge1": 0.9708670633474195,
1950
+ "gen_val/rouge2": 0.9542004447299458,
1951
+ "gen_val/rougeL": 0.9708336765952826,
1952
+ "gen_val_full/n": 128.0,
1953
+ "gen_val_full/seq_acc": 0.0625,
1954
+ "gen_val_full/bit_acc": 0.3532414452475822,
1955
+ "gen_val_full/lev": 573.65625,
1956
+ "gen_val_full/lev_norm": 0.6462366040228255,
1957
+ "gen_val_full/token_acc": 0.03648509763617677,
1958
+ "gen_val_full/bleu": 0.10978491718305944,
1959
+ "gen_val_full/rouge1": 0.33893754392655,
1960
+ "gen_val_full/rouge2": 0.23012860870149393,
1961
+ "gen_val_full/rougeL": 0.2863141438838297
1962
+ }
1963
+ ]
1964
+ }
tokenizer_cipher.json ADDED
The diff for this file is too large to render. See raw diff
 
tokenizer_plain.json ADDED
The diff for this file is too large to render. See raw diff