andyjzhao commited on
Commit
3353078
·
verified ·
1 Parent(s): 0355e3e

Upload folder using huggingface_hub

Browse files
checkpoint-20/optimizer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b733d59b356122c9ec251414b9f1830f05e11883e5c3d56e7223894fdfb9eb9d
3
  size 558664971
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d7e8b4099cc856e696097aabaee14e621bd73ddc512db76d7da20110fd3b27a6
3
  size 558664971
checkpoint-20/pytorch_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:ff0c41a557a265632e67741268f9a9812ee7dddccf952ee20a8a47f01ad5ad88
3
  size 279336283
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:027fb67d9027b2f17e742728bfc764e361abf16b22fdc66666f19fb7636f4b93
3
  size 279336283
checkpoint-20/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9c213a373f3e5d95993ad095a3790a902d821a1b4b93a10cc7d382c8726fcb9d
3
+ size 15429
checkpoint-20/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8fb125336725f7741cb4daa1e3d06e225bbacfde8d41c4dcabb6762c222e62c6
3
+ size 15429
checkpoint-20/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:153c865f77c7129ba565bded50f334683d51c80f20e3cfec39e62f8737b86f0d
3
+ size 15429
checkpoint-20/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d017ce00fcebac7edc058ddd138f194eb0340f2d8ad0879bdab08f922ed0846e
3
+ size 15429
checkpoint-20/trainer_state.json CHANGED
@@ -1,8 +1,8 @@
1
  {
2
  "best_global_step": 20,
3
- "best_metric": 124944.5436483137,
4
  "best_model_checkpoint": "/network/scratch/j/jianan.zhao/DNAFM/output/gencode_human_12.8k_12800/CKPT_DEBUG/checkpoint-20",
5
- "epoch": 0.0007091444172605751,
6
  "eval_steps": 10,
7
  "global_step": 20,
8
  "is_hyper_param_search": false,
@@ -12,118 +12,118 @@
12
  {
13
  "comp/rl_weight": 0.03,
14
  "comp/strictness": 0.0,
15
- "epoch": 0.00035457220863028757,
16
- "grad_norm": 1130.100830078125,
17
- "loss": 146.4512,
18
- "loss_ce": 54.19949722290039,
19
- "loss_region": 0.10721047222614288,
20
- "loss_total": 54.30670928955078,
21
  "lr": 2.20454076850486e-05,
22
- "router/selected_tokens_s0": 24.15625,
23
  "router/selected_tokens_s1": 1.0,
24
  "step": 10,
25
- "tokens_trained": 0.00409568
26
  },
27
  {
28
- "epoch": 0.00035457220863028757,
29
- "eval_ppl": 7.294625624845928e+17,
30
- "eval_runtime": 6.6402,
31
  "step": 10,
32
- "tokens_trained": 0.00409568
33
  },
34
  {
35
- "epoch": 0.00035457220863028757,
36
  "eval_F": 7.813110399249941e-05,
37
- "eval_F_cds": 4.744345662481978e-05,
38
  "eval_F_dig": 7.813110399249941e-05,
39
- "eval_F_exon": 6.977782739756614e-05,
40
- "eval_F_intron": 8.580132276805985e-05,
41
- "eval_F_nig": 8.365496566708533e-05,
42
- "eval_F_promoter": 4.770454407466195e-05,
43
- "eval_F_utr": 7.9026626119178e-05,
44
- "eval_G": 0.00016736338285193968,
45
- "eval_G_cds": 0.00016425546555209688,
46
- "eval_G_dig": 0.000140951360132302,
47
- "eval_G_exon": 0.0001619063651334152,
48
- "eval_G_intron": 0.00016938910394235382,
49
- "eval_G_nig": 0.0001631118140770298,
50
- "eval_G_promoter": 0.00016429247622091106,
51
- "eval_G_utr": 0.00019291894904831453,
52
  "eval_avg_bp_per_token": 12799.0,
53
- "eval_bp_per_token/cds": 21077.722222222223,
54
  "eval_bp_per_token/dig": 12799.0,
55
- "eval_bp_per_token/exon": 14331.2,
56
- "eval_bp_per_token/intron": 11654.831973898858,
57
- "eval_bp_per_token/nig": 11953.86301369863,
58
- "eval_bp_per_token/promoter": 20962.363636363636,
59
- "eval_bp_per_token/utr": 12653.962962962964,
60
- "eval_ppl_cds": 2.311600502337928e+20,
61
- "eval_ppl_dig": 9.652397143812314e+16,
62
- "eval_ppl_exon": 2.2747493338461386e+18,
63
- "eval_ppl_intron": 2.793470782953012e+17,
64
- "eval_ppl_nig": 1.1026701972335622e+17,
65
- "eval_ppl_promoter": 1.5383074335283397e+20,
66
- "eval_ppl_utr": 4.222450573520073e+17,
67
  "step": 10,
68
- "tokens_trained": 0.00409568
69
  },
70
  {
71
  "comp/rl_weight": 0.03,
72
  "comp/strictness": 0.0,
73
- "epoch": 0.0007091444172605751,
74
- "grad_norm": 299.7831115722656,
75
- "loss": 26.1005,
76
- "loss_ce": 12.792510032653809,
77
- "loss_region": 0.09937700629234314,
78
- "loss_total": 12.891886711120605,
79
  "lr": 4.654030511288038e-05,
80
  "router/selected_tokens_s0": 1.0,
81
  "router/selected_tokens_s1": 1.0,
82
  "step": 20,
83
- "tokens_trained": 0.00819136
84
  },
85
  {
86
- "epoch": 0.0007091444172605751,
87
- "eval_ppl": 124944.5436483137,
88
- "eval_runtime": 6.5805,
89
  "step": 20,
90
- "tokens_trained": 0.00819136
91
  },
92
  {
93
- "epoch": 0.0007091444172605751,
94
  "eval_F": 7.813110399249941e-05,
95
- "eval_F_cds": 4.744345662481978e-05,
96
  "eval_F_dig": 7.813110399249941e-05,
97
- "eval_F_exon": 6.977782739756614e-05,
98
- "eval_F_intron": 8.580132276805985e-05,
99
- "eval_F_nig": 8.365496566708533e-05,
100
- "eval_F_promoter": 4.770454407466195e-05,
101
- "eval_F_utr": 7.9026626119178e-05,
102
  "eval_G": 7.813110399249941e-05,
103
- "eval_G_cds": 4.744345662481978e-05,
104
  "eval_G_dig": 7.813110399249941e-05,
105
- "eval_G_exon": 6.977782739756614e-05,
106
- "eval_G_intron": 8.580132276805985e-05,
107
- "eval_G_nig": 8.365496566708533e-05,
108
- "eval_G_promoter": 4.770454407466195e-05,
109
- "eval_G_utr": 7.9026626119178e-05,
110
  "eval_avg_bp_per_token": 12799.0,
111
- "eval_bp_per_token/cds": 21077.722222222223,
112
  "eval_bp_per_token/dig": 12799.0,
113
- "eval_bp_per_token/exon": 14331.2,
114
- "eval_bp_per_token/intron": 11654.831973898858,
115
- "eval_bp_per_token/nig": 11953.86301369863,
116
- "eval_bp_per_token/promoter": 20962.363636363636,
117
- "eval_bp_per_token/utr": 12653.962962962964,
118
- "eval_ppl_cds": 287230.0213028405,
119
- "eval_ppl_dig": 25004.34707335865,
120
- "eval_ppl_exon": 108975.3496962627,
121
- "eval_ppl_intron": 104946.6793164646,
122
- "eval_ppl_nig": 100265.49289711946,
123
- "eval_ppl_promoter": 307025.0147172088,
124
- "eval_ppl_utr": 179407.97736471496,
125
  "step": 20,
126
- "tokens_trained": 0.00819136
127
  }
128
  ],
129
  "logging_steps": 10,
 
1
  {
2
  "best_global_step": 20,
3
+ "best_metric": 6586376.056868001,
4
  "best_model_checkpoint": "/network/scratch/j/jianan.zhao/DNAFM/output/gencode_human_12.8k_12800/CKPT_DEBUG/checkpoint-20",
5
+ "epoch": 0.0028368794326241137,
6
  "eval_steps": 10,
7
  "global_step": 20,
8
  "is_hyper_param_search": false,
 
12
  {
13
  "comp/rl_weight": 0.03,
14
  "comp/strictness": 0.0,
15
+ "epoch": 0.0014184397163120568,
16
+ "grad_norm": 904.2174682617188,
17
+ "loss": 144.6514,
18
+ "loss_ce": 54.04480743408203,
19
+ "loss_region": 0.1044125035405159,
20
+ "loss_total": 54.14921951293945,
21
  "lr": 2.20454076850486e-05,
22
+ "router/selected_tokens_s0": 22.09375,
23
  "router/selected_tokens_s1": 1.0,
24
  "step": 10,
25
+ "tokens_trained": 0.01638272
26
  },
27
  {
28
+ "epoch": 0.0014184397163120568,
29
+ "eval_ppl": 1.698420053510286e+17,
30
+ "eval_runtime": 1.7385,
31
  "step": 10,
32
+ "tokens_trained": 0.01638272
33
  },
34
  {
35
+ "epoch": 0.0014184397163120568,
36
  "eval_F": 7.813110399249941e-05,
37
+ "eval_F_cds": 2.439649178448139e-05,
38
  "eval_F_dig": 7.813110399249941e-05,
39
+ "eval_F_exon": 6.169412055031155e-05,
40
+ "eval_F_intron": 8.631567698207511e-05,
41
+ "eval_F_nig": 8.867784676802712e-05,
42
+ "eval_F_promoter": 2.898993785611503e-05,
43
+ "eval_F_utr": 0.00013418766144453016,
44
+ "eval_G": 0.00021740636410105978,
45
+ "eval_G_cds": 0.00017406325095451275,
46
+ "eval_G_dig": 0.00015412581061020393,
47
+ "eval_G_exon": 0.00024593300789684745,
48
+ "eval_G_intron": 0.000209744205032615,
49
+ "eval_G_nig": 0.0001933390126892979,
50
+ "eval_G_promoter": 0.000290331756895652,
51
+ "eval_G_utr": 0.0002710944576302459,
52
  "eval_avg_bp_per_token": 12799.0,
53
+ "eval_bp_per_token/cds": 40989.5,
54
  "eval_bp_per_token/dig": 12799.0,
55
+ "eval_bp_per_token/exon": 16209.0,
56
+ "eval_bp_per_token/intron": 11585.380952380952,
57
+ "eval_bp_per_token/nig": 11276.77358490566,
58
+ "eval_bp_per_token/promoter": 34494.72727272727,
59
+ "eval_bp_per_token/utr": 7452.25,
60
+ "eval_ppl_cds": 2.889123597693874e+19,
61
+ "eval_ppl_dig": 3.430802770811005e+17,
62
+ "eval_ppl_exon": 4.6340490808362584e+16,
63
+ "eval_ppl_intron": 9.889931203838952e+16,
64
+ "eval_ppl_nig": 3.3020604346559384e+16,
65
+ "eval_ppl_promoter": 7.55131215186589e+19,
66
+ "eval_ppl_utr": 1.4233813097089611e+17,
67
  "step": 10,
68
+ "tokens_trained": 0.01638272
69
  },
70
  {
71
  "comp/rl_weight": 0.03,
72
  "comp/strictness": 0.0,
73
+ "epoch": 0.0028368794326241137,
74
+ "grad_norm": 529.8057250976562,
75
+ "loss": 24.7968,
76
+ "loss_ce": 16.434017181396484,
77
+ "loss_region": 0.09944300353527069,
78
+ "loss_total": 16.53346061706543,
79
  "lr": 4.654030511288038e-05,
80
  "router/selected_tokens_s0": 1.0,
81
  "router/selected_tokens_s1": 1.0,
82
  "step": 20,
83
+ "tokens_trained": 0.03276544
84
  },
85
  {
86
+ "epoch": 0.0028368794326241137,
87
+ "eval_ppl": 6586376.056868001,
88
+ "eval_runtime": 1.7138,
89
  "step": 20,
90
+ "tokens_trained": 0.03276544
91
  },
92
  {
93
+ "epoch": 0.0028368794326241137,
94
  "eval_F": 7.813110399249941e-05,
95
+ "eval_F_cds": 2.439649178448139e-05,
96
  "eval_F_dig": 7.813110399249941e-05,
97
+ "eval_F_exon": 6.169412055031155e-05,
98
+ "eval_F_intron": 8.631567698207511e-05,
99
+ "eval_F_nig": 8.867784676802712e-05,
100
+ "eval_F_promoter": 2.898993785611503e-05,
101
+ "eval_F_utr": 0.00013418766144453016,
102
  "eval_G": 7.813110399249941e-05,
103
+ "eval_G_cds": 2.439649178448139e-05,
104
  "eval_G_dig": 7.813110399249941e-05,
105
+ "eval_G_exon": 6.169412055031155e-05,
106
+ "eval_G_intron": 8.631567698207511e-05,
107
+ "eval_G_nig": 8.867784676802712e-05,
108
+ "eval_G_promoter": 2.898993785611503e-05,
109
+ "eval_G_utr": 0.00013418766144453016,
110
  "eval_avg_bp_per_token": 12799.0,
111
+ "eval_bp_per_token/cds": 40989.5,
112
  "eval_bp_per_token/dig": 12799.0,
113
+ "eval_bp_per_token/exon": 16209.0,
114
+ "eval_bp_per_token/intron": 11585.380952380952,
115
+ "eval_bp_per_token/nig": 11276.77358490566,
116
+ "eval_bp_per_token/promoter": 34494.72727272727,
117
+ "eval_bp_per_token/utr": 7452.25,
118
+ "eval_ppl_cds": 24871529.498565275,
119
+ "eval_ppl_dig": 4569.419833469421,
120
+ "eval_ppl_exon": 10972869.282852469,
121
+ "eval_ppl_intron": 4123822.139661733,
122
+ "eval_ppl_nig": 7623851.6146739,
123
+ "eval_ppl_promoter": 24012533.741014417,
124
+ "eval_ppl_utr": 14099401.427840032,
125
  "step": 20,
126
+ "tokens_trained": 0.03276544
127
  }
128
  ],
129
  "logging_steps": 10,
checkpoint-20/training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:c4dbdc20d0327174126de42adf35e1c0844ec1a2a3097ba550495f97e879180f
3
  size 5969
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:912c85f83733bc75ddebd721c1bb21205af4115ec87e6bbd7790ca686f00482a
3
  size 5969