andyjzhao commited on
Commit
237eea6
·
verified ·
1 Parent(s): fcaef65

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 sha256:0deb9077d610bc6c6c3e247c7af784e7ca623d5cc46ce437b0fb27bdad27191e
3
  size 558664971
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6a07fd6ba116375cef4a07ea3297665705fff56cb308fba302f4ed3303a55077
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:e5fb880ce8ce6d6a3efccf929c14283e1399abfde74c8ba9c9a5cbad8426d40c
3
  size 279336283
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f9aa72091fd08cce6bde598dcee41240f0f985c1da6f41ff69c36b0e054f05bc
3
  size 279336283
checkpoint-20/trainer_state.json CHANGED
@@ -1,8 +1,8 @@
1
  {
2
  "best_global_step": 20,
3
- "best_metric": 68320.79411816206,
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": 1132.7886962890625,
17
- "loss": 146.4728,
18
- "loss_ce": 54.09164047241211,
19
- "loss_region": 0.10727831721305847,
20
- "loss_total": 54.198917388916016,
21
  "lr": 2.20454076850486e-05,
22
- "router/selected_tokens_s0": 23.78125,
23
  "router/selected_tokens_s1": 1.0,
24
  "step": 10,
25
- "tokens_trained": 0.00409568
26
  },
27
  {
28
- "epoch": 0.00035457220863028757,
29
- "eval_ppl": 1.0186400757990682e+18,
30
- "eval_runtime": 7.2215,
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": 0.00017417084202725643,
45
- "eval_G_cds": 0.0001794316059741855,
46
- "eval_G_dig": 0.0001354323335807485,
47
- "eval_G_exon": 0.0001921070810539243,
48
- "eval_G_intron": 0.00017461964504566645,
49
- "eval_G_nig": 0.00016813731929190464,
50
- "eval_G_promoter": 0.00017862840321777992,
51
- "eval_G_utr": 0.0001872366236459373,
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": 3.3014647482544744e+20,
61
- "eval_ppl_dig": 1.1622268373935285e+17,
62
- "eval_ppl_exon": 3.2522437808130043e+18,
63
- "eval_ppl_intron": 3.9138014555579206e+17,
64
- "eval_ppl_nig": 1.5178663639690314e+17,
65
- "eval_ppl_promoter": 2.1860490407785362e+20,
66
- "eval_ppl_utr": 5.7390657388765485e+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": 303.0186462402344,
75
- "loss": 25.9086,
76
- "loss_ce": 12.05529499053955,
77
- "loss_region": 0.0993824377655983,
78
- "loss_total": 12.154677391052246,
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": 68320.79411816206,
88
- "eval_runtime": 7.246,
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": 154033.63476438122,
119
- "eval_ppl_dig": 15444.189120194795,
120
- "eval_ppl_exon": 60520.166471376135,
121
- "eval_ppl_intron": 57146.884551315,
122
- "eval_ppl_nig": 54740.4715093694,
123
- "eval_ppl_promoter": 169717.99548180104,
124
- "eval_ppl_utr": 97889.76239710047,
125
  "step": 20,
126
- "tokens_trained": 0.00819136
127
  }
128
  ],
129
  "logging_steps": 10,
 
1
  {
2
  "best_global_step": 20,
3
+ "best_metric": 1207329.5664974535,
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": 917.1181030273438,
17
+ "loss": 144.7748,
18
+ "loss_ce": 54.116615295410156,
19
+ "loss_region": 0.10434026271104813,
20
+ "loss_total": 54.22095489501953,
21
  "lr": 2.20454076850486e-05,
22
+ "router/selected_tokens_s0": 22.84375,
23
  "router/selected_tokens_s1": 1.0,
24
  "step": 10,
25
+ "tokens_trained": 0.01638272
26
  },
27
  {
28
+ "epoch": 0.0014184397163120568,
29
+ "eval_ppl": 2.1194747907582294e+17,
30
+ "eval_runtime": 1.7718,
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": 0.00022184538364639257,
45
+ "eval_G_cds": 0.0001901449227851035,
46
+ "eval_G_dig": 0.00015412581061020393,
47
+ "eval_G_exon": 0.000246234248719847,
48
+ "eval_G_intron": 0.00021249317768522492,
49
+ "eval_G_nig": 0.00019619516090846272,
50
+ "eval_G_promoter": 0.00030238687665044985,
51
+ "eval_G_utr": 0.0002784000847059613,
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": 3.730425344687072e+19,
61
+ "eval_ppl_dig": 4.017993685449341e+17,
62
+ "eval_ppl_exon": 6.054858771331785e+16,
63
+ "eval_ppl_intron": 1.2271560232082048e+17,
64
+ "eval_ppl_nig": 3.849790568976093e+16,
65
+ "eval_ppl_promoter": 9.761527176325515e+19,
66
+ "eval_ppl_utr": 1.747612158170368e+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": 518.6296997070312,
75
+ "loss": 24.6671,
76
+ "loss_ce": 15.385003089904785,
77
+ "loss_region": 0.09944470971822739,
78
+ "loss_total": 15.484447479248047,
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": 1207329.5664974535,
88
+ "eval_runtime": 1.7227,
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": 4840595.678625624,
119
+ "eval_ppl_dig": 1044.7725056220102,
120
+ "eval_ppl_exon": 2136300.4799613026,
121
+ "eval_ppl_intron": 763217.2143647178,
122
+ "eval_ppl_nig": 1411875.7684863594,
123
+ "eval_ppl_promoter": 4137739.958306599,
124
+ "eval_ppl_utr": 2728184.3584358147,
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 sha256:27e50a78a0ad195455e56f8e4877f004f4700345080b847afe4f3e05a3937021
3
  size 5969
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a7a44597d0ca5c121252f255044e4012aab2d8b6f4e4e8d122c2bec3966364fd
3
  size 5969