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:
|
| 3 |
size 558664971
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5bc559b8ebcbbf2224ee35f0e7716ff47bb8acc1f62036e9041fa8313c080afb
|
| 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:
|
| 3 |
size 279336283
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5ec335ae107fe4b8d9033d7edfbfa9153f969d01dfe9d625d40290a0ea00a703
|
| 3 |
size 279336283
|
checkpoint-20/trainer_state.json
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_global_step": 20,
|
| 3 |
-
"best_metric":
|
| 4 |
"best_model_checkpoint": "/network/scratch/j/jianan.zhao/DNAFM/output/gencode_human_12.8k_12800/CKPT_DEBUG/checkpoint-20",
|
| 5 |
-
"epoch": 0.
|
| 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.
|
| 16 |
-
"grad_norm":
|
| 17 |
-
"loss":
|
| 18 |
-
"loss_ce": 54.
|
| 19 |
-
"loss_region": 0.
|
| 20 |
-
"loss_total": 54.
|
| 21 |
"lr": 2.20454076850486e-05,
|
| 22 |
-
"router/selected_tokens_s0": 22.
|
| 23 |
"router/selected_tokens_s1": 1.0,
|
| 24 |
"step": 10,
|
| 25 |
-
"tokens_trained": 0.
|
| 26 |
},
|
| 27 |
{
|
| 28 |
-
"epoch": 0.
|
| 29 |
-
"eval_ppl": 1.
|
| 30 |
-
"eval_runtime":
|
| 31 |
"step": 10,
|
| 32 |
-
"tokens_trained": 0.
|
| 33 |
},
|
| 34 |
{
|
| 35 |
-
"epoch": 0.
|
| 36 |
"eval_F": 7.813110399249941e-05,
|
| 37 |
-
"eval_F_cds":
|
| 38 |
"eval_F_dig": 7.813110399249941e-05,
|
| 39 |
-
"eval_F_exon": 6.
|
| 40 |
-
"eval_F_intron": 8.
|
| 41 |
-
"eval_F_nig": 8.
|
| 42 |
-
"eval_F_promoter":
|
| 43 |
-
"eval_F_utr":
|
| 44 |
-
"eval_G": 0.
|
| 45 |
-
"eval_G_cds": 0.
|
| 46 |
-
"eval_G_dig": 0.
|
| 47 |
-
"eval_G_exon": 0.
|
| 48 |
-
"eval_G_intron": 0.
|
| 49 |
-
"eval_G_nig": 0.
|
| 50 |
-
"eval_G_promoter": 0.
|
| 51 |
-
"eval_G_utr": 0.
|
| 52 |
"eval_avg_bp_per_token": 12799.0,
|
| 53 |
-
"eval_bp_per_token/cds":
|
| 54 |
"eval_bp_per_token/dig": 12799.0,
|
| 55 |
-
"eval_bp_per_token/exon":
|
| 56 |
-
"eval_bp_per_token/intron":
|
| 57 |
-
"eval_bp_per_token/nig":
|
| 58 |
-
"eval_bp_per_token/promoter":
|
| 59 |
-
"eval_bp_per_token/utr":
|
| 60 |
-
"eval_ppl_cds":
|
| 61 |
-
"eval_ppl_dig":
|
| 62 |
-
"eval_ppl_exon":
|
| 63 |
-
"eval_ppl_intron":
|
| 64 |
-
"eval_ppl_nig":
|
| 65 |
-
"eval_ppl_promoter":
|
| 66 |
-
"eval_ppl_utr":
|
| 67 |
"step": 10,
|
| 68 |
-
"tokens_trained": 0.
|
| 69 |
},
|
| 70 |
{
|
| 71 |
"comp/rl_weight": 0.03,
|
| 72 |
"comp/strictness": 0.0,
|
| 73 |
-
"epoch": 0.
|
| 74 |
-
"grad_norm":
|
| 75 |
-
"loss":
|
| 76 |
-
"loss_ce":
|
| 77 |
-
"loss_region": 0.
|
| 78 |
-
"loss_total":
|
| 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.
|
| 84 |
},
|
| 85 |
{
|
| 86 |
-
"epoch": 0.
|
| 87 |
-
"eval_ppl":
|
| 88 |
-
"eval_runtime":
|
| 89 |
"step": 20,
|
| 90 |
-
"tokens_trained": 0.
|
| 91 |
},
|
| 92 |
{
|
| 93 |
-
"epoch": 0.
|
| 94 |
"eval_F": 7.813110399249941e-05,
|
| 95 |
-
"eval_F_cds":
|
| 96 |
"eval_F_dig": 7.813110399249941e-05,
|
| 97 |
-
"eval_F_exon": 6.
|
| 98 |
-
"eval_F_intron": 8.
|
| 99 |
-
"eval_F_nig": 8.
|
| 100 |
-
"eval_F_promoter":
|
| 101 |
-
"eval_F_utr":
|
| 102 |
"eval_G": 7.813110399249941e-05,
|
| 103 |
-
"eval_G_cds":
|
| 104 |
"eval_G_dig": 7.813110399249941e-05,
|
| 105 |
-
"eval_G_exon": 6.
|
| 106 |
-
"eval_G_intron": 8.
|
| 107 |
-
"eval_G_nig": 8.
|
| 108 |
-
"eval_G_promoter":
|
| 109 |
-
"eval_G_utr":
|
| 110 |
"eval_avg_bp_per_token": 12799.0,
|
| 111 |
-
"eval_bp_per_token/cds":
|
| 112 |
"eval_bp_per_token/dig": 12799.0,
|
| 113 |
-
"eval_bp_per_token/exon":
|
| 114 |
-
"eval_bp_per_token/intron":
|
| 115 |
-
"eval_bp_per_token/nig":
|
| 116 |
-
"eval_bp_per_token/promoter":
|
| 117 |
-
"eval_bp_per_token/utr":
|
| 118 |
-
"eval_ppl_cds":
|
| 119 |
-
"eval_ppl_dig":
|
| 120 |
-
"eval_ppl_exon":
|
| 121 |
-
"eval_ppl_intron":
|
| 122 |
-
"eval_ppl_nig":
|
| 123 |
-
"eval_ppl_promoter":
|
| 124 |
-
"eval_ppl_utr":
|
| 125 |
"step": 20,
|
| 126 |
-
"tokens_trained": 0.
|
| 127 |
}
|
| 128 |
],
|
| 129 |
"logging_steps": 10,
|
|
|
|
| 1 |
{
|
| 2 |
"best_global_step": 20,
|
| 3 |
+
"best_metric": 261006.7330645563,
|
| 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 |
{
|
| 13 |
"comp/rl_weight": 0.03,
|
| 14 |
"comp/strictness": 0.0,
|
| 15 |
+
"epoch": 0.00035457220863028757,
|
| 16 |
+
"grad_norm": 1104.44580078125,
|
| 17 |
+
"loss": 146.6099,
|
| 18 |
+
"loss_ce": 54.38197326660156,
|
| 19 |
+
"loss_region": 0.10671775788068771,
|
| 20 |
+
"loss_total": 54.48868942260742,
|
| 21 |
"lr": 2.20454076850486e-05,
|
| 22 |
+
"router/selected_tokens_s0": 22.71875,
|
| 23 |
"router/selected_tokens_s1": 1.0,
|
| 24 |
"step": 10,
|
| 25 |
+
"tokens_trained": 0.00409568
|
| 26 |
},
|
| 27 |
{
|
| 28 |
+
"epoch": 0.00035457220863028757,
|
| 29 |
+
"eval_ppl": 1.1411666197284078e+18,
|
| 30 |
+
"eval_runtime": 7.5718,
|
| 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": 0.00018249983824419877,
|
| 45 |
+
"eval_G_cds": 0.0001800133237040688,
|
| 46 |
+
"eval_G_dig": 0.00018303075842513738,
|
| 47 |
+
"eval_G_exon": 0.00019645910961166313,
|
| 48 |
+
"eval_G_intron": 0.00018081985788333596,
|
| 49 |
+
"eval_G_nig": 0.00017549950537569102,
|
| 50 |
+
"eval_G_promoter": 0.00019468167686080247,
|
| 51 |
+
"eval_G_utr": 0.00019923581984270775,
|
| 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": 3.4552045457243275e+20,
|
| 61 |
+
"eval_ppl_dig": 1.0972195397037683e+17,
|
| 62 |
+
"eval_ppl_exon": 3.635291150333688e+18,
|
| 63 |
+
"eval_ppl_intron": 4.403435383603975e+17,
|
| 64 |
+
"eval_ppl_nig": 1.714937073151653e+17,
|
| 65 |
+
"eval_ppl_promoter": 2.442053915928566e+20,
|
| 66 |
+
"eval_ppl_utr": 6.521360367361032e+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": 302.3475036621094,
|
| 75 |
+
"loss": 26.3678,
|
| 76 |
+
"loss_ce": 14.019078254699707,
|
| 77 |
+
"loss_region": 0.09937329590320587,
|
| 78 |
+
"loss_total": 14.118451118469238,
|
| 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": 261006.7330645563,
|
| 88 |
+
"eval_runtime": 6.5924,
|
| 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": 772152.5478942868,
|
| 119 |
+
"eval_ppl_dig": 89170.01956551318,
|
| 120 |
+
"eval_ppl_exon": 265189.5246595459,
|
| 121 |
+
"eval_ppl_intron": 205887.97033035877,
|
| 122 |
+
"eval_ppl_nig": 188452.7328225424,
|
| 123 |
+
"eval_ppl_promoter": 865149.3774534287,
|
| 124 |
+
"eval_ppl_utr": 336103.40127256134,
|
| 125 |
"step": 20,
|
| 126 |
+
"tokens_trained": 0.00819136
|
| 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:
|
| 3 |
size 5969
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:41cc3ab09601671003ffee48a30ae9c578b1fa64cc5e7591626680344f1fa2b0
|
| 3 |
size 5969
|