Upload folder using huggingface_hub

checkpoint-20/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1d76997c5c7ec0559b0a17d06e5b50b2feac552b8961ba2f07364561f6d547ef
+oid sha256:81b058994c7bb01c1974d22357374a9fcaa0183d790223af50af90811785860c
 size 558664971

checkpoint-20/pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:129016d30aa705b92e6102d79906622950127aaa87e5a4c618c7cbb4f6ee188f
+oid sha256:816ae2edf36355d926b630d48a4a45b67f1986fecc140c4dbd67b92e8acb7e68
 size 279336283

checkpoint-20/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:52e7a1fb60ae8bfda35fbaf20a1fa4b5ee1f711ba533efbed972766f18d6fbe2
-size 16359
+oid sha256:b52f36f0d65f68e848f95af11c72cdc4e8a1e2cb2e5a2889df74b6af1f5f8006
+size 14709

checkpoint-20/trainer_state.json

@@ -1,7 +1,7 @@
 {
   "best_global_step": 20,
-  "best_metric": 15562359.077386793,
-  "best_model_checkpoint": "/gpfs/scratch/guoh/DNAFM/output/gencode_human_12.8k_12800/CKPT_DEBUG/checkpoint-20",
+  "best_metric": 148816.18235103838,
+  "best_model_checkpoint": "/network/scratch/j/jianan.zhao/DNAFM/output/gencode_human_12.8k_12800/CKPT_DEBUG/checkpoint-20",
   "epoch": 0.0007091444172605751,
   "eval_steps": 10,
   "global_step": 20,
@@ -13,57 +13,57 @@
       "comp/rl_weight": 0.03,
       "comp/strictness": 0.0,
       "epoch": 0.00035457220863028757,
-      "grad_norm": 1504.030029296875,
-      "loss": 87.8092,
-      "loss_ce": 59.68216323852539,
-      "loss_region": 0.0935576930642128,
-      "loss_total": 59.77572250366211,
+      "grad_norm": 1111.61279296875,
+      "loss": 146.1968,
+      "loss_ce": 54.305824279785156,
+      "loss_region": 0.10640381276607513,
+      "loss_total": 54.412227630615234,
       "lr": 2.20454076850486e-05,
-      "router/selected_tokens_s0": 793.03125,
-      "router/selected_tokens_s1": 148.84375,
+      "router/selected_tokens_s0": 21.6875,
+      "router/selected_tokens_s1": 1.0,
       "step": 10,
       "tokens_trained": 0.00409568
     },
     {
       "epoch": 0.00035457220863028757,
-      "eval_ppl": 9.153268951196861e+22,
-      "eval_runtime": 4.9797,
+      "eval_ppl": 7.424721446549467e+17,
+      "eval_runtime": 7.6118,
       "step": 10,
       "tokens_trained": 0.00409568
     },
     {
       "epoch": 0.00035457220863028757,
-      "eval_F": 0.00177827022776477,
-      "eval_F_cds": 0.003658417655291659,
-      "eval_F_dig": 0.0012045211865510327,
-      "eval_F_exon": 0.0016746678575415876,
-      "eval_F_intron": 0.001447564894073858,
-      "eval_F_nig": 0.0013862277187481856,
-      "eval_F_promoter": 0.0033154658131890054,
-      "eval_F_utr": 0.0016332169397963454,
-      "eval_G": 0.011702613646642867,
-      "eval_G_cds": 0.023833423264492,
-      "eval_G_dig": 0.009547610734562596,
-      "eval_G_exon": 0.01037959719586171,
-      "eval_G_intron": 0.008864032313383523,
-      "eval_G_nig": 0.008525809138569675,
-      "eval_G_promoter": 0.02530388628896865,
-      "eval_G_utr": 0.010295956315471365,
-      "eval_avg_bp_per_token": 562.3442288953848,
-      "eval_bp_per_token/cds": 273.34221902017293,
-      "eval_bp_per_token/dig": 830.2054054054054,
-      "eval_bp_per_token/exon": 597.1333333333333,
-      "eval_bp_per_token/intron": 690.8153161864243,
-      "eval_bp_per_token/nig": 721.3821989528795,
-      "eval_bp_per_token/promoter": 301.61674296926094,
-      "eval_bp_per_token/utr": 612.2885304659499,
-      "eval_ppl_cds": 6.414658338854343e+21,
-      "eval_ppl_dig": 4.0669318513627916e+21,
-      "eval_ppl_exon": 2.4904150523692485e+23,
-      "eval_ppl_intron": 1.535523404049312e+23,
-      "eval_ppl_nig": 7.19246164907062e+23,
-      "eval_ppl_promoter": 1.0487349115377585e+21,
-      "eval_ppl_utr": 3.14597976396868e+23,
+      "eval_F": 7.813110399249941e-05,
+      "eval_F_cds": 4.744345662481978e-05,
+      "eval_F_dig": 7.813110399249941e-05,
+      "eval_F_exon": 6.977782739756614e-05,
+      "eval_F_intron": 8.580132276805985e-05,
+      "eval_F_nig": 8.365496566708533e-05,
+      "eval_F_promoter": 4.770454407466195e-05,
+      "eval_F_utr": 7.9026626119178e-05,
+      "eval_G": 0.00015868149956862495,
+      "eval_G_cds": 0.00015302779857089767,
+      "eval_G_dig": 0.00013616989934109436,
+      "eval_G_exon": 0.0001662856506549812,
+      "eval_G_intron": 0.00016059590085930095,
+      "eval_G_nig": 0.00015651285421575188,
+      "eval_G_promoter": 0.00015211858489077828,
+      "eval_G_utr": 0.00018231460938894855,
+      "eval_avg_bp_per_token": 12799.0,
+      "eval_bp_per_token/cds": 21077.722222222223,
+      "eval_bp_per_token/dig": 12799.0,
+      "eval_bp_per_token/exon": 14331.2,
+      "eval_bp_per_token/intron": 11654.831973898858,
+      "eval_bp_per_token/nig": 11953.86301369863,
+      "eval_bp_per_token/promoter": 20962.363636363636,
+      "eval_bp_per_token/utr": 12653.962962962964,
+      "eval_ppl_cds": 2.4835679876749204e+20,
+      "eval_ppl_dig": 9.028849885879438e+16,
+      "eval_ppl_exon": 2.2687957801872694e+18,
+      "eval_ppl_intron": 2.8188580486001277e+17,
+      "eval_ppl_nig": 1.1219396355209006e+17,
+      "eval_ppl_promoter": 1.6162428794991283e+20,
+      "eval_ppl_utr": 4.3084586855038605e+17,
       "step": 10,
       "tokens_trained": 0.00409568
     },
@@ -71,11 +71,11 @@
       "comp/rl_weight": 0.03,
       "comp/strictness": 0.0,
       "epoch": 0.0007091444172605751,
-      "grad_norm": 566.407958984375,
-      "loss": 41.7086,
-      "loss_ce": 21.22186279296875,
-      "loss_region": 0.10101450234651566,
-      "loss_total": 21.322877883911133,
+      "grad_norm": 330.2630310058594,
+      "loss": 25.8642,
+      "loss_ce": 12.18574047088623,
+      "loss_region": 0.09936444461345673,
+      "loss_total": 12.285104751586914,
       "lr": 4.654030511288038e-05,
       "router/selected_tokens_s0": 1.0,
       "router/selected_tokens_s1": 1.0,
@@ -84,8 +84,8 @@
     },
     {
       "epoch": 0.0007091444172605751,
-      "eval_ppl": 15562359.077386793,
-      "eval_runtime": 4.5207,
+      "eval_ppl": 148816.18235103838,
+      "eval_runtime": 7.2278,
       "step": 20,
       "tokens_trained": 0.00819136
     },
@@ -99,9 +99,9 @@
       "eval_F_nig": 8.365496566708533e-05,
       "eval_F_promoter": 4.770454407466195e-05,
       "eval_F_utr": 7.9026626119178e-05,
-      "eval_G": 7.816417754056831e-05,
+      "eval_G": 7.813110399249941e-05,
       "eval_G_cds": 4.744345662481978e-05,
-      "eval_G_dig": 8.086518396619527e-05,
+      "eval_G_dig": 7.813110399249941e-05,
       "eval_G_exon": 6.977782739756614e-05,
       "eval_G_intron": 8.580132276805985e-05,
       "eval_G_nig": 8.365496566708533e-05,
@@ -115,13 +115,13 @@
       "eval_bp_per_token/nig": 11953.86301369863,
       "eval_bp_per_token/promoter": 20962.363636363636,
       "eval_bp_per_token/utr": 12653.962962962964,
-      "eval_ppl_cds": 5593566.178288604,
-      "eval_ppl_dig": 35486712.16234024,
-      "eval_ppl_exon": 17259683.491786722,
-      "eval_ppl_intron": 17053510.68256213,
-      "eval_ppl_nig": 19611828.658571318,
-      "eval_ppl_promoter": 9113844.845808096,
-      "eval_ppl_utr": 14151555.844300926,
+      "eval_ppl_cds": 395673.42673693807,
+      "eval_ppl_dig": 31301.43244868628,
+      "eval_ppl_exon": 140758.6210649053,
+      "eval_ppl_intron": 122319.47964181146,
+      "eval_ppl_nig": 111897.23080114595,
+      "eval_ppl_promoter": 421597.51324185927,
+      "eval_ppl_utr": 202983.65571501417,
       "step": 20,
       "tokens_trained": 0.00819136
     }

checkpoint-20/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:faf9107e8232f8c02d0de570892b7fdd717781cc6eb0d429077eedff87e08462
+oid sha256:09a783c7f9bdabaac56f2ebff5b9206706ee40f8914b7eb767a104570a9abfac
 size 5969