Upload folder using huggingface_hub

checkpoint-20/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:925e7c3900591222762661453ccd0b69bba92ee0317e75507f9a70ee3f70724c
+oid sha256:b733d59b356122c9ec251414b9f1830f05e11883e5c3d56e7223894fdfb9eb9d
 size 558664971

checkpoint-20/pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a6415cb16517110e0c48fc221ff175d37dfac542b01c1183d83871327aa7a6f3
+oid sha256:ff0c41a557a265632e67741268f9a9812ee7dddccf952ee20a8a47f01ad5ad88
 size 279336283

checkpoint-20/trainer_state.json

@@ -1,6 +1,6 @@
 {
   "best_global_step": 20,
-  "best_metric": 83052.92336765953,
+  "best_metric": 124944.5436483137,
   "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,
@@ -13,21 +13,21 @@
       "comp/rl_weight": 0.03,
       "comp/strictness": 0.0,
       "epoch": 0.00035457220863028757,
-      "grad_norm": 1119.484619140625,
-      "loss": 146.4954,
-      "loss_ce": 55.18280029296875,
-      "loss_region": 0.10638566315174103,
-      "loss_total": 55.2891845703125,
+      "grad_norm": 1130.100830078125,
+      "loss": 146.4512,
+      "loss_ce": 54.19949722290039,
+      "loss_region": 0.10721047222614288,
+      "loss_total": 54.30670928955078,
       "lr": 2.20454076850486e-05,
-      "router/selected_tokens_s0": 22.90625,
+      "router/selected_tokens_s0": 24.15625,
       "router/selected_tokens_s1": 1.0,
       "step": 10,
       "tokens_trained": 0.00409568
     },
     {
       "epoch": 0.00035457220863028757,
-      "eval_ppl": 1.089793926702946e+18,
-      "eval_runtime": 7.4688,
+      "eval_ppl": 7.294625624845928e+17,
+      "eval_runtime": 6.6402,
       "step": 10,
       "tokens_trained": 0.00409568
     },
@@ -41,14 +41,14 @@
       "eval_F_nig": 8.365496566708533e-05,
       "eval_F_promoter": 4.770454407466195e-05,
       "eval_F_utr": 7.9026626119178e-05,
-      "eval_G": 0.00018098983848678323,
-      "eval_G_cds": 0.00017869544990893493,
-      "eval_G_dig": 0.00016462976436961221,
-      "eval_G_exon": 0.0001940150685218265,
-      "eval_G_intron": 0.00017927827939164204,
-      "eval_G_nig": 0.0001736500060544804,
-      "eval_G_promoter": 0.00019476532074800724,
-      "eval_G_utr": 0.00020035056174760066,
+      "eval_G": 0.00016736338285193968,
+      "eval_G_cds": 0.00016425546555209688,
+      "eval_G_dig": 0.000140951360132302,
+      "eval_G_exon": 0.0001619063651334152,
+      "eval_G_intron": 0.00016938910394235382,
+      "eval_G_nig": 0.0001631118140770298,
+      "eval_G_promoter": 0.00016429247622091106,
+      "eval_G_utr": 0.00019291894904831453,
       "eval_avg_bp_per_token": 12799.0,
       "eval_bp_per_token/cds": 21077.722222222223,
       "eval_bp_per_token/dig": 12799.0,
@@ -57,13 +57,13 @@
       "eval_bp_per_token/nig": 11953.86301369863,
       "eval_bp_per_token/promoter": 20962.363636363636,
       "eval_bp_per_token/utr": 12653.962962962964,
-      "eval_ppl_cds": 3.6165174451293454e+20,
-      "eval_ppl_dig": 1.0595141039292915e+17,
-      "eval_ppl_exon": 3.0494664121994173e+18,
-      "eval_ppl_intron": 4.124064470122155e+17,
-      "eval_ppl_nig": 1.7108246674137136e+17,
-      "eval_ppl_promoter": 2.3044177457874593e+20,
-      "eval_ppl_utr": 6.966228632508205e+17,
+      "eval_ppl_cds": 2.311600502337928e+20,
+      "eval_ppl_dig": 9.652397143812314e+16,
+      "eval_ppl_exon": 2.2747493338461386e+18,
+      "eval_ppl_intron": 2.793470782953012e+17,
+      "eval_ppl_nig": 1.1026701972335622e+17,
+      "eval_ppl_promoter": 1.5383074335283397e+20,
+      "eval_ppl_utr": 4.222450573520073e+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": 296.3909606933594,
-      "loss": 26.0585,
-      "loss_ce": 12.013338088989258,
-      "loss_region": 0.09937021881341934,
-      "loss_total": 12.11270809173584,
+      "grad_norm": 299.7831115722656,
+      "loss": 26.1005,
+      "loss_ce": 12.792510032653809,
+      "loss_region": 0.09937700629234314,
+      "loss_total": 12.891886711120605,
       "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": 83052.92336765953,
-      "eval_runtime": 6.5101,
+      "eval_ppl": 124944.5436483137,
+      "eval_runtime": 6.5805,
       "step": 20,
       "tokens_trained": 0.00819136
     },
@@ -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": 190899.46793400217,
-      "eval_ppl_dig": 22914.483406702355,
-      "eval_ppl_exon": 75671.08690466816,
-      "eval_ppl_intron": 69044.63564426819,
-      "eval_ppl_nig": 65731.35949006735,
-      "eval_ppl_promoter": 211928.04291369749,
-      "eval_ppl_utr": 113084.05555305106,
+      "eval_ppl_cds": 287230.0213028405,
+      "eval_ppl_dig": 25004.34707335865,
+      "eval_ppl_exon": 108975.3496962627,
+      "eval_ppl_intron": 104946.6793164646,
+      "eval_ppl_nig": 100265.49289711946,
+      "eval_ppl_promoter": 307025.0147172088,
+      "eval_ppl_utr": 179407.97736471496,
       "step": 20,
       "tokens_trained": 0.00819136
     }

checkpoint-20/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:598e63fc194a844bc1f682ff5be48e182568d89fdb52af1791c3d1ed5721322a
+oid sha256:c4dbdc20d0327174126de42adf35e1c0844ec1a2a3097ba550495f97e879180f
 size 5969