Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+super_binary_tune_3072_gpu5.log filter=lfs diff=lfs merge=lfs -text
+super_binary_tune_3072_gpu5_v2.log filter=lfs diff=lfs merge=lfs -text

EXC/len2_3072/lr1e-5_wd0.01_wr0.03_ep10_seed42/results/hg38_len2_3072_binary_EXC_lr1e-5_wd0.01_wr0.03_ep10_seed42/eval_results.json

@@ -0,0 +1 @@
+{"eval_loss": 0.6178473234176636, "eval_accuracy": 0.663172133919513, "eval_f1": 0.6496316443912218, "eval_matthews_correlation": 0.29935397381737294, "eval_precision": 0.6491809348546145, "eval_recall": 0.6501746884204495, "eval_runtime": 26.6799, "eval_samples_per_second": 1440.823, "eval_steps_per_second": 11.282, "epoch": 10.0}

EXC/len2_3072/lr1e-5_wd0.01_wr0.03_ep6_seed42/results/hg38_len2_3072_binary_EXC_lr1e-5_wd0.01_wr0.03_ep6_seed42/eval_results.json

@@ -0,0 +1 @@
+{"eval_loss": 0.6081615686416626, "eval_accuracy": 0.672250982024401, "eval_f1": 0.650826607955981, "eval_matthews_correlation": 0.30430116531411755, "eval_precision": 0.6556507960451163, "eval_recall": 0.6487290806798853, "eval_runtime": 21.2705, "eval_samples_per_second": 1807.249, "eval_steps_per_second": 14.151, "epoch": 6.0}

EXC/len2_3072/lr1e-5_wd0.01_wr0.03_ep8_seed42/results/hg38_len2_3072_binary_EXC_lr1e-5_wd0.01_wr0.03_ep8_seed42/eval_results.json

@@ -0,0 +1 @@
+{"eval_loss": 0.6351748704910278, "eval_accuracy": 0.6623657032855544, "eval_f1": 0.6519756439575786, "eval_matthews_correlation": 0.3053418860572037, "eval_precision": 0.6509430797262692, "eval_recall": 0.6544185853867015, "eval_runtime": 22.0743, "eval_samples_per_second": 1741.435, "eval_steps_per_second": 13.636, "epoch": 8.0}

EXC/len2_3072/lr1e-5_wd0.01_wr0.06_ep10_seed42/checkpoint-14418/config.json

@@ -0,0 +1,27 @@
+{
+  "_name_or_path": "/home/n5huang/dna_token/pretrain/models/len2_3072/checkpoint-100000",
+  "architectures": [
+    "BertForSequenceClassification"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "classifier_dropout": null,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "layer_norm_eps": 1e-12,
+  "max_length": 512,
+  "max_position_embeddings": 512,
+  "model_type": "bert",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "pad_token_id": 0,
+  "position_embedding_type": "absolute",
+  "problem_type": "single_label_classification",
+  "torch_dtype": "float32",
+  "transformers_version": "4.35.2",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "vocab_size": 3072
+}

EXC/len2_3072/lr1e-5_wd0.01_wr0.06_ep10_seed42/checkpoint-14418/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f384c42f6b34781334dd0ee6a7b230774f3ad5c085a0c47fb2ae122a6beef593
+size 353632152

EXC/len2_3072/lr1e-5_wd0.01_wr0.06_ep10_seed42/checkpoint-14418/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eeb3767a5e40dd97969021fa20767b361be9c4618ffc4396035809393a5a7bc2
+size 707385995

EXC/len2_3072/lr1e-5_wd0.01_wr0.06_ep10_seed42/checkpoint-14418/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b7a0ca6d9c03ca6cd21dacb7fbc3af34a636617c1d1ce5e8095e429ca7da8fbd
+size 14645

EXC/len2_3072/lr1e-5_wd0.01_wr0.06_ep10_seed42/checkpoint-14418/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:feeb0288e2a148e2562ef3fce3ddc09a85dc5e469890aad408b74813b4638e83
+size 1465

EXC/len2_3072/lr1e-5_wd0.01_wr0.06_ep10_seed42/checkpoint-14418/special_tokens_map.json

@@ -0,0 +1,7 @@
+{
+  "cls_token": "[CLS]",
+  "mask_token": "[MASK]",
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "unk_token": "[UNK]"
+}

EXC/len2_3072/lr1e-5_wd0.01_wr0.06_ep10_seed42/checkpoint-14418/tokenizer_config.json

@@ -0,0 +1,15 @@
+{
+  "added_tokens_decoder": {},
+  "cache_dir": null,
+  "clean_up_tokenization_spaces": true,
+  "cls_token": "[CLS]",
+  "mask_token": "[MASK]",
+  "model_max_length": 200,
+  "pad_token": "[PAD]",
+  "padding_side": "right",
+  "sep_token": "[SEP]",
+  "tokenizer_class": "PreTrainedTokenizerFast",
+  "trust_remote_code": true,
+  "unk_token": "[UNK]",
+  "use_fast": true
+}

EXC/len2_3072/lr1e-5_wd0.01_wr0.06_ep10_seed42/checkpoint-14418/trainer_state.json

@@ -0,0 +1,961 @@
+{
+  "best_metric": 0.6707499651450133,
+  "best_model_checkpoint": "/home/n5huang/dna_token/Finetune-species/super_binary_out_tune_1gpu_3072/EXC/len2_3072/lr1e-5_wd0.01_wr0.06_ep10_seed42/checkpoint-14418",
+  "epoch": 6.0,
+  "eval_steps": 100,
+  "global_step": 14418,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.934812760055479e-07,
+      "loss": 0.6906,
+      "step": 100
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.3869625520110958e-06,
+      "loss": 0.6709,
+      "step": 200
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.0804438280166435e-06,
+      "loss": 0.6667,
+      "step": 300
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.7739251040221917e-06,
+      "loss": 0.6581,
+      "step": 400
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.46740638002774e-06,
+      "loss": 0.6526,
+      "step": 500
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.160887656033287e-06,
+      "loss": 0.654,
+      "step": 600
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.854368932038836e-06,
+      "loss": 0.6472,
+      "step": 700
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 5.547850208044383e-06,
+      "loss": 0.6395,
+      "step": 800
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.24133148404993e-06,
+      "loss": 0.6426,
+      "step": 900
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.93481276005548e-06,
+      "loss": 0.6422,
+      "step": 1000
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 7.628294036061027e-06,
+      "loss": 0.6363,
+      "step": 1100
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.321775312066574e-06,
+      "loss": 0.6389,
+      "step": 1200
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.015256588072124e-06,
+      "loss": 0.6314,
+      "step": 1300
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.708737864077671e-06,
+      "loss": 0.632,
+      "step": 1400
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.974322649194264e-06,
+      "loss": 0.6355,
+      "step": 1500
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.930051354701612e-06,
+      "loss": 0.6311,
+      "step": 1600
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.885780060208962e-06,
+      "loss": 0.6249,
+      "step": 1700
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.84150876571631e-06,
+      "loss": 0.624,
+      "step": 1800
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.797237471223659e-06,
+      "loss": 0.6267,
+      "step": 1900
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.752966176731009e-06,
+      "loss": 0.612,
+      "step": 2000
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.708694882238358e-06,
+      "loss": 0.623,
+      "step": 2100
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.664423587745706e-06,
+      "loss": 0.6229,
+      "step": 2200
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.620152293253056e-06,
+      "loss": 0.6218,
+      "step": 2300
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.575880998760403e-06,
+      "loss": 0.6148,
+      "step": 2400
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.6617413698915221,
+      "eval_f1": 0.625381396050946,
+      "eval_loss": 0.6078959703445435,
+      "eval_matthews_correlation": 0.2657819752253823,
+      "eval_precision": 0.6426361126509434,
+      "eval_recall": 0.6238116649455647,
+      "eval_runtime": 23.2922,
+      "eval_samples_per_second": 1650.378,
+      "eval_steps_per_second": 12.923,
+      "step": 2403
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.531609704267753e-06,
+      "loss": 0.6099,
+      "step": 2500
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.487338409775103e-06,
+      "loss": 0.6078,
+      "step": 2600
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.443067115282452e-06,
+      "loss": 0.6086,
+      "step": 2700
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 9.398795820789802e-06,
+      "loss": 0.6043,
+      "step": 2800
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.35452452629715e-06,
+      "loss": 0.6084,
+      "step": 2900
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 9.310253231804499e-06,
+      "loss": 0.607,
+      "step": 3000
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 9.265981937311847e-06,
+      "loss": 0.6049,
+      "step": 3100
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 9.221710642819196e-06,
+      "loss": 0.5999,
+      "step": 3200
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 9.177439348326546e-06,
+      "loss": 0.6054,
+      "step": 3300
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 9.133168053833896e-06,
+      "loss": 0.5965,
+      "step": 3400
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 9.088896759341243e-06,
+      "loss": 0.6064,
+      "step": 3500
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.044625464848593e-06,
+      "loss": 0.6093,
+      "step": 3600
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.000354170355941e-06,
+      "loss": 0.6044,
+      "step": 3700
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 8.95608287586329e-06,
+      "loss": 0.6022,
+      "step": 3800
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.91181158137064e-06,
+      "loss": 0.6043,
+      "step": 3900
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.86754028687799e-06,
+      "loss": 0.6009,
+      "step": 4000
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.823268992385337e-06,
+      "loss": 0.6057,
+      "step": 4100
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.778997697892687e-06,
+      "loss": 0.5978,
+      "step": 4200
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.734726403400036e-06,
+      "loss": 0.6022,
+      "step": 4300
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.690455108907386e-06,
+      "loss": 0.5945,
+      "step": 4400
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 8.646183814414734e-06,
+      "loss": 0.601,
+      "step": 4500
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.601912519922083e-06,
+      "loss": 0.6051,
+      "step": 4600
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 8.557641225429433e-06,
+      "loss": 0.5986,
+      "step": 4700
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 8.51336993093678e-06,
+      "loss": 0.5895,
+      "step": 4800
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.6756327879087433,
+      "eval_f1": 0.6517439082900512,
+      "eval_loss": 0.5998880863189697,
+      "eval_matthews_correlation": 0.30735273006946745,
+      "eval_precision": 0.6582144798629759,
+      "eval_recall": 0.6492684183567907,
+      "eval_runtime": 23.0561,
+      "eval_samples_per_second": 1667.279,
+      "eval_steps_per_second": 13.055,
+      "step": 4806
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 8.46909863644413e-06,
+      "loss": 0.5826,
+      "step": 4900
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 8.42482734195148e-06,
+      "loss": 0.591,
+      "step": 5000
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 8.38055604745883e-06,
+      "loss": 0.5801,
+      "step": 5100
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 8.336284752966177e-06,
+      "loss": 0.5769,
+      "step": 5200
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 8.292013458473527e-06,
+      "loss": 0.5786,
+      "step": 5300
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 8.247742163980875e-06,
+      "loss": 0.5863,
+      "step": 5400
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 8.203470869488224e-06,
+      "loss": 0.5793,
+      "step": 5500
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 8.159199574995574e-06,
+      "loss": 0.584,
+      "step": 5600
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 8.114928280502923e-06,
+      "loss": 0.5804,
+      "step": 5700
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 8.070656986010271e-06,
+      "loss": 0.5844,
+      "step": 5800
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 8.02638569151762e-06,
+      "loss": 0.5859,
+      "step": 5900
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.982114397024969e-06,
+      "loss": 0.5858,
+      "step": 6000
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.937843102532318e-06,
+      "loss": 0.5807,
+      "step": 6100
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.894014520984595e-06,
+      "loss": 0.5845,
+      "step": 6200
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.849743226491944e-06,
+      "loss": 0.5836,
+      "step": 6300
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 7.805471931999292e-06,
+      "loss": 0.5729,
+      "step": 6400
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 7.761200637506642e-06,
+      "loss": 0.5756,
+      "step": 6500
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 7.71692934301399e-06,
+      "loss": 0.5746,
+      "step": 6600
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 7.672658048521339e-06,
+      "loss": 0.5735,
+      "step": 6700
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.628386754028688e-06,
+      "loss": 0.5764,
+      "step": 6800
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 7.584115459536037e-06,
+      "loss": 0.5845,
+      "step": 6900
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 7.539844165043386e-06,
+      "loss": 0.5833,
+      "step": 7000
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 7.495572870550736e-06,
+      "loss": 0.5826,
+      "step": 7100
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 7.451301576058084e-06,
+      "loss": 0.5809,
+      "step": 7200
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.6780780937020369,
+      "eval_f1": 0.6487373780206568,
+      "eval_loss": 0.5931949615478516,
+      "eval_matthews_correlation": 0.30668433633331155,
+      "eval_precision": 0.661206003802502,
+      "eval_recall": 0.6458619405196494,
+      "eval_runtime": 23.2902,
+      "eval_samples_per_second": 1650.519,
+      "eval_steps_per_second": 12.924,
+      "step": 7209
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 7.407030281565434e-06,
+      "loss": 0.5566,
+      "step": 7300
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 7.362758987072783e-06,
+      "loss": 0.5556,
+      "step": 7400
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 7.318487692580132e-06,
+      "loss": 0.5595,
+      "step": 7500
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 7.27421639808748e-06,
+      "loss": 0.557,
+      "step": 7600
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 7.2299451035948296e-06,
+      "loss": 0.5599,
+      "step": 7700
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 7.185673809102178e-06,
+      "loss": 0.5505,
+      "step": 7800
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 7.141402514609528e-06,
+      "loss": 0.5577,
+      "step": 7900
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 7.0971312201168765e-06,
+      "loss": 0.5608,
+      "step": 8000
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 7.052859925624226e-06,
+      "loss": 0.5561,
+      "step": 8100
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 7.009031344076501e-06,
+      "loss": 0.556,
+      "step": 8200
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 6.9647600495838505e-06,
+      "loss": 0.5593,
+      "step": 8300
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 6.920488755091199e-06,
+      "loss": 0.5581,
+      "step": 8400
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 6.876217460598549e-06,
+      "loss": 0.5599,
+      "step": 8500
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 6.8319461661058975e-06,
+      "loss": 0.5548,
+      "step": 8600
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 6.787674871613247e-06,
+      "loss": 0.5591,
+      "step": 8700
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 6.743403577120595e-06,
+      "loss": 0.5646,
+      "step": 8800
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 6.6991322826279444e-06,
+      "loss": 0.5555,
+      "step": 8900
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 6.654860988135293e-06,
+      "loss": 0.5594,
+      "step": 9000
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 6.610589693642643e-06,
+      "loss": 0.5588,
+      "step": 9100
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 6.566318399149991e-06,
+      "loss": 0.5582,
+      "step": 9200
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 6.522047104657341e-06,
+      "loss": 0.5581,
+      "step": 9300
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 6.47777581016469e-06,
+      "loss": 0.561,
+      "step": 9400
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 6.433504515672039e-06,
+      "loss": 0.5624,
+      "step": 9500
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 6.389233221179387e-06,
+      "loss": 0.5597,
+      "step": 9600
+    },
+    {
+      "epoch": 4.0,
+      "eval_accuracy": 0.6898884004058167,
+      "eval_f1": 0.6598930398850054,
+      "eval_loss": 0.5905744433403015,
+      "eval_matthews_correlation": 0.3312815820267371,
+      "eval_precision": 0.6753522746800809,
+      "eval_recall": 0.656467155602009,
+      "eval_runtime": 22.4224,
+      "eval_samples_per_second": 1714.403,
+      "eval_steps_per_second": 13.424,
+      "step": 9612
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 6.344961926686737e-06,
+      "loss": 0.5417,
+      "step": 9700
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 6.300690632194086e-06,
+      "loss": 0.5321,
+      "step": 9800
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 6.256419337701435e-06,
+      "loss": 0.5311,
+      "step": 9900
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 6.212148043208784e-06,
+      "loss": 0.5333,
+      "step": 10000
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 6.167876748716133e-06,
+      "loss": 0.5291,
+      "step": 10100
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 6.123605454223483e-06,
+      "loss": 0.5294,
+      "step": 10200
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 6.0793341597308305e-06,
+      "loss": 0.5319,
+      "step": 10300
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 6.035505578183106e-06,
+      "loss": 0.5302,
+      "step": 10400
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 5.991234283690456e-06,
+      "loss": 0.5329,
+      "step": 10500
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 5.9469629891978045e-06,
+      "loss": 0.5447,
+      "step": 10600
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 5.902691694705154e-06,
+      "loss": 0.5297,
+      "step": 10700
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 5.858420400212502e-06,
+      "loss": 0.5351,
+      "step": 10800
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 5.8141491057198515e-06,
+      "loss": 0.5394,
+      "step": 10900
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 5.7698778112272e-06,
+      "loss": 0.5325,
+      "step": 11000
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.72560651673455e-06,
+      "loss": 0.5324,
+      "step": 11100
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.681335222241899e-06,
+      "loss": 0.5281,
+      "step": 11200
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 5.637063927749248e-06,
+      "loss": 0.5363,
+      "step": 11300
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 5.5927926332565975e-06,
+      "loss": 0.5241,
+      "step": 11400
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 5.548521338763946e-06,
+      "loss": 0.5294,
+      "step": 11500
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 5.504250044271296e-06,
+      "loss": 0.5324,
+      "step": 11600
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 5.459978749778644e-06,
+      "loss": 0.53,
+      "step": 11700
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 5.415707455285993e-06,
+      "loss": 0.5348,
+      "step": 11800
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 5.371436160793342e-06,
+      "loss": 0.541,
+      "step": 11900
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 5.3271648663006914e-06,
+      "loss": 0.5291,
+      "step": 12000
+    },
+    {
+      "epoch": 5.0,
+      "eval_accuracy": 0.6896542753830546,
+      "eval_f1": 0.6685080615373584,
+      "eval_loss": 0.592842161655426,
+      "eval_matthews_correlation": 0.3396584444714288,
+      "eval_precision": 0.673753091582037,
+      "eval_recall": 0.6659939657048923,
+      "eval_runtime": 22.9039,
+      "eval_samples_per_second": 1678.361,
+      "eval_steps_per_second": 13.142,
+      "step": 12015
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 5.28289357180804e-06,
+      "loss": 0.5016,
+      "step": 12100
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 5.23862227731539e-06,
+      "loss": 0.4907,
+      "step": 12200
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 5.1943509828227375e-06,
+      "loss": 0.4981,
+      "step": 12300
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 5.150079688330087e-06,
+      "loss": 0.5075,
+      "step": 12400
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 5.106251106782363e-06,
+      "loss": 0.5116,
+      "step": 12500
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 5.0619798122897115e-06,
+      "loss": 0.5062,
+      "step": 12600
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 5.017708517797061e-06,
+      "loss": 0.5009,
+      "step": 12700
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 4.97343722330441e-06,
+      "loss": 0.5103,
+      "step": 12800
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.9291659288117585e-06,
+      "loss": 0.5113,
+      "step": 12900
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.884894634319108e-06,
+      "loss": 0.5119,
+      "step": 13000
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 4.840623339826457e-06,
+      "loss": 0.5076,
+      "step": 13100
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 4.7963520453338055e-06,
+      "loss": 0.5095,
+      "step": 13200
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.752080750841155e-06,
+      "loss": 0.5023,
+      "step": 13300
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 4.707809456348504e-06,
+      "loss": 0.5063,
+      "step": 13400
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 4.663538161855853e-06,
+      "loss": 0.5059,
+      "step": 13500
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.619266867363202e-06,
+      "loss": 0.5014,
+      "step": 13600
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 4.574995572870551e-06,
+      "loss": 0.5076,
+      "step": 13700
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 4.5307242783779e-06,
+      "loss": 0.5064,
+      "step": 13800
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 4.48645298388525e-06,
+      "loss": 0.5039,
+      "step": 13900
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 4.4421816893925985e-06,
+      "loss": 0.5061,
+      "step": 14000
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 4.397910394899947e-06,
+      "loss": 0.5012,
+      "step": 14100
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 4.353639100407297e-06,
+      "loss": 0.5125,
+      "step": 14200
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.3093678059146454e-06,
+      "loss": 0.5031,
+      "step": 14300
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 4.265096511421994e-06,
+      "loss": 0.504,
+      "step": 14400
+    },
+    {
+      "epoch": 6.0,
+      "eval_accuracy": 0.6868707890013267,
+      "eval_f1": 0.6707499651450133,
+      "eval_loss": 0.6041229963302612,
+      "eval_matthews_correlation": 0.34167357699550605,
+      "eval_precision": 0.6717097549461897,
+      "eval_recall": 0.6699682601805124,
+      "eval_runtime": 23.2629,
+      "eval_samples_per_second": 1652.461,
+      "eval_steps_per_second": 12.939,
+      "step": 14418
+    }
+  ],
+  "logging_steps": 100,
+  "max_steps": 24030,
+  "num_train_epochs": 10,
+  "save_steps": 100,
+  "total_flos": 1.5455891716687834e+17,
+  "trial_name": null,
+  "trial_params": null
+}

EXC/len2_3072/lr1e-5_wd0.01_wr0.06_ep10_seed42/checkpoint-14418/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:69bf5ebb2f7dc68596fa5b7663ed9a0eed64bcb4042b7f62ec4c374c5c8e84ca
+size 5457

EXC/len2_3072/lr1e-5_wd0.01_wr0.06_ep10_seed42/results/hg38_len2_3072_binary_EXC_lr1e-5_wd0.01_wr0.06_ep10_seed42/eval_results.json

@@ -0,0 +1 @@
+{"eval_loss": 0.6228747963905334, "eval_accuracy": 0.6745402044691865, "eval_f1": 0.6584560741429251, "eval_matthews_correlation": 0.3171264005593038, "eval_precision": 0.659454481751973, "eval_recall": 0.6576769006846301, "eval_runtime": 22.5541, "eval_samples_per_second": 1704.388, "eval_steps_per_second": 13.346, "epoch": 10.0}

EXC/len2_3072/lr1e-5_wd0.01_wr0.06_ep6_seed42/results/hg38_len2_3072_binary_EXC_lr1e-5_wd0.01_wr0.06_ep6_seed42/eval_results.json

@@ -0,0 +1 @@
+{"eval_loss": 0.6058430671691895, "eval_accuracy": 0.6695455373169272, "eval_f1": 0.652633533274055, "eval_matthews_correlation": 0.3056244314511564, "eval_precision": 0.6539471364332141, "eval_recall": 0.6516856618186666, "eval_runtime": 27.4612, "eval_samples_per_second": 1399.829, "eval_steps_per_second": 10.961, "epoch": 6.0}

EXC/len2_3072/lr1e-5_wd0.01_wr0.06_ep8_seed42/results/hg38_len2_3072_binary_EXC_lr1e-5_wd0.01_wr0.06_ep8_seed42/eval_results.json

@@ -0,0 +1 @@
+{"eval_loss": 0.6170939207077026, "eval_accuracy": 0.6655133841471346, "eval_f1": 0.6521925426902411, "eval_matthews_correlation": 0.3044970206569984, "eval_precision": 0.651691755189383, "eval_recall": 0.6528073089292692, "eval_runtime": 21.9351, "eval_samples_per_second": 1752.489, "eval_steps_per_second": 13.722, "epoch": 8.0}

EXC/len2_3072/lr1e-5_wd0.01_wr0.10_ep10_seed42/results/hg38_len2_3072_binary_EXC_lr1e-5_wd0.01_wr0.10_ep10_seed42/eval_results.json

@@ -0,0 +1 @@
+{"eval_loss": 0.6097246408462524, "eval_accuracy": 0.6775578158736766, "eval_f1": 0.6533419889978405, "eval_matthews_correlation": 0.31187318588704666, "eval_precision": 0.6613316063587961, "eval_recall": 0.650721991602536, "eval_runtime": 27.471, "eval_samples_per_second": 1399.33, "eval_steps_per_second": 10.957, "epoch": 10.0}

EXC/len2_3072/lr1e-5_wd0.01_wr0.10_ep6_seed42/results/hg38_len2_3072_binary_EXC_lr1e-5_wd0.01_wr0.10_ep6_seed42/eval_results.json

@@ -0,0 +1 @@
+{"eval_loss": 0.6046963334083557, "eval_accuracy": 0.6697536484482713, "eval_f1": 0.6525406637024938, "eval_matthews_correlation": 0.3055307435290016, "eval_precision": 0.6540498733014011, "eval_recall": 0.6514915806823575, "eval_runtime": 22.5014, "eval_samples_per_second": 1708.38, "eval_steps_per_second": 13.377, "epoch": 6.0}

EXC/len2_3072/lr1e-5_wd0.01_wr0.10_ep8_seed42/results/hg38_len2_3072_binary_EXC_lr1e-5_wd0.01_wr0.10_ep8_seed42/eval_results.json

@@ -0,0 +1 @@
+{"eval_loss": 0.6201969981193542, "eval_accuracy": 0.669259384511329, "eval_f1": 0.6538326577116802, "eval_matthews_correlation": 0.3077225984117871, "eval_precision": 0.6542970380198336, "eval_recall": 0.6534267909295997, "eval_runtime": 21.3455, "eval_samples_per_second": 1800.896, "eval_steps_per_second": 14.101, "epoch": 8.0}

EXC/len2_3072/lr1e-5_wd0.05_wr0.03_ep10_seed42/results/hg38_len2_3072_binary_EXC_lr1e-5_wd0.05_wr0.03_ep10_seed42/eval_results.json

@@ -0,0 +1 @@
+{"eval_loss": 0.626980185508728, "eval_accuracy": 0.6718347597617128, "eval_f1": 0.6568772796679739, "eval_matthews_correlation": 0.31377797157494003, "eval_precision": 0.6571690026059576, "eval_recall": 0.6566094678550054, "eval_runtime": 27.6494, "eval_samples_per_second": 1390.299, "eval_steps_per_second": 10.886, "epoch": 10.0}

EXC/len2_3072/lr1e-5_wd0.05_wr0.03_ep6_seed42/results/hg38_len2_3072_binary_EXC_lr1e-5_wd0.05_wr0.03_ep6_seed42/eval_results.json

@@ -0,0 +1 @@
+{"eval_loss": 0.6019596457481384, "eval_accuracy": 0.6723290236986551, "eval_f1": 0.6535178148818668, "eval_matthews_correlation": 0.3081737231167641, "eval_precision": 0.6562529276040414, "eval_recall": 0.6519508227396431, "eval_runtime": 22.0449, "eval_samples_per_second": 1743.758, "eval_steps_per_second": 13.654, "epoch": 6.0}

EXC/len2_3072/lr1e-5_wd0.05_wr0.03_ep8_seed42/results/hg38_len2_3072_binary_EXC_lr1e-5_wd0.05_wr0.03_ep8_seed42/eval_results.json

@@ -0,0 +1 @@
+{"eval_loss": 0.6227311491966248, "eval_accuracy": 0.6639005228792175, "eval_f1": 0.6497639694777707, "eval_matthews_correlation": 0.29954785220580965, "eval_precision": 0.6495359297275902, "eval_recall": 0.6500123012652759, "eval_runtime": 24.7843, "eval_samples_per_second": 1551.024, "eval_steps_per_second": 12.145, "epoch": 8.0}

EXC/len2_3072/lr1e-5_wd0.05_wr0.06_ep10_seed42/results/hg38_len2_3072_binary_EXC_lr1e-5_wd0.05_wr0.06_ep10_seed42/eval_results.json

@@ -0,0 +1 @@
+{"eval_loss": 0.6587177515029907, "eval_accuracy": 0.658593689029942, "eval_f1": 0.6519040755496552, "eval_matthews_correlation": 0.3096925066030533, "eval_precision": 0.6519677386972387, "eval_recall": 0.6577792916251126, "eval_runtime": 26.7851, "eval_samples_per_second": 1435.163, "eval_steps_per_second": 11.238, "epoch": 10.0}

EXC/len2_3072/lr1e-5_wd0.05_wr0.06_ep6_seed42/results/hg38_len2_3072_binary_EXC_lr1e-5_wd0.05_wr0.06_ep6_seed42/eval_results.json

@@ -0,0 +1 @@
+{"eval_loss": 0.6155278086662292, "eval_accuracy": 0.664706953513176, "eval_f1": 0.6482201274365782, "eval_matthews_correlation": 0.29663342264178255, "eval_precision": 0.6491129323502097, "eval_recall": 0.6475247418874441, "eval_runtime": 32.0159, "eval_samples_per_second": 1200.684, "eval_steps_per_second": 9.402, "epoch": 6.0}

EXC/len2_3072/lr1e-5_wd0.05_wr0.06_ep8_seed42/checkpoint-14418/config.json

@@ -0,0 +1,27 @@
+{
+  "_name_or_path": "/home/n5huang/dna_token/pretrain/models/len2_3072/checkpoint-100000",
+  "architectures": [
+    "BertForSequenceClassification"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "classifier_dropout": null,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "layer_norm_eps": 1e-12,
+  "max_length": 512,
+  "max_position_embeddings": 512,
+  "model_type": "bert",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "pad_token_id": 0,
+  "position_embedding_type": "absolute",
+  "problem_type": "single_label_classification",
+  "torch_dtype": "float32",
+  "transformers_version": "4.35.2",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "vocab_size": 3072
+}

EXC/len2_3072/lr1e-5_wd0.05_wr0.06_ep8_seed42/checkpoint-14418/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd93a748485af39a14f99e10db550d5cf1dd534fcd6d6c12d933065b8518a97c
+size 353632152

EXC/len2_3072/lr1e-5_wd0.05_wr0.06_ep8_seed42/checkpoint-14418/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a3760db4c2799f5cbf558818c0a482adf538433b5c7ae352439ce7e995629845
+size 707385995

EXC/len2_3072/lr1e-5_wd0.05_wr0.06_ep8_seed42/checkpoint-14418/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b7a0ca6d9c03ca6cd21dacb7fbc3af34a636617c1d1ce5e8095e429ca7da8fbd
+size 14645

EXC/len2_3072/lr1e-5_wd0.05_wr0.06_ep8_seed42/checkpoint-14418/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c4c0966dcb7ecdca801991ee2834048a5dc64198cf3bf00b8284813bfdfc713a
+size 1465

EXC/len2_3072/lr1e-5_wd0.05_wr0.06_ep8_seed42/checkpoint-14418/special_tokens_map.json

@@ -0,0 +1,7 @@
+{
+  "cls_token": "[CLS]",
+  "mask_token": "[MASK]",
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "unk_token": "[UNK]"
+}

EXC/len2_3072/lr1e-5_wd0.05_wr0.06_ep8_seed42/checkpoint-14418/tokenizer_config.json

@@ -0,0 +1,15 @@
+{
+  "added_tokens_decoder": {},
+  "cache_dir": null,
+  "clean_up_tokenization_spaces": true,
+  "cls_token": "[CLS]",
+  "mask_token": "[MASK]",
+  "model_max_length": 200,
+  "pad_token": "[PAD]",
+  "padding_side": "right",
+  "sep_token": "[SEP]",
+  "tokenizer_class": "PreTrainedTokenizerFast",
+  "trust_remote_code": true,
+  "unk_token": "[UNK]",
+  "use_fast": true
+}

EXC/len2_3072/lr1e-5_wd0.05_wr0.06_ep8_seed42/checkpoint-14418/trainer_state.json

@@ -0,0 +1,961 @@
+{
+  "best_metric": 0.6679233877734426,
+  "best_model_checkpoint": "/home/n5huang/dna_token/Finetune-species/super_binary_out_tune_1gpu_3072/EXC/len2_3072/lr1e-5_wd0.05_wr0.06_ep8_seed42/checkpoint-14418",
+  "epoch": 6.0,
+  "eval_steps": 100,
+  "global_step": 14418,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.04,
+      "learning_rate": 8.665511265164646e-07,
+      "loss": 0.6918,
+      "step": 100
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.7331022530329292e-06,
+      "loss": 0.6725,
+      "step": 200
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.599653379549394e-06,
+      "loss": 0.6653,
+      "step": 300
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 3.4662045060658585e-06,
+      "loss": 0.6558,
+      "step": 400
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.332755632582323e-06,
+      "loss": 0.6513,
+      "step": 500
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 5.199306759098788e-06,
+      "loss": 0.6549,
+      "step": 600
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 6.065857885615252e-06,
+      "loss": 0.648,
+      "step": 700
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.932409012131717e-06,
+      "loss": 0.6384,
+      "step": 800
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.79896013864818e-06,
+      "loss": 0.6461,
+      "step": 900
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.665511265164645e-06,
+      "loss": 0.6383,
+      "step": 1000
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.532062391681109e-06,
+      "loss": 0.6338,
+      "step": 1100
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.974543442169342e-06,
+      "loss": 0.6406,
+      "step": 1200
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.919203099059215e-06,
+      "loss": 0.6309,
+      "step": 1300
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.863862755949087e-06,
+      "loss": 0.6315,
+      "step": 1400
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.80852241283896e-06,
+      "loss": 0.6359,
+      "step": 1500
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.753182069728832e-06,
+      "loss": 0.6312,
+      "step": 1600
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.697841726618706e-06,
+      "loss": 0.6233,
+      "step": 1700
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.642501383508579e-06,
+      "loss": 0.6223,
+      "step": 1800
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.587161040398452e-06,
+      "loss": 0.6244,
+      "step": 1900
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.531820697288324e-06,
+      "loss": 0.6094,
+      "step": 2000
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.476480354178197e-06,
+      "loss": 0.6207,
+      "step": 2100
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.421140011068069e-06,
+      "loss": 0.6207,
+      "step": 2200
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.365799667957943e-06,
+      "loss": 0.6204,
+      "step": 2300
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.310459324847814e-06,
+      "loss": 0.6141,
+      "step": 2400
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.6602585780806951,
+      "eval_f1": 0.6328390998865477,
+      "eval_loss": 0.6084668040275574,
+      "eval_matthews_correlation": 0.27131168354961854,
+      "eval_precision": 0.640820039047717,
+      "eval_recall": 0.6306810275872483,
+      "eval_runtime": 28.9619,
+      "eval_samples_per_second": 1327.294,
+      "eval_steps_per_second": 10.393,
+      "step": 2403
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.255118981737688e-06,
+      "loss": 0.6073,
+      "step": 2500
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.19977863862756e-06,
+      "loss": 0.6051,
+      "step": 2600
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.144438295517433e-06,
+      "loss": 0.608,
+      "step": 2700
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 9.089097952407306e-06,
+      "loss": 0.6038,
+      "step": 2800
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.033757609297178e-06,
+      "loss": 0.6093,
+      "step": 2900
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.978417266187051e-06,
+      "loss": 0.6088,
+      "step": 3000
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.923076923076925e-06,
+      "loss": 0.6035,
+      "step": 3100
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 8.867736579966796e-06,
+      "loss": 0.5986,
+      "step": 3200
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 8.81239623685667e-06,
+      "loss": 0.6038,
+      "step": 3300
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.757055893746541e-06,
+      "loss": 0.5994,
+      "step": 3400
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.701715550636415e-06,
+      "loss": 0.6055,
+      "step": 3500
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 8.646375207526288e-06,
+      "loss": 0.6069,
+      "step": 3600
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 8.59103486441616e-06,
+      "loss": 0.6029,
+      "step": 3700
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 8.535694521306033e-06,
+      "loss": 0.6009,
+      "step": 3800
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.480354178195905e-06,
+      "loss": 0.6037,
+      "step": 3900
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.425013835085778e-06,
+      "loss": 0.6018,
+      "step": 4000
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.36967349197565e-06,
+      "loss": 0.6034,
+      "step": 4100
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.314333148865523e-06,
+      "loss": 0.5966,
+      "step": 4200
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.258992805755395e-06,
+      "loss": 0.6022,
+      "step": 4300
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.20365246264527e-06,
+      "loss": 0.5941,
+      "step": 4400
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 8.148312119535142e-06,
+      "loss": 0.6019,
+      "step": 4500
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.092971776425015e-06,
+      "loss": 0.6037,
+      "step": 4600
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 8.037631433314887e-06,
+      "loss": 0.5979,
+      "step": 4700
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 7.98229109020476e-06,
+      "loss": 0.59,
+      "step": 4800
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.6743320933378424,
+      "eval_f1": 0.6501833897374352,
+      "eval_loss": 0.5996325016021729,
+      "eval_matthews_correlation": 0.30435005906363133,
+      "eval_precision": 0.6567598119741866,
+      "eval_recall": 0.6477243390469378,
+      "eval_runtime": 28.7984,
+      "eval_samples_per_second": 1334.829,
+      "eval_steps_per_second": 10.452,
+      "step": 4806
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 7.926950747094632e-06,
+      "loss": 0.5815,
+      "step": 4900
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 7.871610403984505e-06,
+      "loss": 0.5899,
+      "step": 5000
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 7.816270060874379e-06,
+      "loss": 0.58,
+      "step": 5100
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 7.76092971776425e-06,
+      "loss": 0.5783,
+      "step": 5200
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 7.705589374654124e-06,
+      "loss": 0.5781,
+      "step": 5300
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 7.650249031543996e-06,
+      "loss": 0.584,
+      "step": 5400
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 7.594908688433868e-06,
+      "loss": 0.5786,
+      "step": 5500
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 7.5395683453237415e-06,
+      "loss": 0.5857,
+      "step": 5600
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 7.484228002213614e-06,
+      "loss": 0.5804,
+      "step": 5700
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.4288876591034875e-06,
+      "loss": 0.5829,
+      "step": 5800
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.37354731599336e-06,
+      "loss": 0.586,
+      "step": 5900
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.3182069728832326e-06,
+      "loss": 0.5876,
+      "step": 6000
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.263420033204206e-06,
+      "loss": 0.5821,
+      "step": 6100
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.208079690094079e-06,
+      "loss": 0.5827,
+      "step": 6200
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.152739346983952e-06,
+      "loss": 0.583,
+      "step": 6300
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 7.097399003873825e-06,
+      "loss": 0.5713,
+      "step": 6400
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 7.042058660763698e-06,
+      "loss": 0.575,
+      "step": 6500
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 6.98671831765357e-06,
+      "loss": 0.5758,
+      "step": 6600
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 6.931377974543443e-06,
+      "loss": 0.5755,
+      "step": 6700
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 6.8760376314333155e-06,
+      "loss": 0.5745,
+      "step": 6800
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.820697288323188e-06,
+      "loss": 0.5851,
+      "step": 6900
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 6.765356945213061e-06,
+      "loss": 0.5831,
+      "step": 7000
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 6.710016602102933e-06,
+      "loss": 0.5813,
+      "step": 7100
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 6.654676258992806e-06,
+      "loss": 0.5767,
+      "step": 7200
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.6805494133867486,
+      "eval_f1": 0.6523568504684876,
+      "eval_loss": 0.5922513008117676,
+      "eval_matthews_correlation": 0.3129693021435863,
+      "eval_precision": 0.6639263959745831,
+      "eval_recall": 0.6493807380774579,
+      "eval_runtime": 40.5043,
+      "eval_samples_per_second": 949.059,
+      "eval_steps_per_second": 7.431,
+      "step": 7209
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 6.599335915882678e-06,
+      "loss": 0.5558,
+      "step": 7300
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 6.543995572772552e-06,
+      "loss": 0.5577,
+      "step": 7400
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 6.488655229662424e-06,
+      "loss": 0.5571,
+      "step": 7500
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 6.4333148865522976e-06,
+      "loss": 0.5563,
+      "step": 7600
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 6.37797454344217e-06,
+      "loss": 0.5623,
+      "step": 7700
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 6.322634200332043e-06,
+      "loss": 0.551,
+      "step": 7800
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 6.267293857221915e-06,
+      "loss": 0.5539,
+      "step": 7900
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 6.2119535141117886e-06,
+      "loss": 0.5598,
+      "step": 8000
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 6.156613171001661e-06,
+      "loss": 0.5575,
+      "step": 8100
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 6.101272827891534e-06,
+      "loss": 0.5549,
+      "step": 8200
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 6.046485888212508e-06,
+      "loss": 0.5634,
+      "step": 8300
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 5.9911455451023805e-06,
+      "loss": 0.5553,
+      "step": 8400
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 5.935805201992253e-06,
+      "loss": 0.5583,
+      "step": 8500
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 5.880464858882126e-06,
+      "loss": 0.5567,
+      "step": 8600
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 5.825124515771998e-06,
+      "loss": 0.5587,
+      "step": 8700
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 5.769784172661871e-06,
+      "loss": 0.5646,
+      "step": 8800
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 5.714443829551743e-06,
+      "loss": 0.5533,
+      "step": 8900
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 5.659103486441616e-06,
+      "loss": 0.557,
+      "step": 9000
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 5.603763143331489e-06,
+      "loss": 0.5551,
+      "step": 9100
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 5.548422800221362e-06,
+      "loss": 0.555,
+      "step": 9200
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 5.493082457111234e-06,
+      "loss": 0.5578,
+      "step": 9300
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 5.437742114001108e-06,
+      "loss": 0.56,
+      "step": 9400
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 5.38240177089098e-06,
+      "loss": 0.5592,
+      "step": 9500
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 5.327061427780853e-06,
+      "loss": 0.5592,
+      "step": 9600
+    },
+    {
+      "epoch": 4.0,
+      "eval_accuracy": 0.687469108503941,
+      "eval_f1": 0.6636032226599546,
+      "eval_loss": 0.591829240322113,
+      "eval_matthews_correlation": 0.3318795450504428,
+      "eval_precision": 0.6713784800486122,
+      "eval_recall": 0.6606736627487391,
+      "eval_runtime": 30.1746,
+      "eval_samples_per_second": 1273.951,
+      "eval_steps_per_second": 9.975,
+      "step": 9612
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 5.271721084670725e-06,
+      "loss": 0.5386,
+      "step": 9700
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 5.216380741560599e-06,
+      "loss": 0.5314,
+      "step": 9800
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 5.161040398450471e-06,
+      "loss": 0.5337,
+      "step": 9900
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 5.105700055340344e-06,
+      "loss": 0.5311,
+      "step": 10000
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 5.050359712230216e-06,
+      "loss": 0.5286,
+      "step": 10100
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 4.995019369120089e-06,
+      "loss": 0.528,
+      "step": 10200
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 4.939679026009961e-06,
+      "loss": 0.5283,
+      "step": 10300
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 4.884338682899835e-06,
+      "loss": 0.5302,
+      "step": 10400
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 4.828998339789707e-06,
+      "loss": 0.5367,
+      "step": 10500
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 4.774211400110681e-06,
+      "loss": 0.5429,
+      "step": 10600
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 4.718871057000553e-06,
+      "loss": 0.5346,
+      "step": 10700
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 4.663530713890427e-06,
+      "loss": 0.5393,
+      "step": 10800
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 4.608190370780299e-06,
+      "loss": 0.5407,
+      "step": 10900
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 4.552850027670172e-06,
+      "loss": 0.5347,
+      "step": 11000
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 4.497509684560044e-06,
+      "loss": 0.5253,
+      "step": 11100
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 4.442169341449918e-06,
+      "loss": 0.5285,
+      "step": 11200
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 4.38682899833979e-06,
+      "loss": 0.5343,
+      "step": 11300
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 4.331488655229663e-06,
+      "loss": 0.5242,
+      "step": 11400
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 4.276148312119535e-06,
+      "loss": 0.5326,
+      "step": 11500
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 4.220807969009409e-06,
+      "loss": 0.5345,
+      "step": 11600
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 4.165467625899281e-06,
+      "loss": 0.5311,
+      "step": 11700
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 4.110127282789154e-06,
+      "loss": 0.5282,
+      "step": 11800
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 4.054786939679026e-06,
+      "loss": 0.5386,
+      "step": 11900
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 3.999446596568899e-06,
+      "loss": 0.5285,
+      "step": 12000
+    },
+    {
+      "epoch": 5.0,
+      "eval_accuracy": 0.685856247236024,
+      "eval_f1": 0.6634388311433181,
+      "eval_loss": 0.6009307503700256,
+      "eval_matthews_correlation": 0.33024419930080245,
+      "eval_precision": 0.6695534611136571,
+      "eval_recall": 0.6608065539557475,
+      "eval_runtime": 29.3151,
+      "eval_samples_per_second": 1311.304,
+      "eval_steps_per_second": 10.268,
+      "step": 12015
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 3.9441062534587715e-06,
+      "loss": 0.5044,
+      "step": 12100
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 3.888765910348645e-06,
+      "loss": 0.4956,
+      "step": 12200
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 3.833425567238517e-06,
+      "loss": 0.5111,
+      "step": 12300
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 3.77808522412839e-06,
+      "loss": 0.5053,
+      "step": 12400
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 3.7227448810182625e-06,
+      "loss": 0.5143,
+      "step": 12500
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 3.667404537908135e-06,
+      "loss": 0.5064,
+      "step": 12600
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 3.612064194798008e-06,
+      "loss": 0.4986,
+      "step": 12700
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 3.5572772551189823e-06,
+      "loss": 0.515,
+      "step": 12800
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 3.501936912008855e-06,
+      "loss": 0.5119,
+      "step": 12900
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 3.4465965688987274e-06,
+      "loss": 0.5117,
+      "step": 13000
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 3.3912562257886e-06,
+      "loss": 0.5092,
+      "step": 13100
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 3.3359158826784725e-06,
+      "loss": 0.5086,
+      "step": 13200
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 3.280575539568346e-06,
+      "loss": 0.5103,
+      "step": 13300
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 3.2252351964582184e-06,
+      "loss": 0.5022,
+      "step": 13400
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 3.169894853348091e-06,
+      "loss": 0.5025,
+      "step": 13500
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 3.114554510237964e-06,
+      "loss": 0.4997,
+      "step": 13600
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 3.0592141671278364e-06,
+      "loss": 0.5135,
+      "step": 13700
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 3.003873824017709e-06,
+      "loss": 0.5126,
+      "step": 13800
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 2.9485334809075815e-06,
+      "loss": 0.5063,
+      "step": 13900
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 2.893193137797455e-06,
+      "loss": 0.509,
+      "step": 14000
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 2.8378527946873275e-06,
+      "loss": 0.5057,
+      "step": 14100
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 2.7825124515772e-06,
+      "loss": 0.5135,
+      "step": 14200
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 2.7271721084670726e-06,
+      "loss": 0.5029,
+      "step": 14300
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 2.6718317653569455e-06,
+      "loss": 0.5077,
+      "step": 14400
+    },
+    {
+      "epoch": 6.0,
+      "eval_accuracy": 0.6845555526651231,
+      "eval_f1": 0.6679233877734426,
+      "eval_loss": 0.6076513528823853,
+      "eval_matthews_correlation": 0.3361089559539716,
+      "eval_precision": 0.6691251450752326,
+      "eval_recall": 0.6669905888657424,
+      "eval_runtime": 29.2066,
+      "eval_samples_per_second": 1316.175,
+      "eval_steps_per_second": 10.306,
+      "step": 14418
+    }
+  ],
+  "logging_steps": 100,
+  "max_steps": 19224,
+  "num_train_epochs": 8,
+  "save_steps": 100,
+  "total_flos": 1.5455891716687834e+17,
+  "trial_name": null,
+  "trial_params": null
+}

EXC/len2_3072/lr1e-5_wd0.05_wr0.06_ep8_seed42/checkpoint-14418/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4f7d486df42091f53295aa47f059d06a4d52d530d26bb2995b6ea56232b1ad31
+size 5457

EXC/len2_3072/lr1e-5_wd0.05_wr0.06_ep8_seed42/results/hg38_len2_3072_binary_EXC_lr1e-5_wd0.05_wr0.06_ep8_seed42/eval_results.json

@@ -0,0 +1 @@
+{"eval_loss": 0.6280951499938965, "eval_accuracy": 0.6684009260945345, "eval_f1": 0.6577154647251351, "eval_matthews_correlation": 0.3165141694343519, "eval_precision": 0.6566466645065003, "eval_recall": 0.6598840609985672, "eval_runtime": 27.2345, "eval_samples_per_second": 1411.482, "eval_steps_per_second": 11.052, "epoch": 8.0}

EXC/len2_3072/lr1e-5_wd0.05_wr0.10_ep10_seed42/results/hg38_len2_3072_binary_EXC_lr1e-5_wd0.05_wr0.10_ep10_seed42/eval_results.json

@@ -0,0 +1 @@
+{"eval_loss": 0.6246327757835388, "eval_accuracy": 0.6734736349210478, "eval_f1": 0.6549222915141233, "eval_matthews_correlation": 0.31089397043936445, "eval_precision": 0.657531505475311, "eval_recall": 0.6533900481746817, "eval_runtime": 18.7099, "eval_samples_per_second": 2054.582, "eval_steps_per_second": 16.088, "epoch": 10.0}

EXC/len2_3072/lr1e-5_wd0.05_wr0.10_ep6_seed42/results/hg38_len2_3072_binary_EXC_lr1e-5_wd0.05_wr0.10_ep6_seed42/eval_results.json

@@ -0,0 +1 @@
+{"eval_loss": 0.616929829120636, "eval_accuracy": 0.675294607320309, "eval_f1": 0.6531504826243963, "eval_matthews_correlation": 0.3096241158109831, "eval_precision": 0.6588842470526675, "eval_recall": 0.6508442386046533, "eval_runtime": 21.6847, "eval_samples_per_second": 1772.727, "eval_steps_per_second": 13.881, "epoch": 6.0}

EXC/len2_3072/lr1e-5_wd0.05_wr0.10_ep8_seed42/results/hg38_len2_3072_binary_EXC_lr1e-5_wd0.05_wr0.10_ep8_seed42/eval_results.json

@@ -0,0 +1 @@
+{"eval_loss": 0.6200296878814697, "eval_accuracy": 0.6700137873624515, "eval_f1": 0.653167576553823, "eval_matthews_correlation": 0.30668119297168356, "eval_precision": 0.654459324144854, "eval_recall": 0.6522299716175286, "eval_runtime": 25.1259, "eval_samples_per_second": 1529.934, "eval_steps_per_second": 11.98, "epoch": 8.0}

EXC/len2_3072/lr1e-5_wd0.0_wr0.03_ep10_seed42/results/hg38_len2_3072_binary_EXC_lr1e-5_wd0.0_wr0.03_ep10_seed42/eval_results.json

@@ -0,0 +1 @@
+{"eval_loss": 0.6209965944290161, "eval_accuracy": 0.6655133841471346, "eval_f1": 0.6543210909015286, "eval_matthews_correlation": 0.3094853204585996, "eval_precision": 0.653322990349209, "eval_recall": 0.6561754753171891, "eval_runtime": 26.6993, "eval_samples_per_second": 1439.773, "eval_steps_per_second": 11.274, "epoch": 10.0}

EXC/len2_3072/lr1e-5_wd0.0_wr0.03_ep6_seed42/results/hg38_len2_3072_binary_EXC_lr1e-5_wd0.0_wr0.03_ep6_seed42/eval_results.json

@@ -0,0 +1 @@
+{"eval_loss": 0.6022893786430359, "eval_accuracy": 0.6749043989490388, "eval_f1": 0.6551795994582454, "eval_matthews_correlation": 0.3120603827002284, "eval_precision": 0.6587866517628707, "eval_recall": 0.6533215817731978, "eval_runtime": 28.1053, "eval_samples_per_second": 1367.748, "eval_steps_per_second": 10.71, "epoch": 6.0}

EXC/len2_3072/lr1e-5_wd0.0_wr0.03_ep8_seed42/results/hg38_len2_3072_binary_EXC_lr1e-5_wd0.0_wr0.03_ep8_seed42/eval_results.json

@@ -0,0 +1 @@
+{"eval_loss": 0.6208837032318115, "eval_accuracy": 0.6688691761400588, "eval_f1": 0.6550503643835527, "eval_matthews_correlation": 0.31012957267721847, "eval_precision": 0.6547706844178358, "eval_recall": 0.6553594471245199, "eval_runtime": 22.0837, "eval_samples_per_second": 1740.693, "eval_steps_per_second": 13.63, "epoch": 8.0}

EXC/len2_3072/lr1e-5_wd0.0_wr0.06_ep10_seed42/results/hg38_len2_3072_binary_EXC_lr1e-5_wd0.0_wr0.06_ep10_seed42/eval_results.json

@@ -0,0 +1 @@
+{"eval_loss": 0.6213700175285339, "eval_accuracy": 0.6717046903046227, "eval_f1": 0.6566279782150553, "eval_matthews_correlation": 0.3132888196567533, "eval_precision": 0.6569769975295723, "eval_recall": 0.6563125267818802, "eval_runtime": 27.6944, "eval_samples_per_second": 1388.042, "eval_steps_per_second": 10.869, "epoch": 10.0}

EXC/len2_3072/lr1e-5_wd0.0_wr0.06_ep6_seed42/results/hg38_len2_3072_binary_EXC_lr1e-5_wd0.0_wr0.06_ep6_seed42/eval_results.json

@@ -0,0 +1 @@
+{"eval_loss": 0.6122373938560486, "eval_accuracy": 0.6657214952784787, "eval_f1": 0.648601060640406, "eval_matthews_correlation": 0.29756061221586916, "eval_precision": 0.6498911478035265, "eval_recall": 0.6476776968482854, "eval_runtime": 21.3164, "eval_samples_per_second": 1803.357, "eval_steps_per_second": 14.121, "epoch": 6.0}

EXC/len2_3072/lr1e-5_wd0.0_wr0.06_ep8_seed42/results/hg38_len2_3072_binary_EXC_lr1e-5_wd0.0_wr0.06_ep8_seed42/eval_results.json

@@ -0,0 +1 @@
+{"eval_loss": 0.6460663676261902, "eval_accuracy": 0.6640305923363076, "eval_f1": 0.6556355886523308, "eval_matthews_correlation": 0.3144874196534266, "eval_precision": 0.6548121803518537, "eval_recall": 0.65971342967893, "eval_runtime": 22.4264, "eval_samples_per_second": 1714.095, "eval_steps_per_second": 13.422, "epoch": 8.0}

EXC/len2_3072/lr1e-5_wd0.0_wr0.10_ep10_seed42/checkpoint-14418/config.json

@@ -0,0 +1,27 @@
+{
+  "_name_or_path": "/home/n5huang/dna_token/pretrain/models/len2_3072/checkpoint-100000",
+  "architectures": [
+    "BertForSequenceClassification"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "classifier_dropout": null,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "layer_norm_eps": 1e-12,
+  "max_length": 512,
+  "max_position_embeddings": 512,
+  "model_type": "bert",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "pad_token_id": 0,
+  "position_embedding_type": "absolute",
+  "problem_type": "single_label_classification",
+  "torch_dtype": "float32",
+  "transformers_version": "4.35.2",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "vocab_size": 3072
+}

EXC/len2_3072/lr1e-5_wd0.0_wr0.10_ep10_seed42/checkpoint-14418/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:107978cdd1ff99264cc91e02f55d1b26a97a0cbe14a37073e58c1573885b5235
+size 353632152

EXC/len2_3072/lr1e-5_wd0.0_wr0.10_ep10_seed42/checkpoint-14418/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b92e16017a2be993322fe01203028fa02cc5a2142014a67dd3173c6aa46bd93f
+size 707385995

EXC/len2_3072/lr1e-5_wd0.0_wr0.10_ep10_seed42/checkpoint-14418/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b7a0ca6d9c03ca6cd21dacb7fbc3af34a636617c1d1ce5e8095e429ca7da8fbd
+size 14645

EXC/len2_3072/lr1e-5_wd0.0_wr0.10_ep10_seed42/checkpoint-14418/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19738f568d14a3d9a947df48b834704669e7d57aa128c4797fd740aed2a9d388
+size 1465