Upload folder using huggingface_hub

README.md

@@ -13,6 +13,43 @@ tags:
 
 Multi-Task Learning peptide classifier covering 22 binary peptide-activity tasks. Built on a frozen ESM-2 (650M) backbone with a parallel Transformer + CNN feature extractor and per-task heads, following a PDeepPP-inspired design.
 
+## Held-out Test Set Performance (Averaged across 22 tasks)
+
+| Metric | Value |
+|---|---|
+| Accuracy | 87.37% |
+| F1       | 84.80% |
+| AUC      | 92.82% |
+| MCC      | 73.42% |
+
+Best Val Avg F1 (used for checkpoint selection): 83.43%
+
+## Per-Task Test Metrics
+
+| Task | ACC | F1 | AUC | MCC |
+|---|---|---|---|---|
+| AntiMRSA | 0.9899 | 0.9667 | 0.9968 | 0.9607 |
+| Anticancer | 0.7267 | 0.7330 | 0.8046 | 0.4540 |
+| ACE_inhibitory | 0.7798 | 0.7901 | 0.8636 | 0.5623 |
+| Antioxidant | 0.7758 | 0.7658 | 0.8393 | 0.5508 |
+| Bitter | 0.8750 | 0.8689 | 0.9609 | 0.7533 |
+| Antimalarial | 0.9736 | 0.7600 | 0.9260 | 0.7523 |
+| Antimicrobial | 0.9700 | 0.9486 | 0.9824 | 0.9275 |
+| Signal_peptide | 0.9919 | 0.9920 | 0.9985 | 0.9839 |
+| Antifungal | 0.9488 | 0.9479 | 0.9851 | 0.8983 |
+| Antimalarial_alt | 0.9755 | 0.9231 | 0.9976 | 0.9124 |
+| Anticancer_alt | 0.9278 | 0.9275 | 0.9676 | 0.8557 |
+| Anti_parasitic | 0.6957 | 0.5882 | 0.8299 | 0.4587 |
+| Umami | 0.8202 | 0.6923 | 0.9245 | 0.5697 |
+| Quorum_sensing | 0.8750 | 0.8718 | 0.9650 | 0.7509 |
+| Antibacterial | 0.9478 | 0.9466 | 0.9739 | 0.8965 |
+| NeuroPred | 0.8876 | 0.8875 | 0.9478 | 0.7753 |
+| Toxicity | 0.9310 | 0.9254 | 0.9623 | 0.8613 |
+| Antiviral | 0.8483 | 0.8432 | 0.9125 | 0.6981 |
+| DPPIV_inhibitory | 0.8346 | 0.8254 | 0.9420 | 0.6729 |
+| BBP | 0.8158 | 0.7879 | 0.9224 | 0.6547 |
+| TTCA | 0.7563 | 0.8154 | 0.7891 | 0.4688 |
+
 ## Architecture
 
 - **Shared encoder**: frozen ESM-2 (`facebook/esm2_t33_650M_UR50D`, 650M params) + learnable base embedding, mixed at `esm_ratio=0.9`
@@ -57,7 +94,7 @@ from transformers import EsmTokenizer
 
 from mtl_peptide_classifier import MTLPeptideClassifier, get_all_peptide_tasks
 
-REPO = "tuankg1028/MTL-Peptide-Classifier"
+REPO = "minhquoc95/MTL-PepPred"
 checkpoint_dir = "MTL-Peptide-Classifier"
 os.makedirs(checkpoint_dir, exist_ok=True)
 
@@ -116,7 +153,7 @@ with torch.no_grad():
 - `shared_backbone.pt` — base embedding, Transformer, CNN, LayerNorm
 - `ablation_config.json` — architecture configuration for reproducibility
 - `test_results.json` — held-out test metrics (per task + averages)
-- `mtl_peptide_classifier.py` — model code.
+- `mtl_peptide_classifier.py` — model code
 
 ## Requirements
 

heads.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ee5355b88708840f0e12e0ad3b542c2c16220b9291ee0a25a8b1269a88c00fb9
-size 60650083
+oid sha256:a56fe107476b4688385df6b23ccf3da067a86a6169be1d2d9fd22586939810b4
+size 57893243

mtl_peptide_classifier.py

@@ -378,7 +378,6 @@ def get_all_peptide_tasks(data_dir: str) -> Dict[str, Dict]:
         "15__Antifungal_AF": "Antifungal",
         "16__AV_Antiviral": "Antiviral",
         "17__Toxicity_2021_Dataset": "Toxicity",
-        "18__Anti_inflammatory_peptides": "Anti_inflammatory",
         "19__Signal_peptides": "Signal_peptide",
         "21__Antioxidant_FRS": "Antioxidant"
     }

shared_backbone.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5179a64d327dd091ce235f20a46dd79bdd354f188bd5425c110d676e5839069f
+oid sha256:59df1051319794dc663d482b1453782f9eef9a02f6014af2a6388d13c50f84c9
 size 360917916

test_results.json

@@ -1,185 +1,177 @@
 {
-  "best_val_avg_f1": 0.8556169559583615,
-  "test_avg_acc": 0.8663440066337095,
-  "test_avg_f1": 0.8498628286458877,
-  "test_avg_auc": 0.9346601239526039,
-  "test_avg_mcc": 0.7267415747182607,
+  "best_val_avg_f1": 0.8342965624786137,
+  "test_avg_acc": 0.8736814868145825,
+  "test_avg_f1": 0.8479511766276302,
+  "test_avg_auc": 0.9281912883424943,
+  "test_avg_mcc": 0.7341971117217908,
   "test_metrics": {
     "AntiMRSA": {
       "accuracy": 0.9899497487437185,
       "precision": 0.9666666666666667,
       "recall": 0.9666666666666667,
       "f1": 0.9666666666666667,
-      "auc": 0.9970414201183432,
+      "auc": 0.9968441814595661,
       "mcc": 0.960749506903353
     },
     "Anticancer": {
-      "accuracy": 0.7034883720930233,
-      "precision": 0.6650943396226415,
-      "recall": 0.8197674418604651,
-      "f1": 0.734375,
-      "auc": 0.800736884802596,
-      "mcc": 0.4184495873764591
+      "accuracy": 0.7267441860465116,
+      "precision": 0.7166666666666667,
+      "recall": 0.75,
+      "f1": 0.7329545454545454,
+      "auc": 0.8045903190914008,
+      "mcc": 0.45397969225022583
     },
     "ACE_inhibitory": {
-      "accuracy": 0.6800518134715026,
-      "precision": 0.6238859180035651,
-      "recall": 0.9067357512953368,
-      "f1": 0.7391763463569165,
-      "auc": 0.8355459206958575,
-      "mcc": 0.40401002524391405
+      "accuracy": 0.7797927461139896,
+      "precision": 0.7547169811320755,
+      "recall": 0.8290155440414507,
+      "f1": 0.7901234567901234,
+      "auc": 0.8635835861365405,
+      "mcc": 0.5623169838897384
     },
     "Antioxidant": {
-      "accuracy": 0.7117437722419929,
-      "precision": 0.6626506024096386,
-      "recall": 0.8148148148148148,
-      "f1": 0.7308970099667774,
-      "auc": 0.817300862506342,
-      "mcc": 0.4382002786365256
+      "accuracy": 0.7758007117437722,
+      "precision": 0.7686567164179104,
+      "recall": 0.762962962962963,
+      "f1": 0.7657992565055762,
+      "auc": 0.8393455098934551,
+      "mcc": 0.5508018931774569
     },
     "Bitter": {
-      "accuracy": 0.8203125,
-      "precision": 0.7411764705882353,
-      "recall": 0.984375,
-      "f1": 0.8456375838926175,
-      "auc": 0.9591064453125,
-      "mcc": 0.6781724859183206
+      "accuracy": 0.875,
+      "precision": 0.9137931034482759,
+      "recall": 0.828125,
+      "f1": 0.8688524590163934,
+      "auc": 0.9609375,
+      "mcc": 0.7533177846892134
     },
     "Antimalarial": {
       "accuracy": 0.9736263736263736,
-      "precision": 0.8333333333333334,
-      "recall": 0.7142857142857143,
-      "f1": 0.7692307692307693,
-      "auc": 0.9177400468384075,
-      "mcc": 0.7578570897962964
-    },
-    "Anti_inflammatory": {
-      "accuracy": 0.9886111111111111,
-      "precision": 0.9829763866007688,
-      "recall": 0.9944444444444445,
-      "f1": 0.9886771610052472,
-      "auc": 0.9979248456790124,
-      "mcc": 0.97728873441332
+      "precision": 0.8636363636363636,
+      "recall": 0.6785714285714286,
+      "f1": 0.76,
+      "auc": 0.9260204081632653,
+      "mcc": 0.7523383774706999
     },
     "Antimicrobial": {
-      "accuracy": 0.9746453702669496,
-      "precision": 0.9742454728370221,
-      "recall": 0.9369195046439629,
-      "f1": 0.9552179917143421,
-      "auc": 0.9915316704315164,
-      "mcc": 0.9378921309522866
+      "accuracy": 0.9699542052943148,
+      "precision": 0.9372874952776729,
+      "recall": 0.9601393188854489,
+      "f1": 0.9485757981265532,
+      "auc": 0.9824353899280912,
+      "mcc": 0.9274915039611056
     },
     "Signal_peptide": {
-      "accuracy": 0.9926793557833089,
-      "precision": 0.9869753979739508,
-      "recall": 0.9985358711566618,
-      "f1": 0.992721979621543,
-      "auc": 0.99969452655904,
-      "mcc": 0.985426311710703
+      "accuracy": 0.9919472913616398,
+      "precision": 0.9912280701754386,
+      "recall": 0.9926793557833089,
+      "f1": 0.991953182150695,
+      "auc": 0.9985401585032017,
+      "mcc": 0.9838956372992339
     },
     "Antifungal": {
-      "accuracy": 0.9465116279069767,
-      "precision": 0.9615384615384616,
+      "accuracy": 0.9488372093023256,
+      "precision": 0.966183574879227,
       "recall": 0.9302325581395349,
-      "f1": 0.9456264775413712,
-      "auc": 0.9863061114115739,
-      "mcc": 0.8934969492599281
+      "f1": 0.9478672985781991,
+      "auc": 0.9851054624121146,
+      "mcc": 0.8982964942016555
     },
     "Antimalarial_alt": {
-      "accuracy": 0.9877300613496932,
+      "accuracy": 0.9754601226993865,
       "precision": 1.0,
-      "recall": 0.9285714285714286,
-      "f1": 0.9629629629629629,
-      "auc": 0.9941798941798942,
-      "mcc": 0.9565644992853402
+      "recall": 0.8571428571428571,
+      "f1": 0.9230769230769231,
+      "auc": 0.9976190476190476,
+      "mcc": 0.9124017073867975
     },
     "Anticancer_alt": {
-      "accuracy": 0.9329896907216495,
-      "precision": 0.9516129032258065,
-      "recall": 0.9123711340206185,
-      "f1": 0.9315789473684211,
-      "auc": 0.9783584865554256,
-      "mcc": 0.866716620594761
+      "accuracy": 0.9278350515463918,
+      "precision": 0.9322916666666666,
+      "recall": 0.9226804123711341,
+      "f1": 0.927461139896373,
+      "auc": 0.9675975130194495,
+      "mcc": 0.8557155775489096
     },
     "Anti_parasitic": {
-      "accuracy": 0.782608695652174,
-      "precision": 0.8823529411764706,
-      "recall": 0.6521739130434783,
-      "f1": 0.75,
-      "auc": 0.9215500945179584,
-      "mcc": 0.5854905538443586
+      "accuracy": 0.6956521739130435,
+      "precision": 0.9090909090909091,
+      "recall": 0.43478260869565216,
+      "f1": 0.5882352941176471,
+      "auc": 0.8298676748582231,
+      "mcc": 0.458682472293863
     },
     "Umami": {
-      "accuracy": 0.8426966292134831,
-      "precision": 0.7916666666666666,
-      "recall": 0.6785714285714286,
-      "f1": 0.7307692307692307,
-      "auc": 0.9297423887587822,
-      "mcc": 0.6242635125050133
+      "accuracy": 0.8202247191011236,
+      "precision": 0.75,
+      "recall": 0.6428571428571429,
+      "f1": 0.6923076923076923,
+      "auc": 0.9244730679156908,
+      "mcc": 0.5697400065060475
     },
     "Quorum_sensing": {
-      "accuracy": 0.925,
-      "precision": 0.9473684210526315,
-      "recall": 0.9,
-      "f1": 0.9230769230769231,
-      "auc": 0.9850000000000001,
-      "mcc": 0.8510644963469901
+      "accuracy": 0.875,
+      "precision": 0.8947368421052632,
+      "recall": 0.85,
+      "f1": 0.8717948717948718,
+      "auc": 0.9650000000000001,
+      "mcc": 0.7509392614826383
     },
     "Antibacterial": {
-      "accuracy": 0.9430678466076696,
-      "precision": 0.9529553679131484,
-      "recall": 0.9321533923303835,
-      "f1": 0.9424396063226961,
-      "auc": 0.9788722687759418,
-      "mcc": 0.886346891077531
+      "accuracy": 0.947787610619469,
+      "precision": 0.9685185185185186,
+      "recall": 0.9256637168141593,
+      "f1": 0.9466063348416289,
+      "auc": 0.9739487125938688,
+      "mcc": 0.8964532191569444
     },
     "NeuroPred": {
-      "accuracy": 0.865979381443299,
-      "precision": 0.9361179361179361,
-      "recall": 0.7855670103092783,
-      "f1": 0.8542600896860987,
-      "auc": 0.9443936656392815,
-      "mcc": 0.7416123458255524
+      "accuracy": 0.8876288659793814,
+      "precision": 0.8884297520661157,
+      "recall": 0.8865979381443299,
+      "f1": 0.8875128998968008,
+      "auc": 0.9477925390583484,
+      "mcc": 0.7752593798707702
     },
     "Toxicity": {
-      "accuracy": 0.9086206896551724,
-      "precision": 0.9390243902439024,
-      "recall": 0.8587360594795539,
-      "f1": 0.8970873786407767,
-      "auc": 0.9699374843113113,
-      "mcc": 0.8178450017419605
+      "accuracy": 0.9310344827586207,
+      "precision": 0.9288389513108615,
+      "recall": 0.9219330855018587,
+      "f1": 0.9253731343283582,
+      "auc": 0.9623292174183292,
+      "mcc": 0.8612929494836734
     },
     "Antiviral": {
-      "accuracy": 0.8306581059390048,
-      "precision": 0.8259493670886076,
-      "recall": 0.8378812199036918,
-      "f1": 0.8318725099601594,
-      "auc": 0.9097928781410304,
-      "mcc": 0.6613852288786377
+      "accuracy": 0.848314606741573,
+      "precision": 0.872852233676976,
+      "recall": 0.8154093097913323,
+      "f1": 0.8431535269709544,
+      "auc": 0.9125084701220472,
+      "mcc": 0.6981426939615929
     },
     "DPPIV_inhibitory": {
-      "accuracy": 0.8646616541353384,
-      "precision": 0.8211920529801324,
-      "recall": 0.9323308270676691,
-      "f1": 0.8732394366197183,
-      "auc": 0.9477924133642377,
-      "mcc": 0.7360958011737431
+      "accuracy": 0.8345864661654135,
+      "precision": 0.8739495798319328,
+      "recall": 0.7819548872180451,
+      "f1": 0.8253968253968254,
+      "auc": 0.9419695856181808,
+      "mcc": 0.6729113576888817
     },
     "BBP": {
-      "accuracy": 0.6578947368421053,
-      "precision": 0.875,
-      "recall": 0.3684210526315789,
-      "f1": 0.5185185185185185,
-      "auc": 0.9141274238227147,
-      "mcc": 0.3872983346207417
+      "accuracy": 0.8157894736842105,
+      "precision": 0.9285714285714286,
+      "recall": 0.6842105263157895,
+      "f1": 0.7878787878787878,
+      "auc": 0.9224376731301939,
+      "mcc": 0.6546536707079772
     },
     "TTCA": {
-      "accuracy": 0.7360406091370558,
-      "precision": 0.7243589743589743,
-      "recall": 0.9262295081967213,
-      "f1": 0.8129496402877698,
-      "auc": 0.7858469945355191,
-      "mcc": 0.4220882576960022
+      "accuracy": 0.7563451776649747,
+      "precision": 0.7681159420289855,
+      "recall": 0.8688524590163934,
+      "f1": 0.8153846153846154,
+      "auc": 0.7890710382513662,
+      "mcc": 0.46875917622682883
     }
   }
 }