ynuozhang commited on 7 days ago

Commit

470021d

1 Parent(s): 1c0b97e

add inference

Browse files

This view is limited to 50 files because it contains too many changes. See raw diff

Files changed (50) hide show

best_models.txt +10 -0
training_classifiers/.ipynb_checkpoints/train_ml-checkpoint.py +1 -1
training_classifiers/half_life/cnn_smiles/best_model.joblib +3 -0
training_classifiers/half_life/enet_smiles/best_model.joblib +3 -0
training_classifiers/half_life/mlp_smiles/best_model.joblib +3 -0
training_classifiers/half_life/svr_smiles/best_model.joblib +3 -0
training_classifiers/half_life/transformer_smiles/best_model.joblib +3 -0
training_classifiers/half_life/xgb_smiles/best_model.json +0 -0
training_classifiers/half_life/xgb_wt/best_model.json +0 -0
training_classifiers/half_life/xgb_wt/best_model_finetuned.json +0 -0
training_classifiers/half_life/xgb_wt/optimization_summary.json +43 -0
training_classifiers/half_life/xgb_wt/study_trials.csv +3 -0
training_classifiers/hemolysis/svm_gpu_smiles/best_model_cuml_svc.joblib +3 -0
training_classifiers/{solubility/xgb/.ipynb_checkpoints/optimization_summary-checkpoint.txt → hemolysis/svm_gpu_smiles/optimization_summary.txt} +12 -17
training_classifiers/hemolysis/svm_gpu_smiles/pr_curve.png +0 -0
training_classifiers/hemolysis/svm_gpu_smiles/roc_curve.png +0 -0
training_classifiers/hemolysis/svm_gpu_smiles/study_trials.csv +3 -0
training_classifiers/hemolysis/svm_gpu_smiles/train_predictions.csv +3 -0
training_classifiers/hemolysis/svm_gpu_smiles/val_predictions.csv +3 -0
training_classifiers/nf/{transformer_smiles_20 → transformer_smiles}/best_model.pt +0 -0
training_classifiers/nf/{transformer_smiles_20 → transformer_smiles}/best_model_benchmark.json +0 -0
training_classifiers/nf/{transformer_smiles_20 → transformer_smiles}/optimization_summary.txt +0 -0
training_classifiers/nf/{transformer_smiles_20 → transformer_smiles}/pr_curve.png +0 -0
training_classifiers/nf/{transformer_smiles_20 → transformer_smiles}/roc_curve.png +0 -0
training_classifiers/nf/{transformer_smiles_20 → transformer_smiles}/study_trials.csv +0 -0
training_classifiers/nf/{transformer_smiles_20 → transformer_smiles}/train_predictions.csv +0 -0
training_classifiers/nf/{transformer_smiles_20 → transformer_smiles}/val_predictions.csv +0 -0
training_classifiers/permeability_penetrance/cnn_wt/best_model.pt +3 -0
training_classifiers/permeability_penetrance/cnn_wt/best_model_benchmark.json +25 -0
training_classifiers/permeability_penetrance/cnn_wt/optimization_summary.txt +17 -0
training_classifiers/permeability_penetrance/cnn_wt/pr_curve.png +0 -0
training_classifiers/permeability_penetrance/cnn_wt/roc_curve.png +0 -0
training_classifiers/permeability_penetrance/cnn_wt/study_trials.csv +3 -0
training_classifiers/permeability_penetrance/cnn_wt/train_predictions.csv +3 -0
training_classifiers/permeability_penetrance/cnn_wt/val_predictions.csv +3 -0
training_classifiers/permeability_penetrance/enet_gpu_wt/best_model_cuml_enet.joblib +3 -0
training_classifiers/permeability_penetrance/enet_gpu_wt/optimization_summary.txt +17 -0
training_classifiers/permeability_penetrance/enet_gpu_wt/pr_curve.png +0 -0
training_classifiers/permeability_penetrance/enet_gpu_wt/roc_curve.png +0 -0
training_classifiers/permeability_penetrance/enet_gpu_wt/study_trials.csv +3 -0
training_classifiers/permeability_penetrance/enet_gpu_wt/train_predictions.csv +3 -0
training_classifiers/permeability_penetrance/enet_gpu_wt/val_predictions.csv +3 -0
training_classifiers/permeability_penetrance/mlp_wt/best_model.pt +3 -0
training_classifiers/permeability_penetrance/mlp_wt/best_model_benchmark.json +25 -0
training_classifiers/permeability_penetrance/mlp_wt/optimization_summary.txt +15 -0
training_classifiers/permeability_penetrance/mlp_wt/pr_curve.png +0 -0
training_classifiers/permeability_penetrance/mlp_wt/roc_curve.png +0 -0
training_classifiers/permeability_penetrance/mlp_wt/study_trials.csv +3 -0
training_classifiers/permeability_penetrance/mlp_wt/train_predictions.csv +3 -0
training_classifiers/permeability_penetrance/mlp_wt/val_predictions.csv +3 -0

best_models.txt ADDED Viewed

	@@ -0,0 +1,10 @@

+Properties, Best_Model_WT, Best_Model_SMILES, Type, Threshold_WT, Threshold_SMILES,
+Hemolysis, XGB, XGB, Classifier, 0.2801, 0.2223,
+Non-Fouling, MLP, XGB, Classifier, 0.57, -,
+Solubility, CNN, -, Classifier, 0.377, -,
+Permeability (Penetrance), SVM, -, Classifier, 0.5493, -,
+Toxicity, -, Transformer, Classifier, -, 0.3301,
+Binding_affinity, unpooled, unpooled, Regression, -, -,
+Permeability_PAMPA, -, CNN, Regression, -, -,
+Permeability_CACO2, -, SVR, Regression, -, -,
+Halflife, XGB, -, Regression, -, -,

training_classifiers/.ipynb_checkpoints/train_ml-checkpoint.py CHANGED Viewed

@@ -33,7 +33,7 @@ def to_gpu(X: np.ndarray):
 def to_cpu(x):
     if isinstance(x, cp.ndarray):
         return cp.asnumpy(x)
-    return np.asarray(x
 @dataclass
 class SplitData:

 def to_cpu(x):
     if isinstance(x, cp.ndarray):
         return cp.asnumpy(x)
+    return np.asarray(x)
 @dataclass
 class SplitData:

training_classifiers/half_life/cnn_smiles/best_model.joblib ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:146bc7d2572bd217777aac94be413d894692279415256c7fa58299d220d778a1
+size 25767165

training_classifiers/half_life/enet_smiles/best_model.joblib ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:8b4c3eee31dd01959a09f8b2d9cc5172c81e7a5b16cff9f48c934375479610f8
+size 22698

training_classifiers/half_life/mlp_smiles/best_model.joblib ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:f165468db13731cf72d4a651f1cb3ddf4cc0163cdf03d2a53ed1b4552ea37671
+size 4751229

training_classifiers/half_life/svr_smiles/best_model.joblib ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:300cc62988781cb0b34f04cdae549f5076449b54bb7531553ce4e96e35534449
+size 1524311

training_classifiers/half_life/transformer_smiles/best_model.joblib ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:10644b9ef9bc9d551941d830ce58eb6d80257ec0a31f6f2a3ec7bcc8a6786ba5
+size 774301

training_classifiers/half_life/xgb_smiles/best_model.json ADDED Viewed

The diff for this file is too large to render. See raw diff

training_classifiers/half_life/xgb_wt/best_model.json ADDED Viewed

The diff for this file is too large to render. See raw diff

training_classifiers/half_life/xgb_wt/best_model_finetuned.json ADDED Viewed

The diff for this file is too large to render. See raw diff

training_classifiers/half_life/xgb_wt/optimization_summary.json ADDED Viewed

	@@ -0,0 +1,43 @@

+{
+  "best_trial_number": 111,
+  "best_value_cv_spearman_rho": 0.5407086346057526,
+  "best_user_attrs": {
+    "cv_spearman_rho": 0.5407086346057526,
+    "cv_rmse": 30.545584633720324,
+    "cv_mae": 11.373348116874695,
+    "cv_r2": 0.13366737365722656,
+    "cv_mean_best_iter": 2323.0
+  },
+  "best_params": {
+    "lambda": 20.41972407049876,
+    "alpha": 4.815751164899995e-09,
+    "gamma": 0.8263102157574537,
+    "max_depth": 12,
+    "min_child_weight": 0.0016449450459304665,
+    "subsample": 0.9291924815601589,
+    "colsample_bytree": 0.9606594930013117,
+    "learning_rate": 0.0023032791506553446,
+    "num_boost_round": 574,
+    "early_stopping_rounds": 136
+  },
+  "best_xgb_params_full": {
+    "objective": "reg:squarederror",
+    "eval_metric": "rmse",
+    "lambda": 20.41972407049876,
+    "alpha": 4.815751164899995e-09,
+    "gamma": 0.8263102157574537,
+    "max_depth": 12,
+    "min_child_weight": 0.0016449450459304665,
+    "subsample": 0.9291924815601589,
+    "colsample_bytree": 0.9606594930013117,
+    "learning_rate": 0.0023032791506553446,
+    "tree_method": "hist",
+    "device": "cuda",
+    "num_boost_round": 574,
+    "early_stopping_rounds": 136
+  },
+  "base_model_json": "/scratch/pranamlab/tong/PeptiVerse/src/stability/xgboost/best_model.json",
+  "target_transform": "none",
+  "esm_model": "facebook/esm2_t33_650M_UR50D",
+  "esm_max_length": 1024
+}

training_classifiers/half_life/xgb_wt/study_trials.csv ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:9c36eaee2a8503fb31b0b3034e937e35f539b370b84f269549fac4eff43ab7ba
+size 68693

training_classifiers/hemolysis/svm_gpu_smiles/best_model_cuml_svc.joblib ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:54e925ec1a8876a6d02f223429338ca1f940ec010a499f7db3070bcb98a9718c
+size 6562749

training_classifiers/{solubility/xgb/.ipynb_checkpoints/optimization_summary-checkpoint.txt → hemolysis/svm_gpu_smiles/optimization_summary.txt} RENAMED Viewed

@@ -1,22 +1,17 @@
 ========================================================================
-MODEL: xgb
-Best trial: 146
-Best F1 (val @ best-threshold): 0.7203
-Val AUC: 0.7734
-Val AP:  0.7847
-Best threshold (picked on val): 0.3940
-Model saved to: /vast/projects/pranam/lab/yz927/projects/Classifier_Weight/training_classifiers/solubility/xgb/best_model.json
 Best params:
 {
-  "lambda": 3.084325776404141e-06,
-  "alpha": 0.01468401502310261,
-  "colsample_bytree": 0.663085270801208,
-  "subsample": 0.9073538986793201,
-  "learning_rate": 0.024335677079212185,
-  "max_depth": 13,
-  "min_child_weight": 32,
-  "gamma": 3.0440628843516766,
-  "num_boost_round": 682,
-  "early_stopping_rounds": 189
 }
 ========================================================================

 ========================================================================
+MODEL: svm_gpu
+Best trial: 6
+Best F1 (val @ best-threshold): 0.4871
+Val AUC: 0.7458
+Val AP:  0.4338
+Best threshold (picked on val): 0.2014
+Model saved to: /vast/projects/pranam/lab/yz927/projects/Classifier_Weight/training_classifiers/hemolysis/svm_gpu_smiles/best_model_cuml_svc.joblib
 Best params:
 {
+  "C": 2.4210955376276835,
+  "kernel": "linear",
+  "class_weight": null,
+  "max_iter": 2093,
+  "tol": 0.0029131530243637046
 }
 ========================================================================

training_classifiers/hemolysis/svm_gpu_smiles/pr_curve.png ADDED Viewed

training_classifiers/hemolysis/svm_gpu_smiles/roc_curve.png ADDED Viewed

training_classifiers/hemolysis/svm_gpu_smiles/study_trials.csv ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:ec7f92a45bf643bbdf29442bdf7af2d90a1d4e177a2ffedbb9312e7e6ade2c25
+size 4800

training_classifiers/hemolysis/svm_gpu_smiles/train_predictions.csv ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:fced3ff0b1064c51b93d17729470e355e81301db9008d6065a101024ab72eeea
+size 1944898

training_classifiers/hemolysis/svm_gpu_smiles/val_predictions.csv ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:45a4ce19c5cbe99191325e58ee98009d0f8d662486796157ddd2b4feda902fb5
+size 472559

training_classifiers/nf/{transformer_smiles_20 → transformer_smiles}/best_model.pt RENAMED Viewed

File without changes

training_classifiers/nf/{transformer_smiles_20 → transformer_smiles}/best_model_benchmark.json RENAMED Viewed

File without changes

training_classifiers/nf/{transformer_smiles_20 → transformer_smiles}/optimization_summary.txt RENAMED Viewed

File without changes

training_classifiers/nf/{transformer_smiles_20 → transformer_smiles}/pr_curve.png RENAMED Viewed

File without changes

training_classifiers/nf/{transformer_smiles_20 → transformer_smiles}/roc_curve.png RENAMED Viewed

File without changes

training_classifiers/nf/{transformer_smiles_20 → transformer_smiles}/study_trials.csv RENAMED Viewed

File without changes

training_classifiers/nf/{transformer_smiles_20 → transformer_smiles}/train_predictions.csv RENAMED Viewed

File without changes

training_classifiers/nf/{transformer_smiles_20 → transformer_smiles}/val_predictions.csv RENAMED Viewed

File without changes

training_classifiers/permeability_penetrance/cnn_wt/best_model.pt ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:6f5f1d1d08d0dddd78d433f5a5a4c701240016d33d5d235bfb4a00bae502eed1
+size 11016505

training_classifiers/permeability_penetrance/cnn_wt/best_model_benchmark.json ADDED Viewed

	@@ -0,0 +1,25 @@

+{
+  "n_samples": 465,
+  "wall_time_s": 1.2838693670928478,
+  "throughput_samples_per_s": 362.1863811993045,
+  "gpu_total_kernel_ms": 26.139296114444733,
+  "gpu_ms_per_sample": 0.05621354003106394,
+  "gpu_avg_ms_per_batch": 0.5227859222888946,
+  "gpu_peak_mem_MB": 143.54931640625,
+  "telemetry_pre": {
+    "cpu_freq_current_MHz": 1113.310625,
+    "cpu_freq_max_MHz": 4100.0,
+    "cpu_util_pct": 57.1,
+    "cpu_count_logical": 32,
+    "cpu_count_physical": 16,
+    "gpu_stats_error": "No module named 'pynvml'"
+  },
+  "telemetry_post": {
+    "cpu_freq_current_MHz": 1882.3210937499998,
+    "cpu_freq_max_MHz": 4100.0,
+    "cpu_util_pct": 30.6,
+    "cpu_count_logical": 32,
+    "cpu_count_physical": 16,
+    "gpu_stats_error": "No module named 'pynvml'"
+  }
+}

training_classifiers/permeability_penetrance/cnn_wt/optimization_summary.txt ADDED Viewed

	@@ -0,0 +1,17 @@

+========================================================================
+MODEL: cnn
+Best Optuna AUC (objective): 0.9279
+Refit best AUC (val): 0.9588
+Refit best F1@thr (val): 0.9163 at thr=0.9354
+Best params:
+{
+  "lr": 0.0006805038675719076,
+  "weight_decay": 3.828265226735556e-06,
+  "dropout": 0.14299889017563588,
+  "batch_size": 32,
+  "channels": 256,
+  "kernel": 7,
+  "layers": 2
+}
+Saved model: ../../penn_work/peptiverse/permeability/cnn_outputs/best_model.pt
+========================================================================

training_classifiers/permeability_penetrance/cnn_wt/pr_curve.png ADDED Viewed

training_classifiers/permeability_penetrance/cnn_wt/roc_curve.png ADDED Viewed

training_classifiers/permeability_penetrance/cnn_wt/study_trials.csv ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:1996f669a573ddc4997400192b22ab842004bf3a08fdf312aed7359b22db71b0
+size 12146

training_classifiers/permeability_penetrance/cnn_wt/train_predictions.csv ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:f96abb09cd4790ef8ba6d4cdf6681a0e930b5dba4ac46e30577da81a752d17a7
+size 74503

training_classifiers/permeability_penetrance/cnn_wt/val_predictions.csv ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:b01927dda3f2731b23666c89990de1a7fb0e3bcf177ca0f0ecbf8e1a416b0887
+size 18783

training_classifiers/permeability_penetrance/enet_gpu_wt/best_model_cuml_enet.joblib ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:6de7df13cd690448315cf058ea9ff2789e528f399843b191b157873e675ca57b
+size 6667

training_classifiers/permeability_penetrance/enet_gpu_wt/optimization_summary.txt ADDED Viewed

	@@ -0,0 +1,17 @@

+========================================================================
+MODEL: enet_gpu
+Best trial: 55
+Best F1 (val @ best-threshold): 0.9052
+Val AUC: 0.9566
+Val AP:  0.9449
+Best threshold (picked on val): 0.3758
+Model saved to: ../../penn_work/peptiverse/permeability/enet_outputs/best_model_cuml_enet.joblib
+Best params:
+{
+  "C": 5.4606611625825305,
+  "l1_ratio": 0.026268494702099424,
+  "class_weight": "balanced",
+  "max_iter": 1564,
+  "tol": 1.5231156120802544e-06
+}
+========================================================================

training_classifiers/permeability_penetrance/enet_gpu_wt/pr_curve.png ADDED Viewed

training_classifiers/permeability_penetrance/enet_gpu_wt/roc_curve.png ADDED Viewed

training_classifiers/permeability_penetrance/enet_gpu_wt/study_trials.csv ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:6363dac10adb105d4a34875d01013db6c3896966693ab44e24ba186a54b06343
+size 47738

training_classifiers/permeability_penetrance/enet_gpu_wt/train_predictions.csv ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:6da95a76b2d20f965dfee23c743a97ac4a0a7ff94a3f0238e561277095c16cea
+size 80210

training_classifiers/permeability_penetrance/enet_gpu_wt/val_predictions.csv ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:7f04b935f70147e873ac1389ef1e2b96784da6e28ffe3b34b7a4c459b9711f87
+size 20042

training_classifiers/permeability_penetrance/mlp_wt/best_model.pt ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:ae99b76dcd58e2e0fe7de459bc5228882062f515be7fce6b543d915072c86a89
+size 5253819

training_classifiers/permeability_penetrance/mlp_wt/best_model_benchmark.json ADDED Viewed

	@@ -0,0 +1,25 @@

+{
+  "n_samples": 465,
+  "wall_time_s": 1.4508731923997402,
+  "throughput_samples_per_s": 320.4966515584255,
+  "gpu_total_kernel_ms": 5.213824018836021,
+  "gpu_ms_per_sample": 0.01121252477169037,
+  "gpu_avg_ms_per_batch": 0.10427648037672042,
+  "gpu_peak_mem_MB": 102.32373046875,
+  "telemetry_pre": {
+    "cpu_freq_current_MHz": 2871.8869999999997,
+    "cpu_freq_max_MHz": 4100.0,
+    "cpu_util_pct": 34.8,
+    "cpu_count_logical": 32,
+    "cpu_count_physical": 16,
+    "gpu_stats_error": "No module named 'pynvml'"
+  },
+  "telemetry_post": {
+    "cpu_freq_current_MHz": 3379.5095624999994,
+    "cpu_freq_max_MHz": 4100.0,
+    "cpu_util_pct": 64.3,
+    "cpu_count_logical": 32,
+    "cpu_count_physical": 16,
+    "gpu_stats_error": "No module named 'pynvml'"
+  }
+}

training_classifiers/permeability_penetrance/mlp_wt/optimization_summary.txt ADDED Viewed

	@@ -0,0 +1,15 @@

+========================================================================
+MODEL: mlp
+Best Optuna AUC (objective): 0.9087
+Refit best AUC (val): 0.9590
+Refit best F1@thr (val): 0.9078 at thr=0.8393
+Best params:
+{
+  "lr": 0.0016146321628309815,
+  "weight_decay": 2.852382206601946e-08,
+  "dropout": 0.11865519727740499,
+  "batch_size": 32,
+  "hidden": 1024
+}
+Saved model: ../../penn_work/peptiverse/permeability/mlp_outputs/best_model.pt
+========================================================================

training_classifiers/permeability_penetrance/mlp_wt/pr_curve.png ADDED Viewed

training_classifiers/permeability_penetrance/mlp_wt/roc_curve.png ADDED Viewed

training_classifiers/permeability_penetrance/mlp_wt/study_trials.csv ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:cf0b2bb6a8fe59a25aab26db66c3ccfe41b8171f82659e7f5300687d560db568
+size 11921

training_classifiers/permeability_penetrance/mlp_wt/train_predictions.csv ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:4f57720374fa2c2cbdb2d6e80a95839aa14de5647a8b2cc5042880515667746b
+size 77411

training_classifiers/permeability_penetrance/mlp_wt/val_predictions.csv ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:9a86408aae05a02dc26372d4740b561a75eac23654418276cbea6360fd32d566
+size 19389