Initial upload: model weights, config, metrics, README, model_def.py, inference.py

README.md

@@ -0,0 +1,37 @@
+# Multi-output DNA Structure Regressor (PyTorch)
+
+This repository contains a PyTorch MLP that predicts multiple structural targets from engineered DNA features.
+
+## Model
+- **Architecture:** 3-layer MLP (512→256→128, dropout 0.3)
+- **Inputs:** 109658 features
+- **Outputs:** 6 targets → mfe_energy, num_pairs, stem_len_mean, num_stems, num_hairpins, num_internal_loops
+- **Loss:** MSE
+- **Optimizer:** Adam (lr=0.0001)
+- **Epochs:** 15
+
+## Metrics (test)
+- Overall MSE: `15022.6787`
+- Overall R²: `-34.0313`
+- Training time (s): `131.85`
+- Prediction time (s): `0.2694`
+
+### MAE per target
+```json
+{
+  "mfe_energy": 139.4054718017578,
+  "num_pairs": 116.53337097167969,
+  "stem_len_mean": 2.4054114818573,
+  "num_stems": 69.17422485351562,
+  "num_hairpins": 14.115099906921387,
+  "num_internal_loops": 94.97564697265625
+}
+```
+
+## Usage
+```bash
+pip install torch numpy
+python inference.py
+```
+
+Ensure to apply any preprocessing (e.g., scaling, SVD) used during training.

config.json

@@ -0,0 +1,15 @@
+{
+  "model_type": "ImprovedNet",
+  "input_size": 109658,
+  "output_size": 6,
+  "hidden": [
+    512,
+    256,
+    128
+  ],
+  "dropout": 0.3,
+  "loss": "MSELoss",
+  "optimizer": "Adam",
+  "learning_rate": 0.0001,
+  "num_epochs": 15
+}

inference.py

@@ -0,0 +1,18 @@
+import torch, json, numpy as np
+from model_def import load_model
+
+DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
+
+with open("config.json") as f:
+    cfg = json.load(f)
+
+model = load_model("pytorch_model.bin", "config.json", device=DEVICE)
+
+x = np.random.randn(1, cfg["input_size"]).astype("float32")
+x_t = torch.from_numpy(x).to(DEVICE)
+
+with torch.no_grad():
+    y_hat = model(x_t).cpu().numpy()
+
+print("Pred shape:", y_hat.shape)
+print("Pred sample:", y_hat[0][: min(5, y_hat.shape[1])])

metrics.json

@@ -0,0 +1,14 @@
+{
+  "overall_mse": 15022.6787109375,
+  "overall_r2": -34.03133010864258,
+  "training_time_seconds": 131.85455417633057,
+  "prediction_time_seconds": 0.2694435119628906,
+  "mae_per_target": {
+    "mfe_energy": 139.4054718017578,
+    "num_pairs": 116.53337097167969,
+    "stem_len_mean": 2.4054114818573,
+    "num_stems": 69.17422485351562,
+    "num_hairpins": 14.115099906921387,
+    "num_internal_loops": 94.97564697265625
+  }
+}

model_def.py

@@ -0,0 +1,28 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+class ImprovedNet(nn.Module):
+    def __init__(self, input_features, output_features, dropout=0.30):
+        super().__init__()
+        self.layer1 = nn.Linear(input_features, 512)
+        self.layer2 = nn.Linear(512, 256)
+        self.layer3 = nn.Linear(256, 128)
+        self.output_layer = nn.Linear(128, output_features)
+        self.dropout = nn.Dropout(p=dropout)
+
+    def forward(self, x):
+        x = F.relu(self.layer1(x)); x = self.dropout(x)
+        x = F.relu(self.layer2(x)); x = self.dropout(x)
+        x = F.relu(self.layer3(x))
+        return self.output_layer(x)
+
+def load_model(weights_path, config_path="config.json", device="cpu"):
+    import json
+    with open(config_path) as f:
+        cfg = json.load(f)
+    model = ImprovedNet(cfg["input_size"], cfg["output_size"], dropout=cfg.get("dropout", 0.30))
+    model.load_state_dict(torch.load(weights_path, map_location=device))
+    model.to(device)
+    model.eval()
+    return model

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f7636d3543892429806ac381d235df25751e679ca5c956b0b9ba8e3d588b086e
+size 225245281