wavedl 1.5.7__tar.gz → 1.6.1__tar.gz

{wavedl-1.5.7/src/wavedl.egg-info → wavedl-1.6.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: wavedl
-Version: 1.5.7
+Version: 1.6.1
 Summary: A Scalable Deep Learning Framework for Wave-Based Inverse Problems
 Author: Ductho Le
 License: MIT
@@ -23,6 +23,7 @@ Description-Content-Type: text/markdown
 License-File: LICENSE
 Requires-Dist: torch>=2.0.0
 Requires-Dist: torchvision>=0.15.0
+Requires-Dist: timm>=0.9.0
 Requires-Dist: accelerate>=0.20.0
 Requires-Dist: numpy>=1.24.0
 Requires-Dist: scipy>=1.10.0
@@ -37,6 +38,7 @@ Requires-Dist: wandb>=0.15.0
 Requires-Dist: optuna>=3.0.0
 Requires-Dist: onnx>=1.14.0
 Requires-Dist: onnxruntime>=1.15.0
+Requires-Dist: onnxscript>=0.1.0
 Requires-Dist: triton>=2.0.0; sys_platform == "linux"
 Provides-Extra: dev
 Requires-Dist: pytest>=7.0.0; extra == "dev"
@@ -117,7 +119,7 @@ Train on datasets larger than RAM:
 
 **🧠 Models? We've Got Options**
 
-38 architectures, ready to go:
+69 architectures, ready to go:
 - CNNs, ResNets, ViTs, EfficientNets...
 - All adapted for regression
 - [Add your own](#adding-custom-models) in one line
@@ -202,7 +204,7 @@ Deploy models anywhere:
 #### From PyPI (recommended for all users)
 
 ```bash
-pip install wavedl
+pip install --upgrade wavedl
 ```
 
 This installs everything you need: training, inference, HPO, ONNX export.
@@ -358,22 +360,10 @@ WaveDL/
 │       ├── hpo.py                # Hyperparameter optimization
 │       ├── hpc.py                # HPC distributed training launcher
 │       │
-│       ├── models/               # Model architectures (38 variants)
+│       ├── models/               # Model Zoo (69 architectures)
 │       │   ├── registry.py       # Model factory (@register_model)
 │       │   ├── base.py           # Abstract base class
-│       │   ├── cnn.py            # Baseline CNN (1D/2D/3D)
-│       │   ├── resnet.py         # ResNet-18/34/50 (1D/2D/3D)
-│       │   ├── resnet3d.py       # ResNet3D-18, MC3-18 (3D only)
-│       │   ├── tcn.py            # TCN (1D only)
-│       │   ├── efficientnet.py   # EfficientNet-B0/B1/B2 (2D)
-│       │   ├── efficientnetv2.py # EfficientNetV2-S/M/L (2D)
-│       │   ├── mobilenetv3.py    # MobileNetV3-Small/Large (2D)
-│       │   ├── regnet.py         # RegNetY variants (2D)
-│       │   ├── swin.py           # Swin Transformer (2D)
-│       │   ├── vit.py            # Vision Transformer (1D/2D)
-│       │   ├── convnext.py       # ConvNeXt (1D/2D/3D)
-│       │   ├── densenet.py       # DenseNet-121/169 (1D/2D/3D)
-│       │   └── unet.py           # U-Net Regression
+│       │   └── ...               # See "Available Models" section
 │       │
 │       └── utils/                # Utilities
 │           ├── data.py           # Memory-mapped data pipeline
@@ -388,7 +378,7 @@ WaveDL/
 ├── configs/                      # YAML config templates
 ├── examples/                     # Ready-to-run examples
 ├── notebooks/                    # Jupyter notebooks
-├── unit_tests/                   # Pytest test suite (903 tests)
+├── unit_tests/                   # Pytest test suite
 │
 ├── pyproject.toml                # Package config, dependencies
 ├── CHANGELOG.md                  # Version history
@@ -411,71 +401,117 @@ WaveDL/
 > ```
 
 <details>
-<summary><b>Available Models</b> — 38 architectures</summary>
+<summary><b>Available Models</b> — 69 architectures</summary>
 
-| Model | Params | Dim |
-|-------|--------|-----|
+| Model | Backbone Params | Dim |
+|-------|-----------------|-----|
+| **── Classic CNNs ──** |||
 | **CNN** — Convolutional Neural Network |||
-| `cnn` | 1.7M | 1D/2D/3D |
+| `cnn` | 1.6M | 1D/2D/3D |
 | **ResNet** — Residual Network |||
-| `resnet18` | 11.4M | 1D/2D/3D |
-| `resnet34` | 21.5M | 1D/2D/3D |
-| `resnet50` | 24.6M | 1D/2D/3D |
-| `resnet18_pretrained` ⭐ | 11.4M | 2D |
-| `resnet50_pretrained` ⭐ | 24.6M | 2D |
-| **ResNet3D** — 3D Residual Network |||
-| `resnet3d_18` | 33.6M | 3D |
-| `mc3_18` — Mixed Convolution 3D | 11.9M | 3D |
-| **TCN** — Temporal Convolutional Network |||
-| `tcn_small` | 1.0M | 1D |
-| `tcn` | 7.0M | 1D |
-| `tcn_large` | 10.2M | 1D |
+| `resnet18` | 11.2M | 1D/2D/3D |
+| `resnet34` | 21.3M | 1D/2D/3D |
+| `resnet50` | 23.5M | 1D/2D/3D |
+| `resnet18_pretrained` ⭐ | 11.2M | 2D |
+| `resnet50_pretrained` ⭐ | 23.5M | 2D |
+| **DenseNet** — Densely Connected Network |||
+| `densenet121` | 7.0M | 1D/2D/3D |
+| `densenet169` | 12.5M | 1D/2D/3D |
+| `densenet121_pretrained` ⭐ | 7.0M | 2D |
+| **── Efficient/Mobile CNNs ──** |||
+| **MobileNetV3** — Mobile Neural Network V3 |||
+| `mobilenet_v3_small` ⭐ | 0.9M | 2D |
+| `mobilenet_v3_large` ⭐ | 3.0M | 2D |
 | **EfficientNet** — Efficient Neural Network |||
-| `efficientnet_b0` ⭐ | 4.7M | 2D |
-| `efficientnet_b1` ⭐ | 7.2M | 2D |
-| `efficientnet_b2` ⭐ | 8.4M | 2D |
+| `efficientnet_b0` ⭐ | 4.0M | 2D |
+| `efficientnet_b1` ⭐ | 6.5M | 2D |
+| `efficientnet_b2` ⭐ | 7.7M | 2D |
 | **EfficientNetV2** — Efficient Neural Network V2 |||
-| `efficientnet_v2_s` ⭐ | 21.0M | 2D |
-| `efficientnet_v2_m` ⭐ | 53.6M | 2D |
-| `efficientnet_v2_l` ⭐ | 118.0M | 2D |
-| **MobileNetV3** — Mobile Neural Network V3 |||
-| `mobilenet_v3_small` ⭐ | 1.1M | 2D |
-| `mobilenet_v3_large` ⭐ | 3.2M | 2D |
+| `efficientnet_v2_s` ⭐ | 20.2M | 2D |
+| `efficientnet_v2_m` ⭐ | 52.9M | 2D |
+| `efficientnet_v2_l` ⭐ | 117.2M | 2D |
 | **RegNet** — Regularized Network |||
-| `regnet_y_400mf` ⭐ | 4.0M | 2D |
-| `regnet_y_800mf` ⭐ | 5.8M | 2D |
-| `regnet_y_1_6gf` ⭐ | 10.5M | 2D |
-| `regnet_y_3_2gf` ⭐ | 18.3M | 2D |
-| `regnet_y_8gf` ⭐ | 37.9M | 2D |
-| **Swin** — Shifted Window Transformer |||
-| `swin_t` ⭐ | 28.0M | 2D |
-| `swin_s` ⭐ | 49.4M | 2D |
-| `swin_b` ⭐ | 87.4M | 2D |
+| `regnet_y_400mf` ⭐ | 3.9M | 2D |
+| `regnet_y_800mf` ⭐ | 5.7M | 2D |
+| `regnet_y_1_6gf` ⭐ | 10.3M | 2D |
+| `regnet_y_3_2gf` ⭐ | 17.9M | 2D |
+| `regnet_y_8gf` ⭐ | 37.4M | 2D |
+| **── Modern CNNs ──** |||
 | **ConvNeXt** — Convolutional Next |||
-| `convnext_tiny` | 28.2M | 1D/2D/3D |
-| `convnext_small` | 49.8M | 1D/2D/3D |
-| `convnext_base` | 88.1M | 1D/2D/3D |
-| `convnext_tiny_pretrained` ⭐ | 28.2M | 2D |
-| **DenseNet** — Densely Connected Network |||
-| `densenet121` | 7.5M | 1D/2D/3D |
-| `densenet169` | 13.3M | 1D/2D/3D |
-| `densenet121_pretrained` ⭐ | 7.5M | 2D |
+| `convnext_tiny` | 27.8M | 1D/2D/3D |
+| `convnext_small` | 49.5M | 1D/2D/3D |
+| `convnext_base` | 87.6M | 1D/2D/3D |
+| `convnext_tiny_pretrained` ⭐ | 27.8M | 2D |
+| **ConvNeXt V2** — ConvNeXt with GRN |||
+| `convnext_v2_tiny` | 27.9M | 1D/2D/3D |
+| `convnext_v2_small` | 49.6M | 1D/2D/3D |
+| `convnext_v2_base` | 87.7M | 1D/2D/3D |
+| `convnext_v2_tiny_pretrained` ⭐ | 27.9M | 2D |
+| **UniRepLKNet** — Large-Kernel ConvNet |||
+| `unireplknet_tiny` | 30.8M | 1D/2D/3D |
+| `unireplknet_small` | 56.0M | 1D/2D/3D |
+| `unireplknet_base` | 97.6M | 1D/2D/3D |
+| **── Vision Transformers ──** |||
 | **ViT** — Vision Transformer |||
-| `vit_tiny` | 5.5M | 1D/2D |
-| `vit_small` | 21.6M | 1D/2D |
-| `vit_base` | 85.6M | 1D/2D |
+| `vit_tiny` | 5.4M | 1D/2D |
+| `vit_small` | 21.4M | 1D/2D |
+| `vit_base` | 85.3M | 1D/2D |
+| **Swin** — Shifted Window Transformer |||
+| `swin_t` ⭐ | 27.5M | 2D |
+| `swin_s` ⭐ | 48.8M | 2D |
+| `swin_b` ⭐ | 86.7M | 2D |
+| **MaxViT** — Multi-Axis ViT |||
+| `maxvit_tiny` ⭐ | 30.1M | 2D |
+| `maxvit_small` ⭐ | 67.6M | 2D |
+| `maxvit_base` ⭐ | 119.1M | 2D |
+| **── Hybrid CNN-Transformer ──** |||
+| **FastViT** — Fast Hybrid CNN-ViT |||
+| `fastvit_t8` ⭐ | 4.0M | 2D |
+| `fastvit_t12` ⭐ | 6.8M | 2D |
+| `fastvit_s12` ⭐ | 8.8M | 2D |
+| `fastvit_sa12` ⭐ | 10.9M | 2D |
+| **CAFormer** — MetaFormer with Attention |||
+| `caformer_s18` ⭐ | 26.3M | 2D |
+| `caformer_s36` ⭐ | 39.2M | 2D |
+| `caformer_m36` ⭐ | 56.9M | 2D |
+| `poolformer_s12` ⭐ | 11.9M | 2D |
+| **EfficientViT** — Memory-Efficient ViT |||
+| `efficientvit_m0` ⭐ | 2.2M | 2D |
+| `efficientvit_m1` ⭐ | 2.6M | 2D |
+| `efficientvit_m2` ⭐ | 3.8M | 2D |
+| `efficientvit_b0` ⭐ | 2.1M | 2D |
+| `efficientvit_b1` ⭐ | 7.5M | 2D |
+| `efficientvit_b2` ⭐ | 21.8M | 2D |
+| `efficientvit_b3` ⭐ | 46.1M | 2D |
+| `efficientvit_l1` ⭐ | 49.5M | 2D |
+| `efficientvit_l2` ⭐ | 60.5M | 2D |
+| **── State Space Models ──** |||
+| **Mamba** — State Space Model |||
+| `mamba_1d` | 3.4M | 1D |
+| **Vision Mamba (ViM)** — 2D Mamba |||
+| `vim_tiny` | 6.6M | 2D |
+| `vim_small` | 51.1M | 2D |
+| `vim_base` | 201.4M | 2D |
+| **── Specialized Architectures ──** |||
+| **TCN** — Temporal Convolutional Network |||
+| `tcn_small` | 0.9M | 1D |
+| `tcn` | 6.9M | 1D |
+| `tcn_large` | 10.0M | 1D |
+| **ResNet3D** — 3D Residual Network |||
+| `resnet3d_18` | 33.2M | 3D |
+| `mc3_18` — Mixed Convolution 3D | 11.5M | 3D |
 | **U-Net** — U-shaped Network |||
-| `unet_regression` | 31.1M | 1D/2D/3D |
+| `unet_regression` | 31.0M | 1D/2D/3D |
+
 
 ⭐ = **Pretrained on ImageNet** (recommended for smaller datasets). Weights are downloaded automatically on first use.
 - **Cache location**: `~/.cache/torch/hub/checkpoints/` (or `./.torch_cache/` on HPC if home is not writable)
-- **Size**: ~20–350 MB per model depending on architecture
 - **Train from scratch**: Use `--no_pretrained` to disable pretrained weights
 
 **💡 HPC Users**: If compute nodes block internet, pre-download weights on the login node:
 
 ```bash
-# Run once on login node (with internet) — downloads ALL pretrained weights (~1.5 GB total)
+# Run once on login node (with internet) — downloads ALL pretrained weights
 python -c "
 import os
 os.environ['TORCH_HOME'] = '.torch_cache'  # Match WaveDL's HPC cache location
@@ -483,24 +519,56 @@ os.environ['TORCH_HOME'] = '.torch_cache'  # Match WaveDL's HPC cache location
 from torchvision import models as m
 from torchvision.models import video as v
 
-# Model name -> Weights class mapping
-weights = {
-    'resnet18': m.ResNet18_Weights, 'resnet50': m.ResNet50_Weights,
-    'efficientnet_b0': m.EfficientNet_B0_Weights, 'efficientnet_b1': m.EfficientNet_B1_Weights,
-    'efficientnet_b2': m.EfficientNet_B2_Weights, 'efficientnet_v2_s': m.EfficientNet_V2_S_Weights,
-    'efficientnet_v2_m': m.EfficientNet_V2_M_Weights, 'efficientnet_v2_l': m.EfficientNet_V2_L_Weights,
-    'mobilenet_v3_small': m.MobileNet_V3_Small_Weights, 'mobilenet_v3_large': m.MobileNet_V3_Large_Weights,
-    'regnet_y_400mf': m.RegNet_Y_400MF_Weights, 'regnet_y_800mf': m.RegNet_Y_800MF_Weights,
-    'regnet_y_1_6gf': m.RegNet_Y_1_6GF_Weights, 'regnet_y_3_2gf': m.RegNet_Y_3_2GF_Weights,
-    'regnet_y_8gf': m.RegNet_Y_8GF_Weights, 'swin_t': m.Swin_T_Weights, 'swin_s': m.Swin_S_Weights,
-    'swin_b': m.Swin_B_Weights, 'convnext_tiny': m.ConvNeXt_Tiny_Weights, 'densenet121': m.DenseNet121_Weights,
-}
-for name, w in weights.items():
-    getattr(m, name)(weights=w.DEFAULT); print(f'✓ {name}')
+# === TorchVision Models (use IMAGENET1K_V1 to match WaveDL) ===
+models = [
+    ('resnet18', m.ResNet18_Weights.IMAGENET1K_V1),
+    ('resnet50', m.ResNet50_Weights.IMAGENET1K_V1),
+    ('efficientnet_b0', m.EfficientNet_B0_Weights.IMAGENET1K_V1),
+    ('efficientnet_b1', m.EfficientNet_B1_Weights.IMAGENET1K_V1),
+    ('efficientnet_b2', m.EfficientNet_B2_Weights.IMAGENET1K_V1),
+    ('efficientnet_v2_s', m.EfficientNet_V2_S_Weights.IMAGENET1K_V1),
+    ('efficientnet_v2_m', m.EfficientNet_V2_M_Weights.IMAGENET1K_V1),
+    ('efficientnet_v2_l', m.EfficientNet_V2_L_Weights.IMAGENET1K_V1),
+    ('mobilenet_v3_small', m.MobileNet_V3_Small_Weights.IMAGENET1K_V1),
+    ('mobilenet_v3_large', m.MobileNet_V3_Large_Weights.IMAGENET1K_V1),
+    ('regnet_y_400mf', m.RegNet_Y_400MF_Weights.IMAGENET1K_V1),
+    ('regnet_y_800mf', m.RegNet_Y_800MF_Weights.IMAGENET1K_V1),
+    ('regnet_y_1_6gf', m.RegNet_Y_1_6GF_Weights.IMAGENET1K_V1),
+    ('regnet_y_3_2gf', m.RegNet_Y_3_2GF_Weights.IMAGENET1K_V1),
+    ('regnet_y_8gf', m.RegNet_Y_8GF_Weights.IMAGENET1K_V1),
+    ('swin_t', m.Swin_T_Weights.IMAGENET1K_V1),
+    ('swin_s', m.Swin_S_Weights.IMAGENET1K_V1),
+    ('swin_b', m.Swin_B_Weights.IMAGENET1K_V1),
+    ('convnext_tiny', m.ConvNeXt_Tiny_Weights.IMAGENET1K_V1),
+    ('densenet121', m.DenseNet121_Weights.IMAGENET1K_V1),
+]
+for name, w in models:
+    getattr(m, name)(weights=w); print(f'✓ {name}')
 
 # 3D video models
-v.r3d_18(weights=v.R3D_18_Weights.DEFAULT); print('✓ r3d_18')
-v.mc3_18(weights=v.MC3_18_Weights.DEFAULT); print('✓ mc3_18')
+v.r3d_18(weights=v.R3D_18_Weights.KINETICS400_V1); print('✓ r3d_18')
+v.mc3_18(weights=v.MC3_18_Weights.KINETICS400_V1); print('✓ mc3_18')
+
+# === Timm Models (MaxViT, FastViT, CAFormer, ConvNeXt V2) ===
+import timm
+
+timm_models = [
+    # MaxViT (no suffix - timm resolves to default)
+    'maxvit_tiny_tf_224', 'maxvit_small_tf_224', 'maxvit_base_tf_224',
+    # FastViT (no suffix)
+    'fastvit_t8', 'fastvit_t12', 'fastvit_s12', 'fastvit_sa12',
+    # CAFormer/PoolFormer (no suffix)
+    'caformer_s18', 'caformer_s36', 'caformer_m36', 'poolformer_s12',
+    # ConvNeXt V2 (no suffix)
+    'convnextv2_tiny',
+    # EfficientViT (no suffix)
+    'efficientvit_m0', 'efficientvit_m1', 'efficientvit_m2',
+    'efficientvit_b0', 'efficientvit_b1', 'efficientvit_b2', 'efficientvit_b3',
+    'efficientvit_l1', 'efficientvit_l2',
+]
+for name in timm_models:
+    timm.create_model(name, pretrained=True); print(f'✓ {name}')
+
 print('\\n✓ All pretrained weights cached!')
 "
 ```

{wavedl-1.5.7 → wavedl-1.6.1}/README.md

@@ -71,7 +71,7 @@ Train on datasets larger than RAM:
 
 **🧠 Models? We've Got Options**
 
-38 architectures, ready to go:
+69 architectures, ready to go:
 - CNNs, ResNets, ViTs, EfficientNets...
 - All adapted for regression
 - [Add your own](#adding-custom-models) in one line
@@ -156,7 +156,7 @@ Deploy models anywhere:
 #### From PyPI (recommended for all users)
 
 ```bash
-pip install wavedl
+pip install --upgrade wavedl
 ```
 
 This installs everything you need: training, inference, HPO, ONNX export.
@@ -312,22 +312,10 @@ WaveDL/
 │       ├── hpo.py                # Hyperparameter optimization
 │       ├── hpc.py                # HPC distributed training launcher
 │       │
-│       ├── models/               # Model architectures (38 variants)
+│       ├── models/               # Model Zoo (69 architectures)
 │       │   ├── registry.py       # Model factory (@register_model)
 │       │   ├── base.py           # Abstract base class
-│       │   ├── cnn.py            # Baseline CNN (1D/2D/3D)
-│       │   ├── resnet.py         # ResNet-18/34/50 (1D/2D/3D)
-│       │   ├── resnet3d.py       # ResNet3D-18, MC3-18 (3D only)
-│       │   ├── tcn.py            # TCN (1D only)
-│       │   ├── efficientnet.py   # EfficientNet-B0/B1/B2 (2D)
-│       │   ├── efficientnetv2.py # EfficientNetV2-S/M/L (2D)
-│       │   ├── mobilenetv3.py    # MobileNetV3-Small/Large (2D)
-│       │   ├── regnet.py         # RegNetY variants (2D)
-│       │   ├── swin.py           # Swin Transformer (2D)
-│       │   ├── vit.py            # Vision Transformer (1D/2D)
-│       │   ├── convnext.py       # ConvNeXt (1D/2D/3D)
-│       │   ├── densenet.py       # DenseNet-121/169 (1D/2D/3D)
-│       │   └── unet.py           # U-Net Regression
+│       │   └── ...               # See "Available Models" section
 │       │
 │       └── utils/                # Utilities
 │           ├── data.py           # Memory-mapped data pipeline
@@ -342,7 +330,7 @@ WaveDL/
 ├── configs/                      # YAML config templates
 ├── examples/                     # Ready-to-run examples
 ├── notebooks/                    # Jupyter notebooks
-├── unit_tests/                   # Pytest test suite (903 tests)
+├── unit_tests/                   # Pytest test suite
 │
 ├── pyproject.toml                # Package config, dependencies
 ├── CHANGELOG.md                  # Version history
@@ -365,71 +353,117 @@ WaveDL/
 > ```
 
 <details>
-<summary><b>Available Models</b> — 38 architectures</summary>
+<summary><b>Available Models</b> — 69 architectures</summary>
 
-| Model | Params | Dim |
-|-------|--------|-----|
+| Model | Backbone Params | Dim |
+|-------|-----------------|-----|
+| **── Classic CNNs ──** |||
 | **CNN** — Convolutional Neural Network |||
-| `cnn` | 1.7M | 1D/2D/3D |
+| `cnn` | 1.6M | 1D/2D/3D |
 | **ResNet** — Residual Network |||
-| `resnet18` | 11.4M | 1D/2D/3D |
-| `resnet34` | 21.5M | 1D/2D/3D |
-| `resnet50` | 24.6M | 1D/2D/3D |
-| `resnet18_pretrained` ⭐ | 11.4M | 2D |
-| `resnet50_pretrained` ⭐ | 24.6M | 2D |
-| **ResNet3D** — 3D Residual Network |||
-| `resnet3d_18` | 33.6M | 3D |
-| `mc3_18` — Mixed Convolution 3D | 11.9M | 3D |
-| **TCN** — Temporal Convolutional Network |||
-| `tcn_small` | 1.0M | 1D |
-| `tcn` | 7.0M | 1D |
-| `tcn_large` | 10.2M | 1D |
+| `resnet18` | 11.2M | 1D/2D/3D |
+| `resnet34` | 21.3M | 1D/2D/3D |
+| `resnet50` | 23.5M | 1D/2D/3D |
+| `resnet18_pretrained` ⭐ | 11.2M | 2D |
+| `resnet50_pretrained` ⭐ | 23.5M | 2D |
+| **DenseNet** — Densely Connected Network |||
+| `densenet121` | 7.0M | 1D/2D/3D |
+| `densenet169` | 12.5M | 1D/2D/3D |
+| `densenet121_pretrained` ⭐ | 7.0M | 2D |
+| **── Efficient/Mobile CNNs ──** |||
+| **MobileNetV3** — Mobile Neural Network V3 |||
+| `mobilenet_v3_small` ⭐ | 0.9M | 2D |
+| `mobilenet_v3_large` ⭐ | 3.0M | 2D |
 | **EfficientNet** — Efficient Neural Network |||
-| `efficientnet_b0` ⭐ | 4.7M | 2D |
-| `efficientnet_b1` ⭐ | 7.2M | 2D |
-| `efficientnet_b2` ⭐ | 8.4M | 2D |
+| `efficientnet_b0` ⭐ | 4.0M | 2D |
+| `efficientnet_b1` ⭐ | 6.5M | 2D |
+| `efficientnet_b2` ⭐ | 7.7M | 2D |
 | **EfficientNetV2** — Efficient Neural Network V2 |||
-| `efficientnet_v2_s` ⭐ | 21.0M | 2D |
-| `efficientnet_v2_m` ⭐ | 53.6M | 2D |
-| `efficientnet_v2_l` ⭐ | 118.0M | 2D |
-| **MobileNetV3** — Mobile Neural Network V3 |||
-| `mobilenet_v3_small` ⭐ | 1.1M | 2D |
-| `mobilenet_v3_large` ⭐ | 3.2M | 2D |
+| `efficientnet_v2_s` ⭐ | 20.2M | 2D |
+| `efficientnet_v2_m` ⭐ | 52.9M | 2D |
+| `efficientnet_v2_l` ⭐ | 117.2M | 2D |
 | **RegNet** — Regularized Network |||
-| `regnet_y_400mf` ⭐ | 4.0M | 2D |
-| `regnet_y_800mf` ⭐ | 5.8M | 2D |
-| `regnet_y_1_6gf` ⭐ | 10.5M | 2D |
-| `regnet_y_3_2gf` ⭐ | 18.3M | 2D |
-| `regnet_y_8gf` ⭐ | 37.9M | 2D |
-| **Swin** — Shifted Window Transformer |||
-| `swin_t` ⭐ | 28.0M | 2D |
-| `swin_s` ⭐ | 49.4M | 2D |
-| `swin_b` ⭐ | 87.4M | 2D |
+| `regnet_y_400mf` ⭐ | 3.9M | 2D |
+| `regnet_y_800mf` ⭐ | 5.7M | 2D |
+| `regnet_y_1_6gf` ⭐ | 10.3M | 2D |
+| `regnet_y_3_2gf` ⭐ | 17.9M | 2D |
+| `regnet_y_8gf` ⭐ | 37.4M | 2D |
+| **── Modern CNNs ──** |||
 | **ConvNeXt** — Convolutional Next |||
-| `convnext_tiny` | 28.2M | 1D/2D/3D |
-| `convnext_small` | 49.8M | 1D/2D/3D |
-| `convnext_base` | 88.1M | 1D/2D/3D |
-| `convnext_tiny_pretrained` ⭐ | 28.2M | 2D |
-| **DenseNet** — Densely Connected Network |||
-| `densenet121` | 7.5M | 1D/2D/3D |
-| `densenet169` | 13.3M | 1D/2D/3D |
-| `densenet121_pretrained` ⭐ | 7.5M | 2D |
+| `convnext_tiny` | 27.8M | 1D/2D/3D |
+| `convnext_small` | 49.5M | 1D/2D/3D |
+| `convnext_base` | 87.6M | 1D/2D/3D |
+| `convnext_tiny_pretrained` ⭐ | 27.8M | 2D |
+| **ConvNeXt V2** — ConvNeXt with GRN |||
+| `convnext_v2_tiny` | 27.9M | 1D/2D/3D |
+| `convnext_v2_small` | 49.6M | 1D/2D/3D |
+| `convnext_v2_base` | 87.7M | 1D/2D/3D |
+| `convnext_v2_tiny_pretrained` ⭐ | 27.9M | 2D |
+| **UniRepLKNet** — Large-Kernel ConvNet |||
+| `unireplknet_tiny` | 30.8M | 1D/2D/3D |
+| `unireplknet_small` | 56.0M | 1D/2D/3D |
+| `unireplknet_base` | 97.6M | 1D/2D/3D |
+| **── Vision Transformers ──** |||
 | **ViT** — Vision Transformer |||
-| `vit_tiny` | 5.5M | 1D/2D |
-| `vit_small` | 21.6M | 1D/2D |
-| `vit_base` | 85.6M | 1D/2D |
+| `vit_tiny` | 5.4M | 1D/2D |
+| `vit_small` | 21.4M | 1D/2D |
+| `vit_base` | 85.3M | 1D/2D |
+| **Swin** — Shifted Window Transformer |||
+| `swin_t` ⭐ | 27.5M | 2D |
+| `swin_s` ⭐ | 48.8M | 2D |
+| `swin_b` ⭐ | 86.7M | 2D |
+| **MaxViT** — Multi-Axis ViT |||
+| `maxvit_tiny` ⭐ | 30.1M | 2D |
+| `maxvit_small` ⭐ | 67.6M | 2D |
+| `maxvit_base` ⭐ | 119.1M | 2D |
+| **── Hybrid CNN-Transformer ──** |||
+| **FastViT** — Fast Hybrid CNN-ViT |||
+| `fastvit_t8` ⭐ | 4.0M | 2D |
+| `fastvit_t12` ⭐ | 6.8M | 2D |
+| `fastvit_s12` ⭐ | 8.8M | 2D |
+| `fastvit_sa12` ⭐ | 10.9M | 2D |
+| **CAFormer** — MetaFormer with Attention |||
+| `caformer_s18` ⭐ | 26.3M | 2D |
+| `caformer_s36` ⭐ | 39.2M | 2D |
+| `caformer_m36` ⭐ | 56.9M | 2D |
+| `poolformer_s12` ⭐ | 11.9M | 2D |
+| **EfficientViT** — Memory-Efficient ViT |||
+| `efficientvit_m0` ⭐ | 2.2M | 2D |
+| `efficientvit_m1` ⭐ | 2.6M | 2D |
+| `efficientvit_m2` ⭐ | 3.8M | 2D |
+| `efficientvit_b0` ⭐ | 2.1M | 2D |
+| `efficientvit_b1` ⭐ | 7.5M | 2D |
+| `efficientvit_b2` ⭐ | 21.8M | 2D |
+| `efficientvit_b3` ⭐ | 46.1M | 2D |
+| `efficientvit_l1` ⭐ | 49.5M | 2D |
+| `efficientvit_l2` ⭐ | 60.5M | 2D |
+| **── State Space Models ──** |||
+| **Mamba** — State Space Model |||
+| `mamba_1d` | 3.4M | 1D |
+| **Vision Mamba (ViM)** — 2D Mamba |||
+| `vim_tiny` | 6.6M | 2D |
+| `vim_small` | 51.1M | 2D |
+| `vim_base` | 201.4M | 2D |
+| **── Specialized Architectures ──** |||
+| **TCN** — Temporal Convolutional Network |||
+| `tcn_small` | 0.9M | 1D |
+| `tcn` | 6.9M | 1D |
+| `tcn_large` | 10.0M | 1D |
+| **ResNet3D** — 3D Residual Network |||
+| `resnet3d_18` | 33.2M | 3D |
+| `mc3_18` — Mixed Convolution 3D | 11.5M | 3D |
 | **U-Net** — U-shaped Network |||
-| `unet_regression` | 31.1M | 1D/2D/3D |
+| `unet_regression` | 31.0M | 1D/2D/3D |
+
 
 ⭐ = **Pretrained on ImageNet** (recommended for smaller datasets). Weights are downloaded automatically on first use.
 - **Cache location**: `~/.cache/torch/hub/checkpoints/` (or `./.torch_cache/` on HPC if home is not writable)
-- **Size**: ~20–350 MB per model depending on architecture
 - **Train from scratch**: Use `--no_pretrained` to disable pretrained weights
 
 **💡 HPC Users**: If compute nodes block internet, pre-download weights on the login node:
 
 ```bash
-# Run once on login node (with internet) — downloads ALL pretrained weights (~1.5 GB total)
+# Run once on login node (with internet) — downloads ALL pretrained weights
 python -c "
 import os
 os.environ['TORCH_HOME'] = '.torch_cache'  # Match WaveDL's HPC cache location
@@ -437,24 +471,56 @@ os.environ['TORCH_HOME'] = '.torch_cache'  # Match WaveDL's HPC cache location
 from torchvision import models as m
 from torchvision.models import video as v
 
-# Model name -> Weights class mapping
-weights = {
-    'resnet18': m.ResNet18_Weights, 'resnet50': m.ResNet50_Weights,
-    'efficientnet_b0': m.EfficientNet_B0_Weights, 'efficientnet_b1': m.EfficientNet_B1_Weights,
-    'efficientnet_b2': m.EfficientNet_B2_Weights, 'efficientnet_v2_s': m.EfficientNet_V2_S_Weights,
-    'efficientnet_v2_m': m.EfficientNet_V2_M_Weights, 'efficientnet_v2_l': m.EfficientNet_V2_L_Weights,
-    'mobilenet_v3_small': m.MobileNet_V3_Small_Weights, 'mobilenet_v3_large': m.MobileNet_V3_Large_Weights,
-    'regnet_y_400mf': m.RegNet_Y_400MF_Weights, 'regnet_y_800mf': m.RegNet_Y_800MF_Weights,
-    'regnet_y_1_6gf': m.RegNet_Y_1_6GF_Weights, 'regnet_y_3_2gf': m.RegNet_Y_3_2GF_Weights,
-    'regnet_y_8gf': m.RegNet_Y_8GF_Weights, 'swin_t': m.Swin_T_Weights, 'swin_s': m.Swin_S_Weights,
-    'swin_b': m.Swin_B_Weights, 'convnext_tiny': m.ConvNeXt_Tiny_Weights, 'densenet121': m.DenseNet121_Weights,
-}
-for name, w in weights.items():
-    getattr(m, name)(weights=w.DEFAULT); print(f'✓ {name}')
+# === TorchVision Models (use IMAGENET1K_V1 to match WaveDL) ===
+models = [
+    ('resnet18', m.ResNet18_Weights.IMAGENET1K_V1),
+    ('resnet50', m.ResNet50_Weights.IMAGENET1K_V1),
+    ('efficientnet_b0', m.EfficientNet_B0_Weights.IMAGENET1K_V1),
+    ('efficientnet_b1', m.EfficientNet_B1_Weights.IMAGENET1K_V1),
+    ('efficientnet_b2', m.EfficientNet_B2_Weights.IMAGENET1K_V1),
+    ('efficientnet_v2_s', m.EfficientNet_V2_S_Weights.IMAGENET1K_V1),
+    ('efficientnet_v2_m', m.EfficientNet_V2_M_Weights.IMAGENET1K_V1),
+    ('efficientnet_v2_l', m.EfficientNet_V2_L_Weights.IMAGENET1K_V1),
+    ('mobilenet_v3_small', m.MobileNet_V3_Small_Weights.IMAGENET1K_V1),
+    ('mobilenet_v3_large', m.MobileNet_V3_Large_Weights.IMAGENET1K_V1),
+    ('regnet_y_400mf', m.RegNet_Y_400MF_Weights.IMAGENET1K_V1),
+    ('regnet_y_800mf', m.RegNet_Y_800MF_Weights.IMAGENET1K_V1),
+    ('regnet_y_1_6gf', m.RegNet_Y_1_6GF_Weights.IMAGENET1K_V1),
+    ('regnet_y_3_2gf', m.RegNet_Y_3_2GF_Weights.IMAGENET1K_V1),
+    ('regnet_y_8gf', m.RegNet_Y_8GF_Weights.IMAGENET1K_V1),
+    ('swin_t', m.Swin_T_Weights.IMAGENET1K_V1),
+    ('swin_s', m.Swin_S_Weights.IMAGENET1K_V1),
+    ('swin_b', m.Swin_B_Weights.IMAGENET1K_V1),
+    ('convnext_tiny', m.ConvNeXt_Tiny_Weights.IMAGENET1K_V1),
+    ('densenet121', m.DenseNet121_Weights.IMAGENET1K_V1),
+]
+for name, w in models:
+    getattr(m, name)(weights=w); print(f'✓ {name}')
 
 # 3D video models
-v.r3d_18(weights=v.R3D_18_Weights.DEFAULT); print('✓ r3d_18')
-v.mc3_18(weights=v.MC3_18_Weights.DEFAULT); print('✓ mc3_18')
+v.r3d_18(weights=v.R3D_18_Weights.KINETICS400_V1); print('✓ r3d_18')
+v.mc3_18(weights=v.MC3_18_Weights.KINETICS400_V1); print('✓ mc3_18')
+
+# === Timm Models (MaxViT, FastViT, CAFormer, ConvNeXt V2) ===
+import timm
+
+timm_models = [
+    # MaxViT (no suffix - timm resolves to default)
+    'maxvit_tiny_tf_224', 'maxvit_small_tf_224', 'maxvit_base_tf_224',
+    # FastViT (no suffix)
+    'fastvit_t8', 'fastvit_t12', 'fastvit_s12', 'fastvit_sa12',
+    # CAFormer/PoolFormer (no suffix)
+    'caformer_s18', 'caformer_s36', 'caformer_m36', 'poolformer_s12',
+    # ConvNeXt V2 (no suffix)
+    'convnextv2_tiny',
+    # EfficientViT (no suffix)
+    'efficientvit_m0', 'efficientvit_m1', 'efficientvit_m2',
+    'efficientvit_b0', 'efficientvit_b1', 'efficientvit_b2', 'efficientvit_b3',
+    'efficientvit_l1', 'efficientvit_l2',
+]
+for name in timm_models:
+    timm.create_model(name, pretrained=True); print(f'✓ {name}')
+
 print('\\n✓ All pretrained weights cached!')
 "
 ```

{wavedl-1.5.7 → wavedl-1.6.1}/pyproject.toml

@@ -52,6 +52,7 @@ dependencies = [
     # Core ML stack
     "torch>=2.0.0",
     "torchvision>=0.15.0",
+    "timm>=0.9.0",  # Pretrained models (MaxViT, FastViT, CAFormer)
     "accelerate>=0.20.0",
     "numpy>=1.24.0",
     "scipy>=1.10.0",
@@ -70,6 +71,7 @@ dependencies = [
     # ONNX export
     "onnx>=1.14.0",
     "onnxruntime>=1.15.0",
+    "onnxscript>=0.1.0",  # Required by torch.onnx.export in PyTorch 2.1+
     # torch.compile backend (Linux only)
     "triton>=2.0.0; sys_platform == 'linux'",
 ]

{wavedl-1.5.7 → wavedl-1.6.1}/src/wavedl/__init__.py

@@ -18,7 +18,7 @@ For inference:
     # or: python -m wavedl.test --checkpoint best_checkpoint --data_path test.npz
 """
 
-__version__ = "1.5.7"
+__version__ = "1.6.1"
 __author__ = "Ductho Le"
 __email__ = "ductho.le@outlook.com"