metadata
language: en
license: mit
tags:
- image-classification
- imagenet
- multi-scale
- crystal-geometry
- david
datasets:
- imagenet-1k
metrics:
- accuracy
model-index:
- name: David-partial_shared-deep_efficiency
results:
- task:
type: image-classification
dataset:
name: ImageNet-1K
type: imagenet-1k
metrics:
- type: accuracy
value: 79.49
David: Multi-Scale Crystal Classifier
David is a multi-scale deep learning classifier that uses crystal geometry (pentachora/4-simplexes) as class prototypes with role-weighted similarity computation (Rose Loss).
Model Details
Architecture
- Preset: clip_vit_l14
- Sharing Mode: partial_shared
- Fusion Mode: deep_efficiency
- Scales: [384, 768, 1024, 1280]
- Feature Dim: 768
- Parameters: ~8.8M
Training Configuration
- Dataset: AbstractPhil/imagenet-clip-features-orderly
- Model Variant: clip_vit_l14
- Epochs: 20
- Batch Size: 1024
- Learning Rate: 0.01
- Rose Loss Weight: 0.1 → 0.5
- Cayley Loss: False
Performance
Best Results
- Validation Accuracy: 79.49%
- Best Epoch: 0
- Final Train Accuracy: 67.10%
Per-Scale Performance
- Scale 384: 79.49%
Usage
Repository Structure
AbstractPhil/gated-david/
├── weights/
│ ├── best_model.pth # Best model weights (PyTorch)
│ ├── best_model.safetensors # Best model weights (SafeTensors)
│ ├── best_model_metadata.json # Training metadata
│ ├── final_model.pth # Final epoch weights
│ ├── final_model.safetensors
│ ├── david_config.json # Model architecture config
│ └── train_config.json # Training configuration
├── runs/
│ └── events.out.tfevents.* # TensorBoard logs
├── README.md # This file
└── best_model.json # Performance summary
Loading the Model
from geovocab2.train.model.core.david import David, DavidArchitectureConfig
from huggingface_hub import hf_hub_download
# Download config
config_path = hf_hub_download(repo_id="AbstractPhil/gated-david",
filename="weights/david_config.json")
config = DavidArchitectureConfig.from_json(config_path)
# Download weights
weights_path = hf_hub_download(repo_id="AbstractPhil/gated-david",
filename="weights/best_model.pth")
# Initialize model
david = David.from_config(config)
checkpoint = torch.load(weights_path)
david.load_state_dict(checkpoint['model_state_dict'])
david.eval()
Inference
import torch
import torch.nn.functional as F
# Assuming you have CLIP features (512-dim for ViT-B/16)
features = get_clip_features(image) # [1, 512]
# Load anchors
anchors_dict = torch.load("anchors.pth")
# Forward pass
with torch.no_grad():
logits, _ = david(features, anchors_dict)
predictions = logits.argmax(dim=-1)
Architecture Overview
Multi-Scale Processing
David processes inputs at multiple scales (384, 768, 1024, 1280), allowing it to capture both coarse and fine-grained features.
Crystal Geometry
Each class is represented by a pentachoron (4-simplex) in embedding space with 5 vertices:
- Anchor: Primary class representative
- Need: Complementary direction
- Relation: Contextual alignment
- Purpose: Functional direction
- Observer: Meta-perspective
Rose Loss
Similarity computation uses role-weighted cosine similarities:
score = w_anchor * sim(z, anchor) + w_need * sim(z, need) + ...
Fusion Strategy
deep_efficiency: Intelligently combines predictions from multiple scales.
Training Details
Loss Components
- Cross-Entropy: Standard classification loss
- Rose Loss: Pentachora role-weighted margin loss (weight: 0.1→0.5)
- Cayley Loss: Geometric regularization (disabled)
Optimization
- Optimizer: AdamW
- Weight Decay: 1e-05
- Scheduler: cosine_restarts
- Gradient Clip: 5.0
- Mixed Precision: False
Citation
@software{david_classifier_2025,
title = {David: Multi-Scale Crystal Classifier},
author = {AbstractPhil},
year = {2025},
url = {https://huggingface.co/AbstractPhil/gated-david},
note = {Run ID: 20251012_040642}
}
License
MIT License
Acknowledgments
Built with crystal lattice geometry and multi-scale deep learning. Special thanks to Claude (Anthropic) for debugging assistance.
Generated on 2025-10-12 04:09:09