Epoch 50: 58.38%

README.md

@@ -5,7 +5,7 @@ tags:
 - cifar100
 - geometric-learning
 - fractal-encoding
-- trained
+- in-training
 - no-attention
 - no-cross-entropy
 datasets:
@@ -25,7 +25,7 @@ model-index:
       type: cifar100
     metrics:
     - type: accuracy
-      value: 67.00
+      value: 58.38
       name: Test Accuracy
       verified: false
 ---
@@ -34,9 +34,9 @@ model-index:
 
 **Geometric Basin Classification for CIFAR-100**
 
-🎉 **Training Complete** 🎉
+🚧 **Training in Progress** 🚧
 
-Final Status: Epoch 200/200
+Current Status: Epoch 50/200
 
 ---
 
@@ -44,24 +44,24 @@ Final Status: Epoch 200/200
 
 | Metric | Value |
 |--------|-------|
-| **Best Test Accuracy** | **67.00%** |
-| **Best Epoch** | 190 |
-| **Current Train Accuracy** | 63.49% |
-| **Current Test Accuracy** | 64.32% |
-| **Current α (Cantor param)** | 0.4422 |
-| **Total Parameters** | 45,161,489 |
-| **Training Time** | 1:10:27 |
+| **Best Test Accuracy** | **58.38%** |
+| **Best Epoch** | 50 |
+| **Current Train Accuracy** | 62.04% |
+| **Current Test Accuracy** | 58.38% |
+| **Current α (Cantor param)** | 0.4225 |
+| **Total Parameters** | 45,235,067 |
+| **Training Time** | 0:12:52 |
 
 ### Comparison to State-of-the-Art
 
 | Model | Accuracy | Status |
 |-------|----------|--------|
-| **geo-beatrix (this model)** | **67.00%** | ✅ Complete |
+| **geo-beatrix (this model)** | **58.38%** | 🔄 Training |
 | vit-beatrix-dualstream | 66.0% | Vision Transformer + Cross-Entropy |
 | CLIP ViT-L/14 (zero-shot) | ~63-65% | 400M image-text pairs |
 | CLIP ViT-B/32 (zero-shot) | ~63.5% | Vision Transformer |
 
-✅ **geo-beatrix has surpassed all baselines!**
+🎯 **Current target**: Beat vit-beatrix (66.0%) - Currently -7.62%
 
 ---
 
@@ -71,7 +71,7 @@ Final Status: Epoch 200/200
 - **Channels**: 64 → 128 → 256 → 512 → 1024
 - **Positional Encoding**: Devil's Staircase (Cantor function, 1883)
 - **PE Levels**: 20
-- **PE Features/Level**: 4
+- **PE Features/Level**: 10
 - **Classification**: Geometric Basin Compatibility (NO cross-entropy)
 - **Attention Mechanisms**: NONE
 
@@ -90,9 +90,9 @@ Final Status: Epoch 200/200
   "weight_decay": 0.05,
   "warmup_epochs": 10,
   "pe_levels": 20,
-  "pe_features_per_level": 4,
+  "pe_features_per_level": 10,
   "dropout": 0.1,
-  "upload_every_n_epochs": 10,
+  "upload_every_n_epochs": 50,
   "alphamix": {
     "enabled": true,
     "range": [
@@ -107,7 +107,7 @@ Final Status: Epoch 200/200
   "loss_function": "Geometric Basin Compatibility",
   "cross_entropy": false,
   "attention_mechanisms": false,
-  "timestamp": "20251009_222052"
+  "timestamp": "20251009_234907"
 }
 ```
 
@@ -116,22 +116,27 @@ Final Status: Epoch 200/200
 ## Files Structure
 
 ```
-weights/geo-beatrix/20251009_222052/
-  ├── model.pt                 (best checkpoint - PyTorch)
-  ├── model.safetensors        (best checkpoint - SafeTensors)
-  ├── config.json              (model configuration)
-  ├── training_log.txt         (training log)
-  └── checkpoints/
-      ├── checkpoint_epoch_10.safetensors
-      ├── checkpoint_epoch_20.safetensors
-      └── checkpoint_epoch_30.safetensors
-      (snapshots every 10 epochs)
-
-runs/geo-beatrix/20251009_222052/
-  ├── events.out.tfevents.*    (TensorBoard logs)
-  └── metrics.csv              (training metrics)
+├── model.pt                 (BEST overall model - easy access!)
+├── model.safetensors        (BEST overall model - easy access!)
+├── best_model_info.json     (which epoch/run this came from)
+├── README.md
+├── weights/geo-beatrix/20251009_234907/
+│   ├── model.pt                 (best from this training run)
+│   ├── model.safetensors        (best from this training run)
+│   ├── config.json
+│   ├── training_log.txt
+│   └── checkpoints/
+│       ├── checkpoint_epoch_10.safetensors
+│       ├── checkpoint_epoch_20.safetensors
+│       └── checkpoint_epoch_30.safetensors
+│       (snapshots every 50 epochs)
+└── runs/geo-beatrix/20251009_234907/
+    ├── events.out.tfevents.*    (TensorBoard logs)
+    └── metrics.csv              (training metrics)
 ```
 
+**Note**: The root `model.pt` and `model.safetensors` always contain the best model across all training runs!
+
 ---
 
 ## Usage
@@ -140,34 +145,34 @@ runs/geo-beatrix/20251009_222052/
 from huggingface_hub import hf_hub_download
 import torch
 
-# Download best model (SafeTensors - recommended)
+# EASIEST: Download BEST overall model from root (recommended!)
 from safetensors.torch import load_file
 model_path = hf_hub_download(
     repo_id="AbstractPhil/geo-beatrix",
-    filename="weights/geo-beatrix/20251009_222052/model.safetensors"
+    filename="model.safetensors"
 )
 state_dict = load_file(model_path)
 # model.load_state_dict(state_dict)
 
-# Or download PyTorch checkpoint (includes optimizer state)
-checkpoint_path = hf_hub_download(
+# Check which epoch/run the best model came from
+info_path = hf_hub_download(
     repo_id="AbstractPhil/geo-beatrix",
-    filename="weights/geo-beatrix/20251009_222052/model.pt"
+    filename="best_model_info.json"
 )
-checkpoint = torch.load(checkpoint_path)
-# model.load_state_dict(checkpoint['model_state_dict'])
+with open(info_path) as f:
+    best_info = json.load(f)
+    print(f"Best model: epoch {best_info['epoch']}, {best_info['test_accuracy']:.2f}%")
 
-# Download specific epoch checkpoint
-epoch_checkpoint = hf_hub_download(
+# Or download from specific training run
+model_path = hf_hub_download(
     repo_id="AbstractPhil/geo-beatrix",
-    filename="weights/geo-beatrix/20251009_222052/checkpoints/checkpoint_epoch_100.safetensors"
+    filename="weights/geo-beatrix/20251009_234907/model.safetensors"
 )
 
-# Download TensorBoard logs for visualization
-import glob
-tensorboard_files = hf_hub_download(
+# Download specific epoch checkpoint
+epoch_checkpoint = hf_hub_download(
     repo_id="AbstractPhil/geo-beatrix",
-    filename="runs/geo-beatrix/20251009_222052/events.out.tfevents.*"
+    filename="weights/geo-beatrix/20251009_234907/checkpoints/checkpoint_epoch_100.safetensors"
 )
 ```
 
@@ -176,27 +181,23 @@ tensorboard_files = hf_hub_download(
 ## Training History
 
 ### Best Checkpoint
-- Epoch: 190
-- Train Acc: 61.47%
-- Test Acc: 67.00%
-- Alpha: 0.4425
-- Loss: 0.5031
+- Epoch: 50
+- Train Acc: 62.04%
+- Test Acc: 58.38%
+- Alpha: 0.4225
+- Loss: 1.1254
 
 ### Latest 5 Epochs
 
-- **Epoch 196**: Train 63.32%, Test 0.00%, α=0.4422, Loss=0.4936
-- **Epoch 197**: Train 65.20%, Test 0.00%, α=0.4422, Loss=0.5069
-- **Epoch 198**: Train 60.80%, Test 0.00%, α=0.4422, Loss=0.4929
-- **Epoch 199**: Train 58.87%, Test 0.00%, α=0.4422, Loss=0.4874
-- **Epoch 200**: Train 63.49%, Test 64.32%, α=0.4422, Loss=0.5049
+- **Epoch 46**: Train 59.73%, Test 0.00%, α=0.4210, Loss=1.1882
+- **Epoch 47**: Train 59.23%, Test 0.00%, α=0.4176, Loss=1.1413
+- **Epoch 48**: Train 60.73%, Test 0.00%, α=0.4184, Loss=1.1374
+- **Epoch 49**: Train 60.30%, Test 0.00%, α=0.4126, Loss=1.1244
+- **Epoch 50**: Train 62.04%, Test 58.38%, α=0.4225, Loss=1.1254
 
 ### Training Milestones
-- 🎯 **50% Accuracy** reached at epoch 45
-- 🎯 **60% Accuracy** reached at epoch 70
-- 🏆 **Beat vit-beatrix (66.0%)** at epoch 140
-- 🚀 **67% Accuracy** reached at epoch 190
-- 📊 **α ≥ 0.40** reached at epoch 10
-- 📊 **α ≥ 0.44** (near triadic equilibrium) at epoch 116
+- 🎯 **50% Accuracy** reached at epoch 35
+- 📊 **α ≥ 0.40** reached at epoch 9
 
 ---