Epoch 10: 25.03%

README.md

@@ -25,7 +25,7 @@ model-index:
       type: cifar100
     metrics:
     - type: accuracy
-      value: 46.26
+      value: 25.03
       name: Test Accuracy
       verified: false
 ---
@@ -36,7 +36,7 @@ model-index:
 
 🚧 **Training in Progress** 🚧
 
-Current Status: Epoch 30/200
+Current Status: Epoch 10/200
 
 ---
 
@@ -44,24 +44,24 @@ Current Status: Epoch 30/200
 
 | Metric | Value |
 |--------|-------|
-| **Best Test Accuracy** | **46.26%** |
-| **Best Epoch** | 30 |
-| **Current Train Accuracy** | 46.33% |
-| **Current Test Accuracy** | 46.26% |
-| **Current α (Cantor param)** | 0.4042 |
+| **Best Test Accuracy** | **25.03%** |
+| **Best Epoch** | 10 |
+| **Current Train Accuracy** | 24.79% |
+| **Current Test Accuracy** | 25.03% |
+| **Current α (Cantor param)** | 0.4097 |
 | **Total Parameters** | 45,161,489 |
-| **Training Time** | 0:09:24 |
+| **Training Time** | 0:02:43 |
 
 ### Comparison to State-of-the-Art
 
 | Model | Accuracy | Status |
 |-------|----------|--------|
-| **geo-beatrix (this model)** | **46.26%** | 🔄 Training |
+| **geo-beatrix (this model)** | **25.03%** | 🔄 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 |
 
-🎯 **Current target**: Beat vit-beatrix (66.0%) - Currently -19.74%
+🎯 **Current target**: Beat vit-beatrix (66.0%) - Currently -40.97%
 
 ---
 
@@ -106,42 +106,29 @@ Current Status: Epoch 30/200
   "loss_function": "Geometric Basin Compatibility",
   "cross_entropy": false,
   "attention_mechanisms": false,
-  "timestamp": "20251009_215951"
+  "timestamp": "20251009_221125"
 }
 ```
 
 ---
 
-## Training History
-
-### Training Progress
+## Files Structure
 
 ```
-Epochs:     30/200 (███                 ) 15.0%
-Best Acc:   46.26%
-Current α:  0.4042 → Target: 0.50 (triadic equilibrium)
-Status:     🔄 Training In Progress
+weights/geo-beatrix/20251009_221125/
+  ├── model.pt                 (best checkpoint - PyTorch)
+  ├── model.safetensors        (best checkpoint - SafeTensors)
+  ├── config.json              (model configuration)
+  ├── training_log.txt         (training log)
+  └── checkpoints/
+      ├── checkpoint_epoch_*.pt
+      └── checkpoint_epoch_*.safetensors
+
+runs/geo-beatrix/20251009_221125/
+  ├── events.out.tfevents.*    (TensorBoard logs)
+  └── metrics.csv              (training metrics)
 ```
 
-### Best Checkpoint
-- Epoch: 30
-- Train Acc: 46.33%
-- Test Acc: 46.26%
-- Alpha: 0.4042
-- Loss: 1.6819
-
-### Latest 5 Epochs
-
-- **Epoch 26**: Train 43.67%, Test 0.00%, α=0.4018, Loss=1.7948
-- **Epoch 27**: Train 44.03%, Test 0.00%, α=0.4000, Loss=1.7660
-- **Epoch 28**: Train 44.65%, Test 0.00%, α=0.4013, Loss=1.7252
-- **Epoch 29**: Train 45.45%, Test 0.00%, α=0.4041, Loss=1.7425
-- **Epoch 30**: Train 46.33%, Test 46.26%, α=0.4042, Loss=1.6819
-
-### Training Milestones
-- 📊 **α ≥ 0.40** reached at epoch 10
-
-
 ---
 
 ## Usage
@@ -150,11 +137,11 @@ Status:     🔄 Training In Progress
 from huggingface_hub import hf_hub_download
 import torch
 
-# Download best model (safetensors format - recommended)
+# Download best model (SafeTensors - recommended)
 from safetensors.torch import load_file
 model_path = hf_hub_download(
     repo_id="AbstractPhil/geo-beatrix",
-    filename="model.safetensors"
+    filename="weights/geo-beatrix/20251009_221125/model.safetensors"
 )
 state_dict = load_file(model_path)
 # model.load_state_dict(state_dict)
@@ -162,7 +149,7 @@ state_dict = load_file(model_path)
 # Or download PyTorch checkpoint (includes optimizer state)
 checkpoint_path = hf_hub_download(
     repo_id="AbstractPhil/geo-beatrix",
-    filename="model.pt"
+    filename="weights/geo-beatrix/20251009_221125/model.pt"
 )
 checkpoint = torch.load(checkpoint_path)
 # model.load_state_dict(checkpoint['model_state_dict'])
@@ -170,12 +157,40 @@ checkpoint = torch.load(checkpoint_path)
 # Download specific epoch checkpoint
 epoch_checkpoint = hf_hub_download(
     repo_id="AbstractPhil/geo-beatrix",
-    filename="checkpoints/checkpoint_epoch_100.safetensors"
+    filename="weights/geo-beatrix/20251009_221125/checkpoints/checkpoint_epoch_100.safetensors"
+)
+
+# Download TensorBoard logs
+tensorboard_log = hf_hub_download(
+    repo_id="AbstractPhil/geo-beatrix",
+    filename="runs/geo-beatrix/20251009_221125/events.out.tfevents.*"
 )
 ```
 
 ---
 
+## Training History
+
+### Best Checkpoint
+- Epoch: 10
+- Train Acc: 24.79%
+- Test Acc: 25.03%
+- Alpha: 0.4097
+- Loss: 2.2922
+
+### Latest 5 Epochs
+
+- **Epoch 6**: Train 11.86%, Test 0.00%, α=0.3382, Loss=2.7329
+- **Epoch 7**: Train 14.90%, Test 0.00%, α=0.3664, Loss=2.5766
+- **Epoch 8**: Train 18.45%, Test 0.00%, α=0.3868, Loss=2.4434
+- **Epoch 9**: Train 22.42%, Test 0.00%, α=0.4014, Loss=2.4042
+- **Epoch 10**: Train 24.79%, Test 25.03%, α=0.4097, Loss=2.2922
+
+### Training Milestones
+- 📊 **α ≥ 0.40** reached at epoch 9
+
+---
+
 ## Innovation
 
 ✅ **NO attention mechanisms**