Shanmuk4622 commited on
Commit
0eb0e2c
·
verified ·
1 Parent(s): db9b5bc

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +37 -4
README.md CHANGED
@@ -21,6 +21,22 @@ co2_eq_emissions:
21
  hardware_used: NVIDIA GeForce GTX 1080 Ti
22
  dataset_info:
23
  dataset_size: "~450,000 images – 300 classes (224 px)"
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
24
  ---
25
 
26
  # EDEN-ConvNeXtV2-Custom-ImageNet300 — *SOTA Optimized*
@@ -28,7 +44,9 @@ dataset_info:
28
  > **Primary KPI:** EAG (Energy-to-Accuracy Gradient) = `-8.4494e-11` ΔAcc/ΔJoules
29
 
30
  ## Abstract
31
- This model is part of **Project EDEN (Energy-Driven Evolution of Networks)**, implementing the **E2AM (Energy Efficient Advanced Model)** Framework. The goal is to shift AI benchmarking from pure accuracy to *Green SOTA* — maximizing predictive power per Joule consumed.
 
 
32
 
33
  **Applied Technique:** Phase 2 – Progressive Unfreezing + AMP (E2AM SOTA)
34
 
@@ -73,15 +91,30 @@ This model is part of **Project EDEN (Energy-Driven Evolution of Networks)**, im
73
  | Estimated CO₂ | 2.0711 kg CO₂e |
74
  | Training Log | `test1\eden_unfrozen_custom_imagenet_convnext.csv` |
75
 
76
- ## Cite This Research
77
- If you use this model, please cite the **EDEN / E2AM Framework**:
 
 
 
 
 
 
 
78
 
 
 
 
 
 
 
 
 
79
  ```bibtex
80
  @misc{eden2025,
81
  title = {Project EDEN: Energy-Driven Evolution of Networks},
82
  author = {EDEN Research Team},
83
  year = {2025},
84
- note = {Hugging Face Organization: ProjectEDEN},
85
  url = {https://huggingface.co/Shanmuk4622}
86
  }
87
  ```
 
21
  hardware_used: NVIDIA GeForce GTX 1080 Ti
22
  dataset_info:
23
  dataset_size: "~450,000 images – 300 classes (224 px)"
24
+ model-index:
25
+ - name: EDEN-ConvNeXtV2-Custom-ImageNet300
26
+ results:
27
+ - task:
28
+ type: image-classification
29
+ name: Image Classification
30
+ dataset:
31
+ name: Custom-ImageNet300
32
+ type: imagenet
33
+ metrics:
34
+ - type: accuracy
35
+ value: 0.9881
36
+ name: Accuracy
37
+ - type: f1
38
+ value: 0.9876
39
+ name: F1 Score
40
  ---
41
 
42
  # EDEN-ConvNeXtV2-Custom-ImageNet300 — *SOTA Optimized*
 
44
  > **Primary KPI:** EAG (Energy-to-Accuracy Gradient) = `-8.4494e-11` ΔAcc/ΔJoules
45
 
46
  ## Abstract
47
+ This model is part of **Project EDEN (Energy-Driven Evolution of Networks)**, implementing the
48
+ **E2AM (Energy Efficient Advanced Model)** Framework. The goal is to shift AI benchmarking from
49
+ pure accuracy to *Green SOTA* — maximising predictive power per Joule consumed.
50
 
51
  **Applied Technique:** Phase 2 – Progressive Unfreezing + AMP (E2AM SOTA)
52
 
 
91
  | Estimated CO₂ | 2.0711 kg CO₂e |
92
  | Training Log | `test1\eden_unfrozen_custom_imagenet_convnext.csv` |
93
 
94
+ ## 📊 Training Visualizations
95
+
96
+ ### Accuracy & Energy over Training
97
+ > Green = accuracy (left axis) · Orange dashed = cumulative energy (right axis)
98
+
99
+ ![Training Curve](training_curve.png)
100
+
101
+ ### EAG Metric Trajectory
102
+ > EAG = ΔAccuracy / ΔJoules — positive means learning more per Joule than baseline
103
 
104
+ ![EAG Curve](eag_curve.png)
105
+
106
+ ### Project-Wide Overview
107
+ *All EDEN models: energy vs accuracy*
108
+
109
+ ![Collection Overview](https://huggingface.co/Shanmuk4622/EDEN-Core-Scripts/resolve/main/energy_accuracy_overview.png)
110
+
111
+ ## Cite This Research
112
  ```bibtex
113
  @misc{eden2025,
114
  title = {Project EDEN: Energy-Driven Evolution of Networks},
115
  author = {EDEN Research Team},
116
  year = {2025},
117
+ note = {Hugging Face: Shanmuk4622},
118
  url = {https://huggingface.co/Shanmuk4622}
119
  }
120
  ```