Hugging Face's logo

Datasets:
Shanmuk4622
/
E2AM_ResNet50

Modalities:
Image
Tabular
Text
Formats:
csv
Languages:
English
Size:
< 1K
Tags:
energy-aware-training
image-classification
model-efficiency
ablation-study
resnet50
Libraries:
Datasets
pandas
Polars
License:
Dataset card Data Studio Files
xet
 Community
1
Shanmuk4622 commited on
Commit
36f2bd4
·
verified ·
1 Parent(s): 6ff7605

progress push (15 files)

Browse files
Files changed (15) hide show
  1. README.md +28 -27
  2. comparison_plots/cifar10/accuracy_bar.png +2 -2
  3. comparison_plots/cifar10/accuracy_vs_energy_scatter.png +2 -2
  4. comparison_plots/cifar10/co2_bar.png +2 -2
  5. comparison_plots/cifar10/energy_bar.png +2 -2
  6. comparison_plots/cifar100/accuracy_bar.png +2 -2
  7. comparison_plots/cifar100/accuracy_vs_energy_scatter.png +2 -2
  8. comparison_plots/cifar100/co2_bar.png +2 -2
  9. comparison_plots/cifar100/energy_bar.png +2 -2
  10. comparison_plots/cross_dataset/accuracy_by_variant_across_datasets.png +2 -2
  11. comparison_plots/cross_dataset/energy_by_variant_across_datasets.png +2 -2
  12. comparison_plots/tiny_imagenet/accuracy_bar.png +2 -2
  13. comparison_plots/tiny_imagenet/accuracy_vs_energy_scatter.png +2 -2
  14. comparison_plots/tiny_imagenet/co2_bar.png +2 -2
  15. comparison_plots/tiny_imagenet/energy_bar.png +2 -2
README.md CHANGED
@@ -23,7 +23,7 @@ Energy-aware training ablation study for **ResNet-50** across three image-classi
23
 
24
  Each dataset has 15 training variants (8 individual-method M0..M7, 7 cumulative ablation C0..C6) at 50 epochs, plus a 5-variant deployment pipeline (FP32 baseline, structured pruning, pruning+finetune, INT8 quantization, pruned+INT8).
25
 
26
- **Status**: 44 completed variants, 0 partial.
27
 
28
  ## Quick links
29
 
@@ -39,7 +39,7 @@ Each dataset has 15 training variants (8 individual-method M0..M7, 7 cumulative
39
  | Dataset | Best variant | Top-1 | Top-5 | Energy (kWh) | CO₂ (kg) | Time (sec) |
40
  |---|---|---|---|---|---|---|
41
  | CIFAR-10 | M5_adaptive_lr_only | 0.9397 | — | 0.1556 | 0.0739 | 7368 |
42
- | CIFAR-100 | M5_adaptive_lr_only | 0.7482 | 0.9368 | 0.1587 | 0.0754 | 7578 |
43
  | Tiny-ImageNet | C5_cache_amp_gradaccum_adaptivelr_l1 | 0.6742 | 0.8651 | 0.5231 | 0.2485 | 24653 |
44
 
45
  ## Cross-dataset comparison
@@ -64,6 +64,7 @@ How the same training variants behave across CIFAR-10, CIFAR-100, and Tiny-Image
64
  |---|---|---|---|---|---|---|---|
65
  | M0_baseline_fp32 | 50 | 0.8298 | — | 0.1590 | 0.0755 | 7588 | completed |
66
  | M1_cache_only | 50 | 0.8543 | — | 0.1561 | 0.0741 | 7369 | completed |
 
67
  | M3_grad_accum_only | 50 | 0.8720 | — | 0.1552 | 0.0737 | 7332 | completed |
68
  | M4_l1_sparsity_only | 50 | 0.8387 | — | 0.1578 | 0.0750 | 7447 | completed |
69
  | M5_adaptive_lr_only | 50 | 0.9397 | — | 0.1556 | 0.0739 | 7368 | completed |
@@ -96,26 +97,26 @@ How the same training variants behave across CIFAR-10, CIFAR-100, and Tiny-Image
96
 
97
  | Variant | Epochs | Top-1 | Top-5 | Energy (kWh) | CO₂ (kg) | Time (s) | Status |
98
  |---|---|---|---|---|---|---|---|
99
- | M0_baseline_fp32 | 50 | 0.5913 | 0.8691 | 0.1569 | 0.0745 | 7414 | completed |
100
- | M1_cache_only | 50 | 0.5947 | 0.8712 | 0.1567 | 0.0744 | 7415 | completed |
101
- | M2_amp_only | 50 | 0.6081 | 0.8730 | 0.0702 | 0.0334 | 3377 | completed |
102
- | M3_grad_accum_only | 50 | 0.6228 | 0.8836 | 0.1559 | 0.0741 | 7380 | completed |
103
- | M4_l1_sparsity_only | 50 | 0.5897 | 0.8637 | 0.1581 | 0.0751 | 7475 | completed |
104
- | M5_adaptive_lr_only | 50 | 0.7482 | 0.9368 | 0.1587 | 0.0754 | 7578 | completed |
105
- | M6_eag_only | 50 | 0.5821 | 0.8643 | 0.1565 | 0.0743 | 7412 | completed |
106
- | M7_full_e2am | 50 | 0.7008 | 0.9097 | 0.0648 | 0.0308 | 3123 | completed |
107
 
108
  **C-matrix (cumulative ablation)**
109
 
110
  | Variant | Epochs | Top-1 | Top-5 | Energy (kWh) | CO₂ (kg) | Time (s) | Status |
111
  |---|---|---|---|---|---|---|---|
112
  | C0_baseline | 50 | 0.5822 | 0.8628 | 0.1657 | 0.0787 | 7918 | completed |
113
- | C1_cache | 50 | 0.5916 | 0.8691 | 0.1621 | 0.0770 | 7754 | completed |
114
- | C2_cache_amp | 50 | 0.6307 | 0.8814 | 0.0670 | 0.0318 | 3233 | completed |
115
- | C3_cache_amp_gradaccum | 50 | 0.6306 | 0.8829 | 0.0659 | 0.0313 | 3188 | completed |
116
- | C4_cache_amp_gradaccum_adaptivelr | 50 | 0.6918 | 0.9044 | 0.0659 | 0.0313 | 3190 | completed |
117
- | C5_cache_amp_gradaccum_adaptivelr_l1 | 50 | 0.6753 | 0.8976 | 0.0667 | 0.0317 | 3225 | completed |
118
- | C6_full_e2am | 50 | 0.6753 | 0.8976 | 0.0670 | 0.0318 | 3224 | completed |
119
 
120
  ![accuracy_bar.png](comparison_plots/cifar100/accuracy_bar.png)
121
 
@@ -131,24 +132,24 @@ How the same training variants behave across CIFAR-10, CIFAR-100, and Tiny-Image
131
 
132
  | Variant | Epochs | Top-1 | Top-5 | Energy (kWh) | CO₂ (kg) | Time (s) | Status |
133
  |---|---|---|---|---|---|---|---|
134
- | M0_baseline_fp32 | 50 | 0.4187 | 0.7041 | 1.2832 | 0.6095 | 60871 | completed |
135
- | M1_cache_only | 50 | 0.4027 | 0.6833 | 1.2088 | 0.5742 | 57541 | completed |
136
- | M2_amp_only | 50 | 0.4889 | 0.7608 | 0.5128 | 0.2436 | 24268 | completed |
137
- | M3_grad_accum_only | 50 | 0.5141 | 0.7703 | 1.3058 | 0.6203 | 61979 | completed |
138
  | M4_l1_sparsity_only | 50 | 0.4112 | 0.6874 | 1.2378 | 0.5880 | 58649 | completed |
139
  | M5_adaptive_lr_only | 50 | 0.6243 | 0.8426 | 1.1917 | 0.5661 | 56539 | completed |
140
- | M6_eag_only | 50 | 0.3996 | 0.6840 | 1.2605 | 0.5987 | 59534 | completed |
141
- | M7_full_e2am | 50 | 0.6641 | 0.8591 | 0.5423 | 0.2576 | 25681 | completed |
142
 
143
  **C-matrix (cumulative ablation)**
144
 
145
  | Variant | Epochs | Top-1 | Top-5 | Energy (kWh) | CO₂ (kg) | Time (s) | Status |
146
  |---|---|---|---|---|---|---|---|
147
- | C0_baseline | 50 | 0.4121 | 0.6944 | 1.3775 | 0.6543 | 65154 | completed |
148
  | C1_cache | 50 | 0.4213 | 0.7007 | 1.3177 | 0.6259 | 62533 | completed |
149
- | C2_cache_amp | 50 | 0.4993 | 0.7619 | 0.4929 | 0.2341 | 22910 | completed |
150
  | C3_cache_amp_gradaccum | 50 | 0.5505 | 0.7980 | 0.5049 | 0.2398 | 23604 | completed |
151
- | C4_cache_amp_gradaccum_adaptivelr | 50 | 0.6667 | 0.8611 | 0.5437 | 0.2583 | 25771 | completed |
152
  | C5_cache_amp_gradaccum_adaptivelr_l1 | 50 | 0.6742 | 0.8651 | 0.5231 | 0.2485 | 24653 | completed |
153
  | C6_full_e2am | 49 | 0.6716 | 0.8621 | 0.5163 | 0.2453 | 24352 | completed |
154
 
@@ -162,7 +163,7 @@ How the same training variants behave across CIFAR-10, CIFAR-100, and Tiny-Image
162
 
163
  ## Deployment results
164
 
165
- No deployment results in this repo yet.
166
 
167
  ## Methodology
168
 
@@ -232,5 +233,5 @@ Each variant directory has a `config.yaml` with the exact configuration used. To
232
 
233
  ---
234
 
235
- _This README was auto-generated on 2026-06-06 12:34 UTC._
236
  _Source repo: Shanmuk4622/E2AM_ResNet50_
 
23
 
24
  Each dataset has 15 training variants (8 individual-method M0..M7, 7 cumulative ablation C0..C6) at 50 epochs, plus a 5-variant deployment pipeline (FP32 baseline, structured pruning, pruning+finetune, INT8 quantization, pruned+INT8).
25
 
26
+ **Status**: 45 completed variants, 0 partial.
27
 
28
  ## Quick links
29
 
 
39
  | Dataset | Best variant | Top-1 | Top-5 | Energy (kWh) | CO₂ (kg) | Time (sec) |
40
  |---|---|---|---|---|---|---|
41
  | CIFAR-10 | M5_adaptive_lr_only | 0.9397 | — | 0.1556 | 0.0739 | 7368 |
42
+ | CIFAR-100 | M5_adaptive_lr_only | 0.7482 | 0.9358 | 0.1587 | 0.0754 | 7578 |
43
  | Tiny-ImageNet | C5_cache_amp_gradaccum_adaptivelr_l1 | 0.6742 | 0.8651 | 0.5231 | 0.2485 | 24653 |
44
 
45
  ## Cross-dataset comparison
 
64
  |---|---|---|---|---|---|---|---|
65
  | M0_baseline_fp32 | 50 | 0.8298 | — | 0.1590 | 0.0755 | 7588 | completed |
66
  | M1_cache_only | 50 | 0.8543 | — | 0.1561 | 0.0741 | 7369 | completed |
67
+ | M2_amp_only | 50 | 0.8744 | — | 0.0622 | 0.0295 | 3006 | completed |
68
  | M3_grad_accum_only | 50 | 0.8720 | — | 0.1552 | 0.0737 | 7332 | completed |
69
  | M4_l1_sparsity_only | 50 | 0.8387 | — | 0.1578 | 0.0750 | 7447 | completed |
70
  | M5_adaptive_lr_only | 50 | 0.9397 | — | 0.1556 | 0.0739 | 7368 | completed |
 
97
 
98
  | Variant | Epochs | Top-1 | Top-5 | Energy (kWh) | CO₂ (kg) | Time (s) | Status |
99
  |---|---|---|---|---|---|---|---|
100
+ | M0_baseline_fp32 | 50 | 0.5913 | 0.8446 | 0.1569 | 0.0745 | 7414 | completed |
101
+ | M1_cache_only | 50 | 0.5947 | 0.8582 | 0.1567 | 0.0744 | 7415 | completed |
102
+ | M2_amp_only | 50 | 0.6081 | 0.8652 | 0.0702 | 0.0334 | 3377 | completed |
103
+ | M3_grad_accum_only | 50 | 0.6228 | 0.8766 | 0.1559 | 0.0741 | 7380 | completed |
104
+ | M4_l1_sparsity_only | 50 | 0.5897 | 0.8576 | 0.1581 | 0.0751 | 7475 | completed |
105
+ | M5_adaptive_lr_only | 50 | 0.7482 | 0.9358 | 0.1587 | 0.0754 | 7578 | completed |
106
+ | M6_eag_only | 50 | 0.5821 | 0.8440 | 0.1565 | 0.0743 | 7412 | completed |
107
+ | M7_full_e2am | 50 | 0.7008 | 0.9087 | 0.0648 | 0.0308 | 3123 | completed |
108
 
109
  **C-matrix (cumulative ablation)**
110
 
111
  | Variant | Epochs | Top-1 | Top-5 | Energy (kWh) | CO₂ (kg) | Time (s) | Status |
112
  |---|---|---|---|---|---|---|---|
113
  | C0_baseline | 50 | 0.5822 | 0.8628 | 0.1657 | 0.0787 | 7918 | completed |
114
+ | C1_cache | 50 | 0.5916 | 0.8505 | 0.1621 | 0.0770 | 7754 | completed |
115
+ | C2_cache_amp | 50 | 0.6307 | 0.8665 | 0.0670 | 0.0318 | 3233 | completed |
116
+ | C3_cache_amp_gradaccum | 50 | 0.6306 | 0.8761 | 0.0659 | 0.0313 | 3188 | completed |
117
+ | C4_cache_amp_gradaccum_adaptivelr | 50 | 0.6918 | 0.9042 | 0.0659 | 0.0313 | 3190 | completed |
118
+ | C5_cache_amp_gradaccum_adaptivelr_l1 | 50 | 0.6753 | 0.8974 | 0.0667 | 0.0317 | 3225 | completed |
119
+ | C6_full_e2am | 50 | 0.6753 | 0.8974 | 0.0670 | 0.0318 | 3224 | completed |
120
 
121
  ![accuracy_bar.png](comparison_plots/cifar100/accuracy_bar.png)
122
 
 
132
 
133
  | Variant | Epochs | Top-1 | Top-5 | Energy (kWh) | CO₂ (kg) | Time (s) | Status |
134
  |---|---|---|---|---|---|---|---|
135
+ | M0_baseline_fp32 | 50 | 0.4187 | 0.6629 | 1.2832 | 0.6095 | 60871 | completed |
136
+ | M1_cache_only | 50 | 0.4027 | 0.6637 | 1.2088 | 0.5742 | 57541 | completed |
137
+ | M2_amp_only | 50 | 0.4889 | 0.7445 | 0.5128 | 0.2436 | 24268 | completed |
138
+ | M3_grad_accum_only | 50 | 0.5141 | 0.7502 | 1.3058 | 0.6203 | 61979 | completed |
139
  | M4_l1_sparsity_only | 50 | 0.4112 | 0.6874 | 1.2378 | 0.5880 | 58649 | completed |
140
  | M5_adaptive_lr_only | 50 | 0.6243 | 0.8426 | 1.1917 | 0.5661 | 56539 | completed |
141
+ | M6_eag_only | 50 | 0.3996 | 0.6543 | 1.2605 | 0.5987 | 59534 | completed |
142
+ | M7_full_e2am | 50 | 0.6641 | 0.8580 | 0.5423 | 0.2576 | 25681 | completed |
143
 
144
  **C-matrix (cumulative ablation)**
145
 
146
  | Variant | Epochs | Top-1 | Top-5 | Energy (kWh) | CO₂ (kg) | Time (s) | Status |
147
  |---|---|---|---|---|---|---|---|
148
+ | C0_baseline | 50 | 0.4121 | 0.6262 | 1.3775 | 0.6543 | 65154 | completed |
149
  | C1_cache | 50 | 0.4213 | 0.7007 | 1.3177 | 0.6259 | 62533 | completed |
150
+ | C2_cache_amp | 50 | 0.4993 | 0.7211 | 0.4929 | 0.2341 | 22910 | completed |
151
  | C3_cache_amp_gradaccum | 50 | 0.5505 | 0.7980 | 0.5049 | 0.2398 | 23604 | completed |
152
+ | C4_cache_amp_gradaccum_adaptivelr | 50 | 0.6667 | 0.8607 | 0.5437 | 0.2583 | 25771 | completed |
153
  | C5_cache_amp_gradaccum_adaptivelr_l1 | 50 | 0.6742 | 0.8651 | 0.5231 | 0.2485 | 24653 | completed |
154
  | C6_full_e2am | 49 | 0.6716 | 0.8621 | 0.5163 | 0.2453 | 24352 | completed |
155
 
 
163
 
164
  ## Deployment results
165
 
166
+ See `paper_tables/deployment_results_table.csv`.
167
 
168
  ## Methodology
169
 
 
233
 
234
  ---
235
 
236
+ _This README was auto-generated on 2026-07-02 08:14 UTC._
237
  _Source repo: Shanmuk4622/E2AM_ResNet50_
comparison_plots/cifar10/accuracy_bar.png CHANGED

Git LFS Details

  • SHA256: 44f4f7d412446ed5ae9e74fcf5f8975eac211b5ea26c1d497c895f8c9d978462
  • Pointer size: 130 Bytes
  • Size of remote file: 32.2 kB

Git LFS Details

  • SHA256: b9a96127f5ae51c5eb9e530f9b54ab6fd68be7b8fd7efae7ef148c3330ca3013
  • Pointer size: 130 Bytes
  • Size of remote file: 25.8 kB
comparison_plots/cifar10/accuracy_vs_energy_scatter.png CHANGED

Git LFS Details

  • SHA256: 594d02f973f69a7468b476d7b40464c57ce66778c880245b2c0adfd29b4a32f7
  • Pointer size: 130 Bytes
  • Size of remote file: 45.4 kB

Git LFS Details

  • SHA256: 439d3c0227d0936a06efa418e79e5628bb8651df6f191baf25866ed2b35b331e
  • Pointer size: 130 Bytes
  • Size of remote file: 42 kB
comparison_plots/cifar10/co2_bar.png CHANGED

Git LFS Details

  • SHA256: 399977b67a5eda81976e119d6d8f55f2c0980611b5a39766b9a2dc03c66d7e4b
  • Pointer size: 130 Bytes
  • Size of remote file: 32.2 kB

Git LFS Details

  • SHA256: 1a74cbc06cbe4f0fbc37c2e4698e5589770c5d1a0519a79765e31702eed2552e
  • Pointer size: 130 Bytes
  • Size of remote file: 30.7 kB
comparison_plots/cifar10/energy_bar.png CHANGED

Git LFS Details

  • SHA256: 23937ca00d8d024766a9231c13c2703cb4f99139d3bbd613077de345106cfa2b
  • Pointer size: 130 Bytes
  • Size of remote file: 31.6 kB

Git LFS Details

  • SHA256: c1d9d2e96c7c9149bf19a0e6baed4ced491ea5df93e4d687121f295f5459031e
  • Pointer size: 130 Bytes
  • Size of remote file: 31.5 kB
comparison_plots/cifar100/accuracy_bar.png CHANGED

Git LFS Details

  • SHA256: e4aa003f10289c826aa1b00b73135b9298b38f6747b101e32afbc5c85a7052ad
  • Pointer size: 130 Bytes
  • Size of remote file: 33.2 kB

Git LFS Details

  • SHA256: fb462e9ac17bc4fe1a5cb91b5edb0a5017a9c349a6c974a5a48cd554df94925c
  • Pointer size: 130 Bytes
  • Size of remote file: 28.4 kB
comparison_plots/cifar100/accuracy_vs_energy_scatter.png CHANGED

Git LFS Details

  • SHA256: 8f8816c49d160a480e390d7d821c540c62ab7906465fe00196cd904db52880c8
  • Pointer size: 130 Bytes
  • Size of remote file: 49.2 kB

Git LFS Details

  • SHA256: 58351da7a2e8a857c6c45cace9731d25301c608c101a81bf284d1bb3125599fe
  • Pointer size: 130 Bytes
  • Size of remote file: 44.4 kB
comparison_plots/cifar100/co2_bar.png CHANGED

Git LFS Details

  • SHA256: 16b91cf2a919181755a3fee06e4548697f2470e9720c813061355734359563ac
  • Pointer size: 130 Bytes
  • Size of remote file: 33.3 kB

Git LFS Details

  • SHA256: 77017fe3e47aa08a54224d95af9072aa9f483e2d864abb51bd65fe5b9eedceac
  • Pointer size: 130 Bytes
  • Size of remote file: 30.7 kB
comparison_plots/cifar100/energy_bar.png CHANGED

Git LFS Details

  • SHA256: be52efa6273f5e08bb3e5860f6f8c87c6c9757cbb283af3ffbc894e8cc573eb3
  • Pointer size: 130 Bytes
  • Size of remote file: 32.7 kB

Git LFS Details

  • SHA256: ebe295c5e632aa1144c8f14aadadc4ac1ce4e2a7c3f327e4af7ab65519e79dfc
  • Pointer size: 130 Bytes
  • Size of remote file: 31.1 kB
comparison_plots/cross_dataset/accuracy_by_variant_across_datasets.png CHANGED

Git LFS Details

  • SHA256: 9b0660b341182d099ec140e4f11658a93541f840a6ee7148760357a65aa25c6d
  • Pointer size: 130 Bytes
  • Size of remote file: 38.5 kB

Git LFS Details

  • SHA256: b3e6e993c32c75fd57d8d45836c917e53b0e7ae88e413c8a3b319a8e1b58935d
  • Pointer size: 130 Bytes
  • Size of remote file: 30.2 kB
comparison_plots/cross_dataset/energy_by_variant_across_datasets.png CHANGED

Git LFS Details

  • SHA256: df79fd6160cf79b8c70bf090951a30f3f3ffe7285f10e20344020a985b249860
  • Pointer size: 130 Bytes
  • Size of remote file: 33.6 kB

Git LFS Details

  • SHA256: 6b159cb047b324fa60cff687a432f41bcbf47caa7afb9a8285bd2926b9c1d5db
  • Pointer size: 130 Bytes
  • Size of remote file: 32.4 kB
comparison_plots/tiny_imagenet/accuracy_bar.png CHANGED

Git LFS Details

  • SHA256: 31d3b748d8e932ea4cbf04d423fc6595408b8501e20b8de6c3b2514f08c68733
  • Pointer size: 130 Bytes
  • Size of remote file: 34.6 kB

Git LFS Details

  • SHA256: fd15ad6ec970cb99a237071f8acef60a63f715280c3c61780f05ca72cfbc26e3
  • Pointer size: 130 Bytes
  • Size of remote file: 29.1 kB
comparison_plots/tiny_imagenet/accuracy_vs_energy_scatter.png CHANGED

Git LFS Details

  • SHA256: dca5d7f03adff0e688c51a3006322b12331194c1eacd95e894dd1d604db40acf
  • Pointer size: 130 Bytes
  • Size of remote file: 46.2 kB

Git LFS Details

  • SHA256: 73c215e3fca96d2611d0e77ff6e40996e74ef0c70dc79e8d595862ee97a17d4f
  • Pointer size: 130 Bytes
  • Size of remote file: 41.7 kB
comparison_plots/tiny_imagenet/co2_bar.png CHANGED

Git LFS Details

  • SHA256: c8e1dbc07eac2991907dc6e83759cb48e53b03102f1397dcc43c2ae37cec0735
  • Pointer size: 130 Bytes
  • Size of remote file: 29.4 kB

Git LFS Details

  • SHA256: 5be227a8b70a6e004127eb309b2715d875390d45e437e469a6c1b7c9fed49508
  • Pointer size: 130 Bytes
  • Size of remote file: 27.6 kB
comparison_plots/tiny_imagenet/energy_bar.png CHANGED

Git LFS Details

  • SHA256: 2cbe35873c39655ea80e8f5ba2dffa85996521b86cf476b70a69ed535cb06107
  • Pointer size: 130 Bytes
  • Size of remote file: 29.9 kB

Git LFS Details

  • SHA256: 37cd8e325e2ee36675cfeb22f50d345b9eff4737b04c5ae06edbdbeecc294062
  • Pointer size: 130 Bytes
  • Size of remote file: 28.9 kB