Skip to content

updated imagenet_akidanet README.md to have collapsible tables/plots #21

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
mtorresBrainChip wants to merge 1 commit into
base: main
Choose a base branch
from
Open

updated imagenet_akidanet README.md to have collapsible tables/plots #21

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
56 changes: 44 additions & 12 deletions akida1/model_zoo/imagenet_akidanet/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -47,7 +47,7 @@ activations that are zero, and therefore the work the hardware skips.
<tbody>
<tr><td colspan="9"><b>224 &times; 224 input</b></td></tr>
<tr>
<td align="center">1.0</td>
<td align="center"><a href="#bench-a100-224">1.0</a></td>
<td align="center">72.06%</td>
<td align="center">90.40%</td>
<td align="center">70.09%</td>
Expand All @@ -58,7 +58,7 @@ activations that are zero, and therefore the work the hardware skips.
<td align="center">4,445,000</td>
</tr>
<tr>
<td align="center">0.5</td>
<td align="center"><a href="#bench-a50-224">0.5</a></td>
<td align="center">65.11%</td>
<td align="center">85.95%</td>
<td align="center">61.94%</td>
Expand All @@ -69,7 +69,7 @@ activations that are zero, and therefore the work the hardware skips.
<td align="center">1,384,856</td>
</tr>
<tr>
<td align="center">0.25</td>
<td align="center"><a href="#bench-a25-224">0.25</a></td>
<td align="center">53.27%</td>
<td align="center">77.21%</td>
<td align="center">46.41%</td>
Expand All @@ -81,7 +81,7 @@ activations that are zero, and therefore the work the hardware skips.
</tr>
<tr><td colspan="9"><b>160 &times; 160 input</b></td></tr>
<tr>
<td align="center">1.0</td>
<td align="center"><a href="#bench-a100-160">1.0</a></td>
<td align="center">69.68%</td>
<td align="center">88.87%</td>
<td align="center">67.04%</td>
Expand All @@ -92,7 +92,7 @@ activations that are zero, and therefore the work the hardware skips.
<td align="center">4,445,000</td>
</tr>
<tr>
<td align="center">0.5</td>
<td align="center"><a href="#bench-a50-160">0.5</a></td>
<td align="center">62.36%</td>
<td align="center">83.97%</td>
<td align="center">57.81%</td>
Expand All @@ -103,7 +103,7 @@ activations that are zero, and therefore the work the hardware skips.
<td align="center">1,384,856</td>
</tr>
<tr>
<td align="center">0.25</td>
<td align="center"><a href="#bench-a25-160">0.25</a></td>
<td align="center">50.00%</td>
<td align="center">74.43%</td>
<td align="center">42.64%</td>
Expand Down Expand Up @@ -185,7 +185,12 @@ mapping — the fastest and lowest-energy of the three modes on every model here

<br><br>

**alpha = 1.0, 224 &times; 224**
*Each model below is collapsed &mdash; click one to open its mapping table and power plot, or click an **Alpha** value in the Model Card above to jump straight to it.*

<details id="bench-a100-224">
<summary><b>alpha = 1.0 &middot; 224 &times; 224</b> </summary>

<br>

<table>
<thead>
Expand Down Expand Up @@ -234,9 +239,14 @@ mapping — the fastest and lowest-energy of the three modes on every model here

<img src="docs/ref_benchmark_results_full_a100_224.png" alt="Power measurements during inference, alpha=1.0 224x224" width="1050">

</details>

<br><br>

**alpha = 0.5, 224 &times; 224**
<details id="bench-a50-224">
<summary><b>alpha = 0.5 &middot; 224 &times; 224</b> </summary>

<br>

<table>
<thead>
Expand Down Expand Up @@ -285,9 +295,14 @@ mapping — the fastest and lowest-energy of the three modes on every model here

<img src="docs/ref_benchmark_results_full_a50_224.png" alt="Power measurements during inference, alpha=0.5 224x224" width="1050">

</details>

<br><br>

**alpha = 0.25, 224 &times; 224**
<details id="bench-a25-224">
<summary><b>alpha = 0.25 &middot; 224 &times; 224</b> </summary>

<br>

<table>
<thead>
Expand Down Expand Up @@ -336,9 +351,14 @@ mapping — the fastest and lowest-energy of the three modes on every model here

<img src="docs/ref_benchmark_results_full_a25_224.png" alt="Power measurements during inference, alpha=0.25 224x224" width="1050">

</details>

<br><br>

**alpha = 1.0, 160 &times; 160**
<details id="bench-a100-160">
<summary><b>alpha = 1.0 &middot; 160 &times; 160</b> </summary>

<br>

<table>
<thead>
Expand Down Expand Up @@ -387,9 +407,14 @@ mapping — the fastest and lowest-energy of the three modes on every model here

<img src="docs/ref_benchmark_results_full_a100_160.png" alt="Power measurements during inference, alpha=1.0 160x160" width="1050">

</details>

<br><br>

**alpha = 0.5, 160 &times; 160**
<details id="bench-a50-160">
<summary><b>alpha = 0.5 &middot; 160 &times; 160</b> </summary>

<br>

<table>
<thead>
Expand Down Expand Up @@ -438,9 +463,14 @@ mapping — the fastest and lowest-energy of the three modes on every model here

<img src="docs/ref_benchmark_results_full_a50_160.png" alt="Power measurements during inference, alpha=0.5 160x160" width="1050">

</details>

<br><br>

**alpha = 0.25, 160 &times; 160**
<details id="bench-a25-160">
<summary><b>alpha = 0.25 &middot; 160 &times; 160</b> </summary>

<br>

<table>
<thead>
Expand Down Expand Up @@ -489,6 +519,8 @@ mapping — the fastest and lowest-energy of the three modes on every model here

<img src="docs/ref_benchmark_results_full_a25_160.png" alt="Power measurements during inference, alpha=0.25 160x160" width="1050">

</details>

<br><br>

### Reading the benchmark tables
Expand Down
56 changes: 44 additions & 12 deletions akida1/model_zoo/imagenet_akidanet/docs/README.md.template
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -47,7 +47,7 @@ activations that are zero, and therefore the work the hardware skips.
<tbody>
<tr><td colspan="9"><b>224 &times; 224 input</b></td></tr>
<tr>
<td align="center">1.0</td>
<td align="center"><a href="#bench-a100-224">1.0</a></td>
<td align="center">{a100_224_float_t1}</td>
<td align="center">{a100_224_float_t5}</td>
<td align="center">{a100_224_qat_t1}</td>
Expand All @@ -58,7 +58,7 @@ activations that are zero, and therefore the work the hardware skips.
<td align="center">{a100_224_params}</td>
</tr>
<tr>
<td align="center">0.5</td>
<td align="center"><a href="#bench-a50-224">0.5</a></td>
<td align="center">{a50_224_float_t1}</td>
<td align="center">{a50_224_float_t5}</td>
<td align="center">{a50_224_qat_t1}</td>
Expand All @@ -69,7 +69,7 @@ activations that are zero, and therefore the work the hardware skips.
<td align="center">{a50_224_params}</td>
</tr>
<tr>
<td align="center">0.25</td>
<td align="center"><a href="#bench-a25-224">0.25</a></td>
<td align="center">{a25_224_float_t1}</td>
<td align="center">{a25_224_float_t5}</td>
<td align="center">{a25_224_qat_t1}</td>
Expand All @@ -81,7 +81,7 @@ activations that are zero, and therefore the work the hardware skips.
</tr>
<tr><td colspan="9"><b>160 &times; 160 input</b></td></tr>
<tr>
<td align="center">1.0</td>
<td align="center"><a href="#bench-a100-160">1.0</a></td>
<td align="center">{a100_160_float_t1}</td>
<td align="center">{a100_160_float_t5}</td>
<td align="center">{a100_160_qat_t1}</td>
Expand All @@ -92,7 +92,7 @@ activations that are zero, and therefore the work the hardware skips.
<td align="center">{a100_160_params}</td>
</tr>
<tr>
<td align="center">0.5</td>
<td align="center"><a href="#bench-a50-160">0.5</a></td>
<td align="center">{a50_160_float_t1}</td>
<td align="center">{a50_160_float_t5}</td>
<td align="center">{a50_160_qat_t1}</td>
Expand All @@ -103,7 +103,7 @@ activations that are zero, and therefore the work the hardware skips.
<td align="center">{a50_160_params}</td>
</tr>
<tr>
<td align="center">0.25</td>
<td align="center"><a href="#bench-a25-160">0.25</a></td>
<td align="center">{a25_160_float_t1}</td>
<td align="center">{a25_160_float_t5}</td>
<td align="center">{a25_160_qat_t1}</td>
Expand Down Expand Up @@ -185,7 +185,12 @@ mapping — the fastest and lowest-energy of the three modes on every model here

<br><br>

**alpha = 1.0, 224 &times; 224**
*Each model below is collapsed &mdash; click one to open its mapping table and power plot, or click an **Alpha** value in the Model Card above to jump straight to it.*

<details id="bench-a100-224">
<summary><b>alpha = 1.0 &middot; 224 &times; 224</b> </summary>

<br>

<table>
<thead>
Expand Down Expand Up @@ -234,9 +239,14 @@ mapping — the fastest and lowest-energy of the three modes on every model here

<img src="docs/ref_benchmark_results_full_a100_224.png" alt="Power measurements during inference, alpha=1.0 224x224" width="1050">

</details>

<br><br>

**alpha = 0.5, 224 &times; 224**
<details id="bench-a50-224">
<summary><b>alpha = 0.5 &middot; 224 &times; 224</b> </summary>

<br>

<table>
<thead>
Expand Down Expand Up @@ -285,9 +295,14 @@ mapping — the fastest and lowest-energy of the three modes on every model here

<img src="docs/ref_benchmark_results_full_a50_224.png" alt="Power measurements during inference, alpha=0.5 224x224" width="1050">

</details>

<br><br>

**alpha = 0.25, 224 &times; 224**
<details id="bench-a25-224">
<summary><b>alpha = 0.25 &middot; 224 &times; 224</b> </summary>

<br>

<table>
<thead>
Expand Down Expand Up @@ -336,9 +351,14 @@ mapping — the fastest and lowest-energy of the three modes on every model here

<img src="docs/ref_benchmark_results_full_a25_224.png" alt="Power measurements during inference, alpha=0.25 224x224" width="1050">

</details>

<br><br>

**alpha = 1.0, 160 &times; 160**
<details id="bench-a100-160">
<summary><b>alpha = 1.0 &middot; 160 &times; 160</b> </summary>

<br>

<table>
<thead>
Expand Down Expand Up @@ -387,9 +407,14 @@ mapping — the fastest and lowest-energy of the three modes on every model here

<img src="docs/ref_benchmark_results_full_a100_160.png" alt="Power measurements during inference, alpha=1.0 160x160" width="1050">

</details>

<br><br>

**alpha = 0.5, 160 &times; 160**
<details id="bench-a50-160">
<summary><b>alpha = 0.5 &middot; 160 &times; 160</b> </summary>

<br>

<table>
<thead>
Expand Down Expand Up @@ -438,9 +463,14 @@ mapping — the fastest and lowest-energy of the three modes on every model here

<img src="docs/ref_benchmark_results_full_a50_160.png" alt="Power measurements during inference, alpha=0.5 160x160" width="1050">

</details>

<br><br>

**alpha = 0.25, 160 &times; 160**
<details id="bench-a25-160">
<summary><b>alpha = 0.25 &middot; 160 &times; 160</b> </summary>

<br>

<table>
<thead>
Expand Down Expand Up @@ -489,6 +519,8 @@ mapping — the fastest and lowest-energy of the three modes on every model here

<img src="docs/ref_benchmark_results_full_a25_160.png" alt="Power measurements during inference, alpha=0.25 160x160" width="1050">

</details>

<br><br>

### Reading the benchmark tables
Expand Down