max_model_size_kb: 32 could never be satisfied by the reference model it gates: AnomalyDetectionAE (640/784 -> 128x4 -> 8 -> 128x4 -> 640, reference/tiny/anomaly_detection_ae.py) has ~266K parameters -- matching this same file's own params: 0.3M -- which is ~1.03MB at FP32 and ~260KB even fully INT8-quantized.
A submitter who trains exactly this reference model and reports the true model size would automatically fail the size gate regardless of reconstruction quality, since even the best-case (INT8) size is ~8x over budget.
Test plan
Manually computed parameter count from the architecture definition (5 linear layers each side: 640->128->128->128->128->8 encoder, 8->128->128->128->128->640 decoder): 265,864 params, matching the YAML's own declared params: 0.3M.
FP32 size: 265,864 * 4 bytes ~= 1.01 MiB. INT8 size: 265,864 bytes ~= 260 KB. Both far exceed the old 32KB budget.
Corrected to 300KB, consistent with the file's own stated 0.3M parameter count at 1 byte/param (INT8).
pre-commit's YAML syntax check passes on the edited file.
🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.
## 📋 Pull Request Information
**Original PR:** https://github.com/harvard-edge/cs249r_book/pull/1937
**Author:** [@Shashank-Tripathi-07](https://github.com/Shashank-Tripathi-07)
**Created:** 7/7/2026
**Status:** 🔄 Open
**Base:** `dev` ← **Head:** `fix/anomaly-ae-model-size-budget`
---
### 📝 Commits (1)
- [`a8ba239`](https://github.com/harvard-edge/cs249r_book/commit/a8ba239ed34bef61b51024ee7c43176cca504c62) fix(mlperf-edu): anomaly-ae-train model-size gate impossible to pass
### 📊 Changes
**1 file changed** (+1 additions, -1 deletions)
<details>
<summary>View changed files</summary>
📝 `mlperf-edu/registry/suites/tiny/anomaly-ae-train.yaml` (+1 -1)
</details>
### 📄 Description
## Summary
- `max_model_size_kb: 32` could never be satisfied by the reference model it gates: `AnomalyDetectionAE` (640/784 -> 128x4 -> 8 -> 128x4 -> 640, `reference/tiny/anomaly_detection_ae.py`) has ~266K parameters -- matching this same file's own `params: 0.3M` -- which is ~1.03MB at FP32 and ~260KB even fully INT8-quantized.
- A submitter who trains exactly this reference model and reports the true model size would automatically fail the size gate regardless of reconstruction quality, since even the best-case (INT8) size is ~8x over budget.
## Test plan
- [x] Manually computed parameter count from the architecture definition (5 linear layers each side: 640->128->128->128->128->8 encoder, 8->128->128->128->128->640 decoder): 265,864 params, matching the YAML's own declared `params: 0.3M`.
- [x] FP32 size: 265,864 * 4 bytes ~= 1.01 MiB. INT8 size: 265,864 bytes ~= 260 KB. Both far exceed the old 32KB budget.
- [x] Corrected to 300KB, consistent with the file's own stated 0.3M parameter count at 1 byte/param (INT8).
- [x] `pre-commit`'s YAML syntax check passes on the edited file.
---
<sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
📋 Pull Request Information
Original PR: https://github.com/harvard-edge/cs249r_book/pull/1937
Author: @Shashank-Tripathi-07
Created: 7/7/2026
Status: 🔄 Open
Base:
dev← Head:fix/anomaly-ae-model-size-budget📝 Commits (1)
a8ba239fix(mlperf-edu): anomaly-ae-train model-size gate impossible to pass📊 Changes
1 file changed (+1 additions, -1 deletions)
View changed files
📝
mlperf-edu/registry/suites/tiny/anomaly-ae-train.yaml(+1 -1)📄 Description
Summary
max_model_size_kb: 32could never be satisfied by the reference model it gates:AnomalyDetectionAE(640/784 -> 128x4 -> 8 -> 128x4 -> 640,reference/tiny/anomaly_detection_ae.py) has ~266K parameters -- matching this same file's ownparams: 0.3M-- which is ~1.03MB at FP32 and ~260KB even fully INT8-quantized.Test plan
params: 0.3M.pre-commit's YAML syntax check passes on the edited file.🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.