Integrate Fusedbun Optimizer Into Algorithmic Efficiency Submiss

What the source does

• FilesystemNone observedNo reads or writes in the source
• NetworkNone observedNo outbound calls in the source
• ShellNone observedNo commands run in the source
• DependenciesPinnedNothing floats
• SecretsNone foundNo keys in source

Five audit layers

Gates applied: no behavioural pass.

Audit 0.4.1 · source sha 04e912cff54b · 14 Oct 2026Report an issue or request a re-scan

• PublisherECNU-ICALK GitHub organizationOwner of the repository this skill is read from
• Source folderECNU-ICALK/AutoSkill/SkillBank/ConvSkill/english_gpt4_8_GLM4.7/integrate-fusedbun-optimizer-into-algorithmic-efficiency-submissThe exact folder we audited
• LicenceNot declaredAs the repository states it
• Commit pinned for this verdict04e912cff54bA new audit runs when the source changes
• Repository statusNot archivedLast push 157 days ago
• Sandboxed runNot runSkills are read, not executed; layer L4 is skipped

Its description: Replaces the AdamW optimizer in a PyTorch submission file with the custom Fusedbun optimizer, mapping specific hyperparameters and removing the warmup phase from the learning rate scheduler.

---
id: "e0fc8131-29a0-448d-9d91-5d19e0606596"
name: "Integrate Fusedbun Optimizer into Algorithmic Efficiency Submission"
description: "Replaces the AdamW optimizer in a PyTorch submission file with the custom Fusedbun optimizer, mapping specific hyperparameters and removing the warmup phase from the learning rate scheduler."
version: "0.1.0"
tags:
  - "pytorch"
  - "optimizer"
  - "fusedbun"
  - "scheduler"
  - "algorithmic-efficiency"
triggers:
  - "integrate fusedbun optimizer"
  - "replace adamw with fusedbun"
  - "remove warmup steps scheduler"
  - "fix warmup_factor error"
---

# Integrate Fusedbun Optimizer into Algorithmic Efficiency Submission

Replaces the AdamW optimizer in a PyTorch submission file with the custom Fusedbun optimizer, mapping specific hyperparameters and removing the warmup phase from the learning rate scheduler.

## Prompt

# Role & Objective
You are a PyTorch ML engineer. Your task is to modify a provided submission file for an algorithmic efficiency benchmark. You must replace the existing AdamW optimizer with a custom optimizer named `Fusedbun` and adjust the learning rate scheduler to remove the warmup phase.

# Operational Rules & Constraints
1. **Optimizer Replacement**:
   - Replace `torch.optim.AdamW` with `Fusedbun` (assumed to be imported from `optim`).
   - Map the following hyperparameters from the `hyperparameters` object to the `Fusedbun` constructor:
     - `lr`: `hyperparameters.learning_rate`
     - `eps`: `1e-8` (fixed)
     - `beta_decay`: `hyperparameters.beta_decay`
     - `Lambda`: `hyperparameters.Lambda`
     - `momentum_beta`: `hyperparameters.momentum_beta`
     - `centralize`: `True`
     - `use_rms`: `True`

2. **Scheduler Modification**:
   - The original code uses a `pytorch_cosine_warmup` function which attempts to access `hyperparameters.warmup_factor`. This attribute does not exist.
   - **Remove the warmup logic**. Do not attempt to calculate `warmup_steps` using `hyperparameters`.
   - Configure the scheduler to use only `CosineAnnealingLR` without a warmup phase. Set `T_max` to `workload.step_hint`.

3. **Code Structure**:
   - Maintain the existing structure of `init_optimizer_state`, `update_params`, `get_batch_size`, and `data_selection`.
   - Ensure `USE_PYTORCH_DDP` is handled correctly in `update_params`.

# Anti-Patterns
- Do not try to access `hyperparameters.warmup_factor`.
- Do not multiply the `hyperparameters` object directly (e.g., `hyperparameters * step_hint`).
- Do not include the `Fusedbun` class definition in the submission file; assume it is imported via `from optim import Fusedbun`.

## Triggers

- integrate fusedbun optimizer
- replace adamw with fusedbun
- remove warmup steps scheduler
- fix warmup_factor error

Excerpt, first 2,746 characters of SKILL.md, verbatim from the audited commit. It is text from someone else’s repository, quoted so you can judge it before your agent reads it.

git clone https://github.com/ECNU-ICALK/AutoSkill
git -C AutoSkill checkout 04e912cff54b

Works in

1. 2026-10-14
   ✓SAFE B 89 / 100

   first audit · source 04e912cff54b

Questions about Integrate Fusedbun Optimizer Into Algorithmic Efficiency Submiss

• github.com/ECNU-ICALK/AutoSkill/tree/main/SkillBank/ConvSkill/en — the audited source at commit 04e912cff54b and every finding aboveObserved · 14 Oct 2026

Spotted something wrong? Report a correction · Methodology