comfy-aimdo 0.2.14: Hotfix async allocator estimations

[rattus128]

This was doing an over-estimate of VRAM used by the async allocator when lots of little small tensors were in play.

Also change the versioning scheme to == so we can roll forward aimdo without worrying about stable regressions downstream in comfyUI core.

This is the deeper root cause fix for a reported regression in LTX2.3 WRT: #13487

Example Test Conditions:

Linux, RTX5090, LTX2.3 T2V (unchanged default workflow)

Before:

Requested to load LTXAV
Model LTXAV prepared for dynamic VRAM loading. 23838MB Staged. 1660 patches attached. Force pre-loaded 1496 weights: 44 KB.
100%|██████████| 8/8 [00:07<00:00,  1.11it/s]                                   
Model LTXAV prepared for dynamic VRAM loading. 23838MB Staged. 1660 patches attached. Force pre-loaded 1496 weights: 44 KB.
100%|██████████| 3/3 [00:13<00:00,  4.58s/it]                                   
Requested to load AudioVAE
loaded completely;  693.46 MB loaded, full load: True
Requested to load VideoVAE
0 models unloaded.
Model VideoVAE prepared for dynamic VRAM loading. 1384MB Staged. 0 patches attached.
Prompt executed in 34.67 seconds
got prompt
Model LTXAV prepared for dynamic VRAM loading. 23838MB Staged. 1660 patches attached. Force pre-loaded 1496 weights: 44 KB.
100%|██████████| 8/8 [00:07<00:00,  1.03it/s]                                   
Model LTXAV prepared for dynamic VRAM loading. 23838MB Staged. 1660 patches attached. Force pre-loaded 1496 weights: 44 KB.
100%|██████████| 3/3 [00:15<00:00,  5.02s/it]                                   
0 models unloaded.
Model VideoVAE prepared for dynamic VRAM loading. 1384MB Staged. 0 patches attached.
Prompt executed in 28.97 seconds

After:

...
Requested to load LTXAV
Model LTXAV prepared for dynamic VRAM loading. 23838MB Staged. 1660 patches attached. Force pre-loaded 1496 weights: 44 KB.
100%|██████████| 8/8 [00:07<00:00,  1.11it/s]                                   
Model LTXAV prepared for dynamic VRAM loading. 23838MB Staged. 1660 patches attached. Force pre-loaded 1496 weights: 44 KB.
100%|██████████| 3/3 [00:13<00:00,  4.58s/it]                                   
Requested to load AudioVAE
loaded completely;  693.46 MB loaded, full load: True
Requested to load VideoVAE
0 models unloaded.
Model VideoVAE prepared for dynamic VRAM loading. 1384MB Staged. 0 patches attached.
got prompt
Prompt executed in 34.53 seconds
Model LTXAV prepared for dynamic VRAM loading. 23838MB Staged. 1660 patches attached. Force pre-loaded 1496 weights: 44 KB.
100%|██████████| 8/8 [00:07<00:00,  1.10it/s]                                   
Model LTXAV prepared for dynamic VRAM loading. 23838MB Staged. 1660 patches attached. Force pre-loaded 1496 weights: 44 KB.
100%|██████████| 3/3 [00:13<00:00,  4.60s/it]                                   
0 models unloaded.
Model VideoVAE prepared for dynamic VRAM loading. 1384MB Staged. 0 patches attached.
Prompt executed in 26.66 seconds

VS revert of #13487 for comparison:

Requested to load LTXAV
Model LTXAV prepared for dynamic VRAM loading. 23838MB Staged. 1660 patches attached.
100%|██████████| 8/8 [00:07<00:00,  1.11it/s]                                   
Model LTXAV prepared for dynamic VRAM loading. 23838MB Staged. 1660 patches attached.
100%|██████████| 3/3 [00:13<00:00,  4.59s/it]                                   
Requested to load AudioVAE
loaded completely;  693.46 MB loaded, full load: True
Requested to load VideoVAE
0 models unloaded.
Model VideoVAE prepared for dynamic VRAM loading. 1384MB Staged. 0 patches attached.
Prompt executed in 34.35 seconds
got prompt
Model LTXAV prepared for dynamic VRAM loading. 23838MB Staged. 1660 patches attached.
100%|██████████| 8/8 [00:07<00:00,  1.10it/s]                                   
Model LTXAV prepared for dynamic VRAM loading. 23838MB Staged. 1660 patches attached.
100%|██████████| 3/3 [00:13<00:00,  4.61s/it]                                   
0 models unloaded.
Model VideoVAE prepared for dynamic VRAM loading. 1384MB Staged. 0 patches attached.
Prompt executed in 26.82 seconds

Regression Tests:

Windows RTX5060, 64GB RAM Flux2 dev FP8 ✅
Windows RTX5060, 64GB RAM LTX2.0 ✅
Windows RTX5060, 64GB wan 2.2 fp16 ✅
Linux RTX5090, Ace-step 1.5 XL turbo ✅
Linux RTX5090, 96GB flux fill inpaint ✅

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: aae03e51-f0cb-48da-a3af-88e495752bdf

📥 Commits

Reviewing files that changed from the base of the PR and between 3cdc0d5 and f1d0eb5.

📒 Files selected for processing (1)

• requirements.txt

---

📝 Walkthrough

Walkthrough

The requirements.txt file was modified to change the version constraint for the comfy-aimdo package from >=0.2.12 to ==0.2.14. This change pins the dependency to a specific version rather than allowing any version at or above the minimum specified version.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title accurately describes the main change: updating comfy-aimdo to version 0.2.14 to fix async allocator VRAM estimation issues.
Description check	✅ Passed	The description clearly explains the fix for VRAM over-estimation, the versioning scheme change, and includes test results demonstrating the improvement.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}