Throttle dynamic VRAM prepare logging

[kijai]

The "prepared for dynamic VRAM loading" line currently fires at INFO on every load() call, which spams the console when the same model is reloaded inside a tight loop (e.g. SDPoseKeypointExtractor running one forward pass per bbox per image). When a tqdm progress bar is active and the message would say the exact same thing as the previous load, drop it to DEBUG; the line is still recoverable under --verbose.

Example when running SDPose detection:

Before:

Requested to load Lotus
Model Lotus prepared for dynamic VRAM loading. 1668MB Staged. 0 patches attached.
Extracting keypoints:  10%|██████                                                       | 1/10 [00:01<00:16,  1.81s/it]Model AutoencoderKL prepared for dynamic VRAM loading. 159MB Staged. 0 patches attached.
Model Lotus prepared for dynamic VRAM loading. 1668MB Staged. 0 patches attached.
Extracting keypoints:  20%|████████████▏                                                | 2/10 [00:02<00:07,  1.08it/s]Model AutoencoderKL prepared for dynamic VRAM loading. 159MB Staged. 0 patches attached.
Model Lotus prepared for dynamic VRAM loading. 1668MB Staged. 0 patches attached.
Extracting keypoints:  30%|██████████████████▎                                          | 3/10 [00:02<00:04,  1.57it/s]Model AutoencoderKL prepared for dynamic VRAM loading. 159MB Staged. 0 patches attached.
Model Lotus prepared for dynamic VRAM loading. 1668MB Staged. 0 patches attached.
Extracting keypoints:  40%|████████████████████████▍                                    | 4/10 [00:02<00:03,  1.99it/s]Model AutoencoderKL prepared for dynamic VRAM loading. 159MB Staged. 0 patches attached.
Model Lotus prepared for dynamic VRAM loading. 1668MB Staged. 0 patches attached.
Extracting keypoints:  50%|██████████████████████████████▌                              | 5/10 [00:03<00:02,  2.32it/s]Model AutoencoderKL prepared for dynamic VRAM loading. 159MB Staged. 0 patches attached.
Model Lotus prepared for dynamic VRAM loading. 1668MB Staged. 0 patches attached.
Extracting keypoints:  60%|████████████████████████████████████▌                        | 6/10 [00:03<00:01,  2.62it/s]Model AutoencoderKL prepared for dynamic VRAM loading. 159MB Staged. 0 patches attached.
Model Lotus prepared for dynamic VRAM loading. 1668MB Staged. 0 patches attached.
Extracting keypoints:  70%|██████████████████████████████████████████▋                  | 7/10 [00:03<00:01,  2.84it/s]Model AutoencoderKL prepared for dynamic VRAM loading. 159MB Staged. 0 patches attached.
Model Lotus prepared for dynamic VRAM loading. 1668MB Staged. 0 patches attached.
Extracting keypoints:  80%|████████████████████████████████████████████████▊            | 8/10 [00:03<00:00,  2.97it/s]Model AutoencoderKL prepared for dynamic VRAM loading. 159MB Staged. 0 patches attached.
Model Lotus prepared for dynamic VRAM loading. 1668MB Staged. 0 patches attached.
Extracting keypoints:  90%|██████████████████████████████████████████████████████▉      | 9/10 [00:04<00:00,  3.07it/s]Model AutoencoderKL prepared for dynamic VRAM loading. 159MB Staged. 0 patches attached.
Model Lotus prepared for dynamic VRAM loading. 1668MB Staged. 0 patches attached.
Extracting keypoints: 100%|████████████████████████████████████████████████████████████| 10/10 [00:04<00:00,  2.23it/s]
Prompt executed in 5.10 seconds

After:

Requested to load Lotus
Model Lotus prepared for dynamic VRAM loading. 1668MB Staged. 0 patches attached.
Extracting keypoints: 100%|████████████████████████████████████████████████████████████| 10/10 [00:03<00:00,  3.33it/s]
Prompt executed in 3.04 seconds

[coderabbitai]

📝 Walkthrough

Walkthrough

The change adds a tqdm import and updates ModelPatcherDynamic.load() to de-duplicate its "prepared for dynamic VRAM loading" message. It computes a log_key from (patches_uuid, allocated_size, num_patches, len(self.backup), self.model.model_loaded_weight_memory), compares it to self._last_prepare_log_key, and selects DEBUG when repeated or INFO otherwise. The message is emitted via logging.log(level, ...) instead of always calling logging.info(...).

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title 'Throttle dynamic VRAM prepare logging' accurately describes the main change: reducing INFO log spam for repeated dynamic VRAM loading messages by implementing throttling/de-duplication logic.
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.
Description check	✅ Passed	The PR description accurately explains the changes: throttling log levels for repeated dynamic VRAM loading messages when inside a tqdm loop, with concrete before/after examples demonstrating the spam reduction.

_{✏️ 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.}