xinyun/TeaCache
1
0
Fork 0
mirror of https://git.datalinker.icu/ali-vilab/TeaCache synced 2025-12-08 20:34:24 +08:00
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #76 from spawner1145/main

Browse Source 
Update Coefficients for lumina2
This commit is contained in:
Feng Liu 2025-06-08 22:23:17 +08:00 committed by GitHub
commit c730b01e42
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 27 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

28
TeaCache4Lumina2/README.md
View File

@ -1,8 +1,17 @@
<!-- ## **TeaCache4LuminaT2X** --> <!-- ## **TeaCache4LuminaT2X** -->
# TeaCache4Lumina2 # TeaCache4Lumina2
[TeaCache](https://github.com/LiewFeng/TeaCache) can speedup [Lumina-Image-2.0](https://github.com/Alpha-VLLM/Lumina-Image-2.0) without much visual quality degradation, in a training-free manner. The following image shows the results generated by TeaCache-Lumina-Image-2.0 with various rel_l1_thresh values: 0 (original), 0.2 (1.25x speedup), 0.3 (1.5625x speedup), 0.4 (2.0833x speedup), 0.5 (2.5x speedup). [TeaCache](https://github.com/LiewFeng/TeaCache) can speedup [Lumina-Image-2.0](https://github.com/Alpha-VLLM/Lumina-Image-2.0) without much visual quality degradation, in a training-free manner. The following image shows the experimental results of Lumina-Image-2.0 and TeaCache with different versions: v1(0 (original), 0.2 (1.25x speedup), 0.3 (1.5625x speedup), 0.4 (2.0833x speedup), 0.5 (2.5x speedup).) and v2(Lumina-Image-2.0 (~25 s), TeaCache (0.2) (~16.7 s, 1.5x speedup), TeaCache (0.3) (~15.6 s, 1.6x speedup), TeaCache (0.5) (~13.79 s, 1.8x speedup), TeaCache (1.1) (~11.9 s, 2.1x speedup)).
The v1 coefficients
`[393.76566581,−603.50993606,209.10239044,−23.00726601,0.86377344]`
exhibit poor quality at low L1 values but perform better with higher L1 settings, though at a slower speed. The v2 coefficients
`[225.7042019806413,−608.8453716535591,304.1869942338369,124.21267720116742,−1.4089066892956552]`
, however, offer faster computation and better quality at low L1 levels but incur significant feature loss at high L1 values.
You can change the value in line 72 to switch versions
## v1
<p align="center"> <p align="center">
<img src="https://github.com/user-attachments/assets/d2c87b99-e4ac-4407-809a-caf9750f41ef" width="150" style="margin: 5px;"> <img src="https://github.com/user-attachments/assets/d2c87b99-e4ac-4407-809a-caf9750f41ef" width="150" style="margin: 5px;">
<img src="https://github.com/user-attachments/assets/411ff763-9c31-438d-8a9b-3ec5c88f6c27" width="150" style="margin: 5px;"> <img src="https://github.com/user-attachments/assets/411ff763-9c31-438d-8a9b-3ec5c88f6c27" width="150" style="margin: 5px;">
@ -11,13 +20,26 @@
<img src="https://github.com/user-attachments/assets/d6a52c4c-bd22-45c0-9f40-00a2daa85fc8" width="150" style="margin: 5px;"> <img src="https://github.com/user-attachments/assets/d6a52c4c-bd22-45c0-9f40-00a2daa85fc8" width="150" style="margin: 5px;">
</p> </p>
## v2
<p align="center">
<img src="https://github.com/user-attachments/assets/aea9907b-830e-497b-b968-aaeef463c7ef" width="150" style="margin: 5px;">
<img src="https://github.com/user-attachments/assets/0e258295-eaaa-49ce-b16f-bba7f7ada6c1" width="150" style="margin: 5px;">
<img src="https://github.com/user-attachments/assets/44600f22-3fd4-4bc4-ab00-29b0ed023d6d" width="150" style="margin: 5px;">
<img src="https://github.com/user-attachments/assets/bcb926ab-95fd-4c83-8b46-f72581a3359e" width="150" style="margin: 5px;">
<img src="https://github.com/user-attachments/assets/ec8db28e-0f9b-4d56-9096-fdc8b3c20f4b" width="150" style="margin: 5px;">
</p>
## 📈 Inference Latency Comparisons on a single 4090 (step 50) ## 📈 Inference Latency Comparisons on a single 4090 (step 50)
## v1
| Lumina-Image-2.0 | TeaCache (0.2) | TeaCache (0.3) | TeaCache (0.4) | TeaCache (0.5) | | Lumina-Image-2.0 | TeaCache (0.2) | TeaCache (0.3) | TeaCache (0.4) | TeaCache (0.5) |
|:-------------------------:|:---------------------------:|:--------------------:|:---------------------:|:---------------------:| |:-------------------------:|:---------------------------:|:--------------------:|:---------------------:|:---------------------:|
| ~25 s | ~20 s | ~16 s | ~12 s | ~10 s | | ~25 s | ~20 s | ~16 s | ~12 s | ~10 s |
## v2
| Lumina-Image-2.0 | TeaCache (0.2) | TeaCache (0.3) | TeaCache (0.5) | TeaCache (1.1) |
|:-------------------------:|:---------------------------:|:--------------------:|:---------------------:|:---------------------:|
| ~25 s | ~16.7 s | ~15.6 s | ~13.79 s | ~11.9 s |
## Installation ## Installation
```shell ```shell

3
TeaCache4Lumina2/teacache_lumina2.py
View File

@ -68,7 +68,8 @@ def teacache_forward_working(
current_cache["accumulated_rel_l1_distance"] = 0.0 current_cache["accumulated_rel_l1_distance"] = 0.0
else: else:
if current_cache["previous_modulated_input"] is not None: if current_cache["previous_modulated_input"] is not None:
coefficients = [393.76566581, -603.50993606, 209.10239044, -23.00726601, 0.86377344] # taken from teacache_lumina_next.py # v1 coefficients，you can switch it to [225.7042019806413, -608.8453716535591, 304.1869942338369, 124.21267720116742, -1.4089066892956552] as v2
coefficients = [393.76566581, -603.50993606, 209.10239044, -23.00726601, 0.86377344]
rescale_func = np.poly1d(coefficients) rescale_func = np.poly1d(coefficients)
prev_mod_input = current_cache["previous_modulated_input"] prev_mod_input = current_cache["previous_modulated_input"]
prev_mean = prev_mod_input.abs().mean() prev_mean = prev_mod_input.abs().mean()