dverdu-freepik commited on
Commit
5c69589
1 Parent(s): 37e9e0c

fix: Update readme

Browse files
Files changed (3) hide show
  1. .gitattributes +0 -1
  2. README.md +9 -9
  3. sample_images/.DS_Store +0 -0
.gitattributes CHANGED
@@ -33,5 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
- flux1-lite_sample.png filter=lfs diff=lfs merge=lfs -text
37
  *.png filter=lfs diff=lfs merge=lfs -text
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
36
  *.png filter=lfs diff=lfs merge=lfs -text
README.md CHANGED
@@ -11,13 +11,13 @@ tags:
11
  - text-to-image
12
  ---
13
 
14
- ![Flux.1 Lite](./sample_images/flux1-lite-8B_sample.png)
15
 
16
  # Flux.1 Lite
17
 
18
  We are thrilled to announce the alpha release of Flux.1 Lite, an 8B parameter transformer model distilled from the FLUX.1-dev model. This version uses 7 GB less RAM and runs 23% faster while maintaining the same precision (bfloat16) as the original model.
19
 
20
- ![Flux.1 Lite vs FLUX.1-dev](./sample_images/models_comparison.png)
21
 
22
  ## Text-to-Image
23
 
@@ -60,16 +60,16 @@ image.save("output.png")
60
 
61
  Inspired by [Ostris](https://ostris.com/2024/09/07/skipping-flux-1-dev-blocks/) findings, we analyzed the mean squared error (MSE) between the input and output of each block to quantify their contribution to the final result, revealing significant variability.
62
 
63
- As Ostris pointed out, not all blocks contribute equally. While skipping just one of the early MMDIT blocks can significantly impact model performance, skipping the rest of the blocks does not have a significant impact over the final image quality.
64
 
65
- ![Flux.1 Lite generated image](./sample_images/skip_blocks/generated_img.png)
66
- ![MSE MMDIT](./sample_images/skip_blocks/mse_mmdit_img.png)
67
- ![MSE DIT](./sample_images/skip_blocks/mse_dit_img.png)
68
 
69
  Furthermore, as displayed in the following image, only when you skip one of the first MMDIT blocks, the performance of the model severely impacts the model's performance.
70
 
71
- ![Skip one MMDIT block](./sample_images/skip_blocks/skip_one_MMDIT_block.png)
72
- ![Skip one DIT block](./sample_images/skip_blocks/skip_one_DIT_block.png)
73
 
74
  ## Future work
75
 
@@ -78,7 +78,7 @@ Stay tuned! Our goal is to distill FLUX.1-dev further until it can run smoothly
78
  ## ComfyUI
79
 
80
  We've also crafted a ComfyUI workflow to make using Flux.1 Lite even more seamless! Find it in `comfy/flux.1-lite_workflow.json`.
81
- ![ComfyUI workflow](./comfy/flux.1-lite_workflow.png)
82
 
83
  The safetensors checkpoint is available here: [flux.1-lite-8B-alpha.safetensors](flux.1-lite-8B-alpha.safetensors)
84
 
 
11
  - text-to-image
12
  ---
13
 
14
+ ![Flux.1 Lite](sample_images/flux1-lite-8B_sample.png)
15
 
16
  # Flux.1 Lite
17
 
18
  We are thrilled to announce the alpha release of Flux.1 Lite, an 8B parameter transformer model distilled from the FLUX.1-dev model. This version uses 7 GB less RAM and runs 23% faster while maintaining the same precision (bfloat16) as the original model.
19
 
20
+ ![Flux.1 Lite vs FLUX.1-dev](sample_images/models_comparison.png)
21
 
22
  ## Text-to-Image
23
 
 
60
 
61
  Inspired by [Ostris](https://ostris.com/2024/09/07/skipping-flux-1-dev-blocks/) findings, we analyzed the mean squared error (MSE) between the input and output of each block to quantify their contribution to the final result, revealing significant variability.
62
 
63
+ As Ostris pointed out, not all blocks contribute equally. While skipping just one of the early MMDiT blocks can significantly impact model performance, skipping the rest of the blocks does not have a significant impact over the final image quality.
64
 
65
+ ![Flux.1 Lite generated image](sample_images/skip_blocks/generated_img.png)
66
+ ![MSE MMDIT](sample_images/skip_blocks/mse_mmdit_img.png)
67
+ ![MSE DIT](sample_images/skip_blocks/mse_dit_img.png)
68
 
69
  Furthermore, as displayed in the following image, only when you skip one of the first MMDIT blocks, the performance of the model severely impacts the model's performance.
70
 
71
+ ![Skip one MMDIT block](sample_images/skip_blocks/skip_one_MMDIT_block.png)
72
+ ![Skip one DIT block](sample_images/skip_blocks/skip_one_DIT_block.png)
73
 
74
  ## Future work
75
 
 
78
  ## ComfyUI
79
 
80
  We've also crafted a ComfyUI workflow to make using Flux.1 Lite even more seamless! Find it in `comfy/flux.1-lite_workflow.json`.
81
+ ![ComfyUI workflow](comfy/flux.1-lite_workflow.png)
82
 
83
  The safetensors checkpoint is available here: [flux.1-lite-8B-alpha.safetensors](flux.1-lite-8B-alpha.safetensors)
84
 
sample_images/.DS_Store DELETED
Binary file (6.15 kB)