Text-to-Image
Diffusers
Safetensors
English
stable-diffusion-xl
lora
rectified-flow
flow-matching
geometric-deep-learning
qwen
aleph
geolip
experimental
Instructions to use AbstractPhil/geolip-sdxl-aleph with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use AbstractPhil/geolip-sdxl-aleph with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-xl-base-1.0", dtype=torch.bfloat16, device_map="cuda") pipe.load_lora_weights("AbstractPhil/geolip-sdxl-aleph") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Inference
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- Draw Things
- DiffusionBee
Upload phase_0_fid_score/SUMMARY.md with huggingface_hub
Browse files- phase_0_fid_score/SUMMARY.md +18 -0
phase_0_fid_score/SUMMARY.md
ADDED
|
@@ -0,0 +1,18 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Phase-0 proto-LoRA FID / KID
|
| 2 |
+
|
| 3 |
+
- prompts per cell: **100**
|
| 4 |
+
- CFG stages: [1.0, 3.0, 5.0]
|
| 5 |
+
- reference: first 100 dataset images (resized as `precompute`)
|
| 6 |
+
- **KID** is the trustworthy ranking at this N; **FID** is inflated/noisy — read it relatively.
|
| 7 |
+
|
| 8 |
+
| config | cfg 1 (FID / KID) | cfg 3 (FID / KID) | cfg 5 (FID / KID) |
|
| 9 |
+
|---|---|---|---|
|
| 10 |
+
| `base` | 183.0 / 0.0070 | 263.6 / 0.0626 | 301.5 / 0.1025 |
|
| 11 |
+
| `clipg_pooled` | 162.5 / -0.0004 | 225.5 / 0.0292 | 272.0 / 0.0719 |
|
| 12 |
+
| `clipg_seq_pooled` | 218.5 / 0.0099 | 273.6 / 0.0624 | 298.4 / 0.0922 |
|
| 13 |
+
| `aleph_clipg_pooled` | 154.5 / -0.0022 | 232.5 / 0.0324 | 288.7 / 0.0772 |
|
| 14 |
+
| `aleph_clipg_seq_pooled` | 241.5 / 0.0228 | 294.6 / 0.0837 | 294.9 / 0.0934 |
|
| 15 |
+
| `clipl` | 180.8 / 0.0052 | 275.3 / 0.0676 | 308.3 / 0.1010 |
|
| 16 |
+
| `aleph_clipl` | 178.2 / 0.0035 | 259.6 / 0.0505 | 297.6 / 0.0806 |
|
| 17 |
+
| `aleph_clipl_clipg_pooled` | 172.7 / -0.0006 | 242.4 / 0.0394 | 287.4 / 0.0788 |
|
| 18 |
+
| `aleph_clipl_clipg_seq_pooled` | 251.9 / 0.0346 | 294.6 / 0.0886 | 298.8 / 0.0990 |
|