VAE Lyra XL Adaptive Cantor Illustrious
Multi-modal VAE trained with custom CLIP weights.
Model source
- Kind
- Other
- File size
- 913 MB–6.46 GB
- Version
- vRevision 3cba3df
- Updated
- Oct 9, 2026
- Downloads
- 190
- Source
- Hugging Face
Source description
VAE Lyra 🎵 - Illustrious Edition
Multi-modal VAE trained with custom CLIP weights.
CLIP Encoders
Uses CLIP weights from AbstractPhil/clips:
- CLIP-L:
IllustriousV01_clip_l.safetensors - CLIP-G:
IllustriousV01_clip_g.safetensors
CLIP Skip: 2 (penultimate layer)
Model Details
- Fusion Strategy: adaptive_cantor
- Latent Dimension: 2048
- Training Steps: 12,125
- Best Loss: 0.0377
- Prompt Source: booru
Quick Load (Safetensors)
from safetensors.torch import load_file
# Load just the weights (fast)
state_dict = load_file("weights/lyra_illustrious_best.safetensors")
# Or specific step
state_dict = load_file("weights/lyra_illustrious_step_5000.safetensors")
T5 Input Format
T5 receives a different input than CLIP to enable richer semantic understanding:
CLIP sees: "masterpiece, 1girl, blue hair, school uniform, smile"
T5 sees: "masterpiece, 1girl, blue hair, school uniform, smile ¶ A cheerful schoolgirl with blue hair smiling warmly"
The pilcrow (¶) separator acts as a mode-switch token.
Learned Parameters
Alpha (Visibility):
- clip_g: 0.7316
- clip_l: 0.7316
- t5_xl_g: 0.7339
- t5_xl_l: 0.7451
Beta (Capacity):
- clip_l_t5_xl_l: 0.5709
- clip_g_t5_xl_g: 0.5763
Usage
from lyra_xl_multimodal import load_lyra_from_hub
model = load_lyra_from_hub("AbstractPhil/vae-lyra-xl-adaptive-cantor-illustrious")
model.eval()
inputs = {
"clip_l": clip_l_embeddings, # [batch, 77, 768]
"clip_g": clip_g_embeddings, # [batch, 77, 1280]
"t5_xl_l": t5_xl_embeddings, # [batch, 512, 2048]
"t5_xl_g": t5_xl_embeddings # [batch, 512, 2048]
}
recons, mu, logvar, _ = model(inputs, target_modalities=["clip_l", "clip_g"])
Files
model.pt- Full checkpoint (model + optimizer + scheduler)checkpoint_lyra_illustrious_XXXX.pt- Step checkpointsconfig.json- Training configurationweights/lyra_illustrious_best.safetensors- Best model weights onlyweights/lyra_illustrious_step_XXXX.safetensors- Step checkpoints (weights only)
Sources
1 source
Sources
1 sourceVerified Oct 9
Download files
Model artifacts
64 artifacts
Model artifacts
64 artifactscheckpoint_2000_lyra_illustrious.pt
pt · 2.59 GB · SHA-256 f482e702c527…5b4f · Hugging Face
Download filecheckpoint_3000_lyra_illustrious.pt
pt · 2.59 GB · SHA-256 a3037d572b70…61f6 · Hugging Face
Download filecheckpoint_lyra_illustrious_10000.pt
pt · 2.58 GB · SHA-256 4a0c64b27be8…9624 · Hugging Face
Download filecheckpoint_lyra_illustrious_11000.pt
pt · 2.58 GB · SHA-256 aa1b58d4df70…9c5e · Hugging Face
Download filecheckpoint_lyra_illustrious_12000.pt
pt · 2.58 GB · SHA-256 56ec3fb0050b…8fc8 · Hugging Face
Download filecheckpoint_lyra_illustrious_13000.pt
pt · 2.58 GB · SHA-256 9f2586f510ef…3335 · Hugging Face
Download filecheckpoint_lyra_illustrious_14000.pt
pt · 2.58 GB · SHA-256 c121f1e62c52…b69f · Hugging Face
Download filecheckpoint_lyra_illustrious_15000.pt
pt · 2.58 GB · SHA-256 559139a15874…ebf3 · Hugging Face
Download filecheckpoint_lyra_illustrious_16000.pt
pt · 2.58 GB · SHA-256 d18ced48fe47…3e37 · Hugging Face
Download filecheckpoint_lyra_illustrious_17000.pt
pt · 2.58 GB · SHA-256 34a00e5d2c71…5257 · Hugging Face
Download filecheckpoint_lyra_illustrious_18000.pt
pt · 2.58 GB · SHA-256 05eb58f065e4…47e0 · Hugging Face
Download filecheckpoint_lyra_illustrious_19000.pt
pt · 2.58 GB · SHA-256 81c17327c43f…db28 · Hugging Face
Download filecheckpoint_lyra_illustrious_20000.pt
pt · 2.58 GB · SHA-256 1e0e322e1862…bc4f · Hugging Face
Download filecheckpoint_lyra_illustrious_21000.pt
pt · 2.58 GB · SHA-256 9e444c33d6d2…760b · Hugging Face
Download filecheckpoint_lyra_illustrious_22000.pt
pt · 2.58 GB · SHA-256 d549f7ae3863…3b81 · Hugging Face
Download filecheckpoint_lyra_illustrious_23000.pt
pt · 2.58 GB · SHA-256 1be74662c7b0…3b18 · Hugging Face
Download filecheckpoint_lyra_illustrious_24000.pt
pt · 2.58 GB · SHA-256 7739e82a17db…4a99 · Hugging Face
Download filecheckpoint_lyra_illustrious_25000.pt
pt · 2.58 GB · SHA-256 9f5a4fcf3a09…5f4b · Hugging Face
Download filecheckpoint_lyra_illustrious_26000.pt
pt · 2.58 GB · SHA-256 ac41efbdf58c…205e · Hugging Face
Download filecheckpoint_lyra_illustrious_27000.pt
pt · 2.58 GB · SHA-256 b5c25ee486f3…6c5a · Hugging Face
Download filecheckpoint_lyra_illustrious_28000.pt
pt · 2.58 GB · SHA-256 d329624b634e…a7ef · Hugging Face
Download filecheckpoint_lyra_illustrious_29000.pt
pt · 2.58 GB · SHA-256 c2cef2b70e01…1e87 · Hugging Face
Download filecheckpoint_lyra_illustrious_30000.pt
pt · 2.58 GB · SHA-256 86c942ae97a9…05ae · Hugging Face
Download filecheckpoint_lyra_illustrious_31000.pt
pt · 2.58 GB · SHA-256 873cb95e459a…c2f7 · Hugging Face
Download filecheckpoint_lyra_illustrious_32000.pt
pt · 2.58 GB · SHA-256 372b974d8942…e562 · Hugging Face
Download filecheckpoint_lyra_illustrious_33000.pt
pt · 2.58 GB · SHA-256 be1a2eab6dcb…ad13 · Hugging Face
Download filecheckpoint_lyra_illustrious_34000.pt
pt · 2.58 GB · SHA-256 733ac1455fbc…4dd7 · Hugging Face
Download filecheckpoint_lyra_illustrious_35000.pt
pt · 2.58 GB · SHA-256 0ea1ed5b2d84…9b5e · Hugging Face
Download filecheckpoint_lyra_illustrious_36000.pt
pt · 2.58 GB · SHA-256 88c3b5ad8fb1…1f10 · Hugging Face
Download filecheckpoint_lyra_illustrious_37000.pt
pt · 2.58 GB · SHA-256 935f4b9d5546…a951 · Hugging Face
Download filecheckpoint_lyra_illustrious_38000.pt
pt · 2.58 GB · SHA-256 e11ee279ce16…21f3 · Hugging Face
Download filecheckpoint_lyra_illustrious_39000.pt
pt · 2.58 GB · SHA-256 b8ceb4759500…3132 · Hugging Face
Download filecheckpoint_lyra_illustrious_4000.pt
pt · 2.58 GB · SHA-256 28196bc43bdf…3fd3 · Hugging Face
Download filecheckpoint_lyra_illustrious_40000.pt
pt · 2.58 GB · SHA-256 2464e6d92116…a40d · Hugging Face
Download filecheckpoint_lyra_illustrious_41000.pt
pt · 2.58 GB · SHA-256 9812ff84c4dc…e40a · Hugging Face
Download filecheckpoint_lyra_illustrious_5000.pt
pt · 2.58 GB · SHA-256 1cb87a8919ff…1376 · Hugging Face
Download filecheckpoint_lyra_illustrious_6000.pt
pt · 2.58 GB · SHA-256 f8cf28181fab…7a91 · Hugging Face
Download filecheckpoint_lyra_illustrious_7000.pt
pt · 2.58 GB · SHA-256 ed301ae8a38b…c25b · Hugging Face
Download filecheckpoint_lyra_illustrious_8000.pt
pt · 2.58 GB · SHA-256 0bc0d3e32f98…1393 · Hugging Face
Download filecheckpoint_lyra_illustrious_9000.pt
pt · 2.58 GB · SHA-256 b26ba85f69a6…01a5 · Hugging Face
Download fileillustriousXL_v01.safetensors
safetensors · 6.46 GB · SHA-256 3e15ba00387d…5aa9 · Hugging Face
Download fileweights/lyra_illustrious_best.safetensors
safetensors · 913 MB · SHA-256 3c4bf295b52c…631d · Hugging Face
Download fileweights/lyra_illustrious_step_10000.safetensors
safetensors · 913 MB · SHA-256 188121373ba6…9bb1 · Hugging Face
Download fileweights/lyra_illustrious_step_11000.safetensors
safetensors · 913 MB · SHA-256 64eaae2334d1…337c · Hugging Face
Download fileweights/lyra_illustrious_step_12000.safetensors
safetensors · 913 MB · SHA-256 ab2841ef2b50…189b · Hugging Face
Download fileweights/lyra_illustrious_step_13000.safetensors
safetensors · 913 MB · SHA-256 8d744ab64651…0ea7 · Hugging Face
Download fileweights/lyra_illustrious_step_14000.safetensors
safetensors · 913 MB · SHA-256 b8a83950c0de…2b6e · Hugging Face
Download fileweights/lyra_illustrious_step_15000.safetensors
safetensors · 913 MB · SHA-256 ec7becbada8a…cb4a · Hugging Face
Download fileweights/lyra_illustrious_step_16000.safetensors
safetensors · 913 MB · SHA-256 c8a4cc9c420d…06ea · Hugging Face
Download fileweights/lyra_illustrious_step_17000.safetensors
safetensors · 913 MB · SHA-256 c0b79191c4c4…26f9 · Hugging Face
Download fileweights/lyra_illustrious_step_18000.safetensors
safetensors · 913 MB · SHA-256 3aabee6d312e…5ea8 · Hugging Face
Download fileweights/lyra_illustrious_step_19000.safetensors
safetensors · 913 MB · SHA-256 74c82b30c835…9ba6 · Hugging Face
Download fileweights/lyra_illustrious_step_20000.safetensors
safetensors · 913 MB · SHA-256 f92cadd3e83d…233d · Hugging Face
Download fileweights/lyra_illustrious_step_21000.safetensors
safetensors · 913 MB · SHA-256 2cee8a2111ef…0f15 · Hugging Face
Download fileweights/lyra_illustrious_step_22000.safetensors
safetensors · 913 MB · SHA-256 dd524069d8f8…cdb9 · Hugging Face
Download fileweights/lyra_illustrious_step_23000.safetensors
safetensors · 913 MB · SHA-256 ddb59a637a96…9f44 · Hugging Face
Download fileweights/lyra_illustrious_step_24000.safetensors
safetensors · 913 MB · SHA-256 f619118f7c97…d186 · Hugging Face
Download fileweights/lyra_illustrious_step_25000.safetensors
safetensors · 913 MB · SHA-256 c1d63fa0ac84…60cc · Hugging Face
Download fileweights/lyra_illustrious_step_26000.safetensors
safetensors · 913 MB · SHA-256 27a9541f0343…3da9 · Hugging Face
Download fileweights/lyra_illustrious_step_27000.safetensors
safetensors · 913 MB · SHA-256 af5adb707c94…c44a · Hugging Face
Download fileweights/lyra_illustrious_step_28000.safetensors
safetensors · 913 MB · SHA-256 f39821c978ea…0fbf · Hugging Face
Download fileSource excerpts
3 excerpts
Source excerpts
3 excerpts--- tags: - vae - multimodal - clip - t5 - sdxl - illustrious - adaptive-cantor - booru license: mit --- # VAE Lyra 🎵 - Illustrious Edition Multi-modal VAE trained with **custom CLIP weights**. ## CLIP Encoders Uses CLIP weights from `AbstractPhil/clips`: - CLIP-L: `IllustriousV01_clip_l.safetensors` - CLIP-G: `IllustriousV01_clip_g.safetensors` **CLIP Skip**: 2 (penultimate layer) ## Model Details - **Fusion Strategy**: adaptive_cantor - **Latent Dimension**: 2048 - **Training Steps**: 12,125 - **Best Loss**: 0.0377 - **Prompt Source**: booru ## Quick Load (Safetensors) ```python from safetensors.torch import load_file # Load just the weights (fast) state_dict = load_file("weights/lyra_illustrious_best.safetensors") # Or specific step state_dict = load_file("weights/lyra_illustrious_step_5000.safetensors") ``` ## T5 Input Format T5 receives a different input than CLIP to enable richer semantic understanding: ``` CLIP sees: "masterpiece, 1girl, blue hair, school uniform, smile" T5 sees: "masterpiece, 1girl, blue hair, school uniform, smile ¶ A cheerful schoolgirl with blue hair smiling warmly" ``` The pilcrow (`¶`) separator acts as a mode-switch token. ## Learned Parameters **Alpha (Visibility):** - clip_g: 0.7316 - clip_l: 0.7316 - t5_xl_g: 0.7339 - t5_xl_l: 0.7451 **Beta (Capacity):** - clip_l_t5_xl_l: 0.5709 - clip_g_t5_xl_g: 0.5763 ## Usage ```python from lyra_xl_multimodal import load_lyra_from_hub model = load_lyra_from_hub("AbstractPhil/vae-lyra-xl-adaptive-cantor-illustrious") model.eval() inputs = { "clip_l": clip_l_embeddings, # [batch, 77, 768] "clip_g": clip_g_embeddings, # [batch, 77, 1280] "t5_xl_l": t5_xl_embeddings, # [batch, 512, 2048] "t5_xl_g": t5_xl_embeddings # [batch, 512, 2048] } recons, mu, logvar, _ = model(inputs, target_modalities=["clip_l", "clip_g"]) ``` ## Files - `model.pt` - Full checkpoint (model + optimizer + scheduler) - `checkpoint_lyra_illustrious_XXXX.pt` - Step checkpoints - `config.json` - Training configuration - `weights/lyra_illustrious_best.safetensors` - Best model weights only - `weights/lyra_illustrious_step_XXXX.safetensors` - Step checkpoints (weights only)
Source context: 190 downloads · 1 likes · Repo AbstractPhil/vae-lyra-xl-adaptive-cantor-illustrious
Source context: 37 downloads · 1 likes · Repo AbstractPhil/vae-lyra-xl-adaptive-cantor-illustrious
