Downloads · 30 days
0
zimage-turbo-42112d9b-latent-eval
LongDuy/zimage-turbo-42112d9b-latent-eval
zimage-turbo-42112d9b-latent-eval is a machine learning model from LongDuy. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for safetensors.
LoRA trained on historical task 42112d9b-5ddd-436d-97b0-d9d34373793f from kipgin/56dataset at pinned dataset revision 6b91363d16a5b15404ec244674886d446c8f6fff.
Access
Public
Updated Sep 19, 2026
Repo size
302 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors302 MB · 100%
At a glance
- Library
- safetensors
- Access
- Public
- Created
- Sep 19, 2026
- Updated
- Sep 19, 2026
- SHA
- 2afdcaf0
Base models
Datasets
Model card
From the Hugging Face model README
Z-Image Turbo historical fine-tune replay
LoRA trained on historical task 42112d9b-5ddd-436d-97b0-d9d34373793f from kipgin/56_dataset at pinned dataset revision 6b91363d16a5b15404ec244674886d446c8f6fff.
Training
- Base model:
gradients-io-tournaments/Z-Image-Turbo - Images: 12 train, 2 test
- Steps: 800
- Hardware allocation: NVIDIA L40S, 12 CPUs
- Time allowance: 4,050 seconds, applying a 1.5x allowance for this host
- LoRA rank: linear 32, convolution 16
- Optimizer: Muon
The uploaded checkpoint/last.safetensors uses Comfy-compatible text-encoder key names. The source training artifact contained 504 AI Toolkit text-encoder tensors that were deterministically renamed before upload.
Evaluation
G.O.D PR #1369 latent prediction MSE:
| Split | Loss |
|---|---|
| Captioned | 0.604045581 |
| Empty caption | 0.605191492 |
| Equal-weight mean | 0.604618536 |
Lower is better. These values are not directly comparable with the historical pixel-RGB L2 opponent losses.
Source code: https://github.com/longduyngn/image_gen_3826/tree/feat/latent-eval-zimage-replay
- Library
- safetensors
- Created
- Sep 19, 2026
- Updated
- Sep 19, 2026