Instructions to use wfen/Cosmos3-Nano-FP8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use wfen/Cosmos3-Nano-FP8 with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("wfen/Cosmos3-Nano-FP8", dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
Update README.md
Browse files
README.md
CHANGED
|
@@ -1,5 +1,7 @@
|
|
| 1 |
---
|
| 2 |
license: openmdw-1.0
|
|
|
|
|
|
|
| 3 |
---
|
| 4 |
|
| 5 |
# Cosmos3-Nano — FP8 (safetensors)
|
|
@@ -62,5 +64,4 @@ on EC-01..04 at 8 steps and EC-01 at 35 steps (seed 123, UniPC `flow_shift=10.0`
|
|
| 62 |
- `action_gen=False` build (matches the reference quantized checkpoint, whose `.pt` is
|
| 63 |
action-adapter-stripped). No action-conditioned generation from this checkpoint.
|
| 64 |
- Verification at the smoke setting (1 frame / 480×480); full-res 720p/189-frame is out of scope.
|
| 65 |
-
- FP8 compute is ModelOpt fake-quant (compute in bf16); real-FP4/FP8 kernel speedups are out of scope.
|
| 66 |
-
|
|
|
|
| 1 |
---
|
| 2 |
license: openmdw-1.0
|
| 3 |
+
base_model:
|
| 4 |
+
- nvidia/Cosmos3-Nano
|
| 5 |
---
|
| 6 |
|
| 7 |
# Cosmos3-Nano — FP8 (safetensors)
|
|
|
|
| 64 |
- `action_gen=False` build (matches the reference quantized checkpoint, whose `.pt` is
|
| 65 |
action-adapter-stripped). No action-conditioned generation from this checkpoint.
|
| 66 |
- Verification at the smoke setting (1 frame / 480×480); full-res 720p/189-frame is out of scope.
|
| 67 |
+
- FP8 compute is ModelOpt fake-quant (compute in bf16); real-FP4/FP8 kernel speedups are out of scope.
|
|
|