Instructions to use Sawfwair/Cosmos3-Super-Text2Image-4Step-MLX-4bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use Sawfwair/Cosmos3-Super-Text2Image-4Step-MLX-4bit with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] hf download Sawfwair/Cosmos3-Super-Text2Image-4Step-MLX-4bit --local-dir Cosmos3-Super-Text2Image-4Step-MLX-4bit
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
Add native MLX qualification evidence
Browse files- .gitattributes +1 -0
- README.md +21 -0
- SHA256SUMS +2 -1
- examples/qualification-768-seed42.png +3 -0
.gitattributes
CHANGED
|
@@ -34,3 +34,4 @@ saved_model/**/* 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 |
text_tokenizer/tokenizer.json 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 |
text_tokenizer/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 37 |
+
examples/qualification-768-seed42.png filter=lfs diff=lfs merge=lfs -text
|
README.md
CHANGED
|
@@ -76,6 +76,27 @@ not by itself establish visual-quality parity with NVIDIA's BF16 checkpoint.
|
|
| 76 |
Record device, prompt, seed, output digest, peak memory, and elapsed time when
|
| 77 |
qualifying a generated image.
|
| 78 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 79 |
## License
|
| 80 |
|
| 81 |
This model is distributed under the
|
|
|
|
| 76 |
Record device, prompt, seed, output digest, peak memory, and elapsed time when
|
| 77 |
qualifying a generated image.
|
| 78 |
|
| 79 |
+
### Native qualification
|
| 80 |
+
|
| 81 |
+
The published artifact completed an end-to-end 768x768 native generation on an
|
| 82 |
+
Apple M4 Max MacBook Pro with 128 GB of unified memory and macOS 26.5.2. The
|
| 83 |
+
command used a debug build of `mere.run`, four steps, seed 42, and this prompt:
|
| 84 |
+
|
| 85 |
+
> A glass observatory above a bioluminescent ocean at blue hour
|
| 86 |
+
|
| 87 |
+
| Measurement | Result |
|
| 88 |
+
| --- | --- |
|
| 89 |
+
| Wall time | 60.85 seconds |
|
| 90 |
+
| Maximum resident memory | 38,657,916,928 bytes |
|
| 91 |
+
| Peak memory footprint | 53,394,599,464 bytes |
|
| 92 |
+
| Swaps | 0 |
|
| 93 |
+
| Output SHA-256 | `006a73d23850c1ab6ae9c05e5521109a8a2678f69e42fe362cd3a837d77c7495` |
|
| 94 |
+
|
| 95 |
+
This is native runtime smoke evidence, not a BF16 parity or broad visual-quality
|
| 96 |
+
evaluation.
|
| 97 |
+
|
| 98 |
+

|
| 99 |
+
|
| 100 |
## License
|
| 101 |
|
| 102 |
This model is distributed under the
|
SHA256SUMS
CHANGED
|
@@ -3,10 +3,11 @@ ce1decfeb17ef0a42c419c3fdcfa6ee03d19125690956760f47876382404e043 CONVERSION.jso
|
|
| 3 |
7d9e73df359fc9c0fcffdc18045018f55f124e5a6b6f71268a93d63c56e1a997 EXPLAINABILITY.md
|
| 4 |
0eddb5d056e34ec31f6247f3c888543753403d519870717b915308bf20664aa6 OPENMDW-1.1-LICENSE.html
|
| 5 |
0298f236e550d4064cd8d83babad8dfc791295bba602210ca6c8106a993ca699 PRIVACY.md
|
| 6 |
-
|
| 7 |
b59cf989cb66eb3b1c5e0692d04c42b9a39751801259b225733ab9493e08f03b SAFETY.md
|
| 8 |
bdcd644dd427a874105c62c7bc5aa0d8f8ccdea7c6a16ec4b0af8a0e78453eeb SOURCE_MANIFEST.json
|
| 9 |
ee730f0b9b5bf548d256f0f40f2211db58f64078f41b1f885ba6af562665de34 UPSTREAM_MODEL_CARD.md
|
|
|
|
| 10 |
f09a2d15751a6372648dba75f351045a20e8b1542d61d1e41cfe97ad309749c5 model_index.json
|
| 11 |
3f90c0c5507576b694c484e42a7e262d450c727d3ad566ef0c060e596209db82 modular_model_index.json
|
| 12 |
0ccbf84dc819354e50be8abd941fad37a2b2936698b24824622bdeed7d8e7735 scheduler/scheduler_config.json
|
|
|
|
| 3 |
7d9e73df359fc9c0fcffdc18045018f55f124e5a6b6f71268a93d63c56e1a997 EXPLAINABILITY.md
|
| 4 |
0eddb5d056e34ec31f6247f3c888543753403d519870717b915308bf20664aa6 OPENMDW-1.1-LICENSE.html
|
| 5 |
0298f236e550d4064cd8d83babad8dfc791295bba602210ca6c8106a993ca699 PRIVACY.md
|
| 6 |
+
1804f608baac0425551c49ee55db72de33865bbe20bba6bbc6725ddcb0707e86 README.md
|
| 7 |
b59cf989cb66eb3b1c5e0692d04c42b9a39751801259b225733ab9493e08f03b SAFETY.md
|
| 8 |
bdcd644dd427a874105c62c7bc5aa0d8f8ccdea7c6a16ec4b0af8a0e78453eeb SOURCE_MANIFEST.json
|
| 9 |
ee730f0b9b5bf548d256f0f40f2211db58f64078f41b1f885ba6af562665de34 UPSTREAM_MODEL_CARD.md
|
| 10 |
+
006a73d23850c1ab6ae9c05e5521109a8a2678f69e42fe362cd3a837d77c7495 examples/qualification-768-seed42.png
|
| 11 |
f09a2d15751a6372648dba75f351045a20e8b1542d61d1e41cfe97ad309749c5 model_index.json
|
| 12 |
3f90c0c5507576b694c484e42a7e262d450c727d3ad566ef0c060e596209db82 modular_model_index.json
|
| 13 |
0ccbf84dc819354e50be8abd941fad37a2b2936698b24824622bdeed7d8e7735 scheduler/scheduler_config.json
|
examples/qualification-768-seed42.png
ADDED
|
Git LFS Details
|