Link reference data generation pipeline
Browse files
README.md
CHANGED
|
@@ -19,10 +19,7 @@ ResNet-50 weights associated with
|
|
| 19 |
[*A MIMO Wireless Channel Foundation Model via CIR-CSI Consistency*](https://arxiv.org/abs/2502.11965).
|
| 20 |
|
| 21 |
- Code: [GREAT-ISAC/CSI-CLIP](https://github.com/GREAT-ISAC/CSI-CLIP)
|
| 22 |
-
|
| 23 |
-
Released by **[GREAT Wireless AI](https://github.com/GREAT-ISAC)**. The
|
| 24 |
-
`GREAT-ISAC` GitHub handle is retained for continuity with existing papers and
|
| 25 |
-
community links.
|
| 26 |
- Paper: [arXiv:2502.11965](https://arxiv.org/abs/2502.11965)
|
| 27 |
- IEEE: [ICMLCN 2025](https://ieeexplore.ieee.org/abstract/document/11140262/)
|
| 28 |
|
|
@@ -36,7 +33,7 @@ an optimizer, scheduler, training data, or a downstream task head.
|
|
| 36 |
| `model.safetensors` | Model-only pre-trained weights | Recommended for safe standalone loading and feature extraction |
|
| 37 |
| `model.pth` | Model-only PyTorch checkpoint with a `model` key | Compatibility with the existing fine-tuning scripts |
|
| 38 |
|
| 39 |
-
These are foundation-model pre-training weights, not task-specific checkpoints
|
| 40 |
for positioning, beam management, or LOS/NLOS classification. A downstream
|
| 41 |
head must be trained before task-level inference.
|
| 42 |
|
|
@@ -100,8 +97,10 @@ python load_pretrained.py \
|
|
| 100 |
|
| 101 |
The model was pre-trained on simulated DeepMIMO CSI covering multiple wireless
|
| 102 |
scenarios. Generated training arrays are not included in this model repository.
|
| 103 |
-
A **reproducible data-generation pipeline**
|
| 104 |
-
|
|
|
|
|
|
|
| 105 |
|
| 106 |
## Intended use
|
| 107 |
|
|
@@ -123,7 +122,7 @@ the simulator configuration, preprocessing, and split definitions.
|
|
| 123 |
|
| 124 |
## License
|
| 125 |
|
| 126 |
-
The original CSI-CLIP code, these model weights, and the
|
| 127 |
data-generation scripts are released under the
|
| 128 |
[Creative Commons Attribution-NonCommercial 4.0 International](https://creativecommons.org/licenses/by-nc/4.0/)
|
| 129 |
license (**CC BY-NC 4.0**). Attribution is required and commercial use is not
|
|
|
|
| 19 |
[*A MIMO Wireless Channel Foundation Model via CIR-CSI Consistency*](https://arxiv.org/abs/2502.11965).
|
| 20 |
|
| 21 |
- Code: [GREAT-ISAC/CSI-CLIP](https://github.com/GREAT-ISAC/CSI-CLIP)
|
| 22 |
+
- Reference data generation: [GREAT-ISAC/Channel-Simulation-Data](https://github.com/GREAT-ISAC/Channel-Simulation-Data)
|
|
|
|
|
|
|
|
|
|
| 23 |
- Paper: [arXiv:2502.11965](https://arxiv.org/abs/2502.11965)
|
| 24 |
- IEEE: [ICMLCN 2025](https://ieeexplore.ieee.org/abstract/document/11140262/)
|
| 25 |
|
|
|
|
| 33 |
| `model.safetensors` | Model-only pre-trained weights | Recommended for safe standalone loading and feature extraction |
|
| 34 |
| `model.pth` | Model-only PyTorch checkpoint with a `model` key | Compatibility with the existing fine-tuning scripts |
|
| 35 |
|
| 36 |
+
These are channel-foundation-model pre-training weights, not task-specific checkpoints
|
| 37 |
for positioning, beam management, or LOS/NLOS classification. A downstream
|
| 38 |
head must be trained before task-level inference.
|
| 39 |
|
|
|
|
| 97 |
|
| 98 |
The model was pre-trained on simulated DeepMIMO CSI covering multiple wireless
|
| 99 |
scenarios. Generated training arrays are not included in this model repository.
|
| 100 |
+
A **reproducible, model-compatible reference data-generation pipeline** is
|
| 101 |
+
available in [Channel Simulation Data](https://github.com/GREAT-ISAC/Channel-Simulation-Data).
|
| 102 |
+
Its committed O1_60 configuration is one runnable reference example; it does
|
| 103 |
+
not reconstruct the complete multi-scenario checkpoint training data.
|
| 104 |
|
| 105 |
## Intended use
|
| 106 |
|
|
|
|
| 122 |
|
| 123 |
## License
|
| 124 |
|
| 125 |
+
The original CSI-CLIP code, these model weights, and the repository-owned
|
| 126 |
data-generation scripts are released under the
|
| 127 |
[Creative Commons Attribution-NonCommercial 4.0 International](https://creativecommons.org/licenses/by-nc/4.0/)
|
| 128 |
license (**CC BY-NC 4.0**). Attribution is required and commercial use is not
|