Skip to content

Commit ff0fafe

Browse files
LTMeyerLTMeyer
committed
Add tables to model READMEs
1 parent a45a9e4 commit ff0fafe

File tree

3 files changed

+67
-2
lines changed

3 files changed

+67
-2
lines changed

‎the_well/benchmark/models/tfno/README.md‎

Lines changed: 22 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -27,7 +27,6 @@ tags:
2727

2828
[The Well](https://github.com/PolymathicAI/the_well) is a 15TB dataset collection of physics simulations. This model is part of the models that have been benchmarked on the Well.
2929

30-
3130
The models have been trained for a fixed time of 12 hours or up to 500 epochs, whichever happens first. The training was performed on a NVIDIA H100 96GB GPU.
3231
In the time dimension, the context length was set to 4. The batch size was set to maximize the memory usage. We experiment with 5 different learning rates for each model on each dataset.
3332
We use the model performing best on the validation set to report test set results.
@@ -48,4 +47,26 @@ For benchmarking on the Well, we used the following parameters.
4847
| Blocks | 4 |
4948
| Hidden Size| 128 |
5049

50+
## Trained Model Versions
51+
52+
Below is the list of checkpoints available for the training of TFNO on different datasets of the Well.
53+
54+
| Dataset | Learning Rate | Epoch | VRMSE |
55+
|---------|----------------|-------|-------|
56+
| [acoustic_scattering_maze](https://huggingface.co/polymathic-ai/TFNO-acoustic_scattering) | 1E-3 | 27 | 0.5034 |
57+
| [active_matter](https://huggingface.co/polymathic-ai/TFNO-active_matter) | 1E-3 | 243 | 0.3342 |
58+
| [convective_envelope_rsg](https://huggingface.co/polymathic-ai/TFNO-convective_envelope_rsg) | 1E-3 | 13 | 0.0195 |
59+
| [gray_scott_reaction_diffusion](https://huggingface.co/polymathic-ai/TFNO-gray_scott_reaction_diffusion) | 5E-3 | 45 | 0.1784 |
60+
| [helmholtz_staircase](https://huggingface.co/polymathic-ai/TFNO-helmholtz_staircase) | 5E-4 | 131 | 0.00031 |
61+
| [MHD_64](https://huggingface.co/polymathic-ai/TFNO-MHD_64) | 1E-3 | 155 | 0.3347 |
62+
| [planetswe](https://huggingface.co/polymathic-ai/TFNO-planetswe) | 5E-4 | 49 | 0.1061 |
63+
| [post_neutron_star_merger](https://huggingface.co/polymathic-ai/TFNO-post_neutron_star_merger) | 5E-4 | 99 | 0.4064 |
64+
| [rayleigh_benard](https://huggingface.co/polymathic-ai/TFNO-rayleigh_benard) | 1E-4 | 31 | 0.8568 |
65+
| [rayleigh_taylor_instability](https://huggingface.co/polymathic-ai/TFNO-rayleigh_taylor_instability) | 1E-4 | 175 | 0.2251 |
66+
| [shear_flow](https://huggingface.co/polymathic-ai/TFNO-shear_flow) | 1E-3 | 24 | 0.3626 |
67+
| [supernova_explosion_64](https://huggingface.co/polymathic-ai/TFNO-supernova_explosion_64) | 1E-4 | 35 | 0.3645 |
68+
| [turbulence_gravity_cooling](https://huggingface.co/polymathic-ai/TFNO-turbulence_gravity_cooling) | 5E-4 | 10 | 0.2789 |
69+
| [turbulent_radiative_layer_2D](https://huggingface.co/polymathic-ai/TFNO-turbulent_radiative_layer_2D) | 1E-3 | 500 | 0.4938 |
70+
| [viscoelastic_instability](https://huggingface.co/polymathic-ai/TFNO-viscoelastic_instability) | 5E-3 | 199 | 0.7021 |
71+
5172
## Loading the model from Hugging Face

‎the_well/benchmark/models/unet_classic/README.md‎

Lines changed: 22 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -27,7 +27,6 @@ tags:
2727

2828
[The Well](https://github.com/PolymathicAI/the_well) is a 15TB dataset collection of physics simulations. This model is part of the models that have been benchmarked on the Well.
2929

30-
3130
The models have been trained for a fixed time of 12 hours or up to 500 epochs, whichever happens first. The training was performed on a NVIDIA H100 96GB GPU.
3231
In the time dimension, the context length was set to 4. The batch size was set to maximize the memory usage. We experiment with 5 different learning rates for each model on each dataset.
3332
We use the model performing best on the validation set to report test set results.
@@ -50,4 +49,26 @@ For benchmarking on the Well, we used the following parameters.
5049
| Up/Down Blocks | 4 |
5150
| Bottleneck Blocks | 1 |
5251

52+
## Trained Model Versions
53+
54+
Below is the list of checkpoints available for the training of U-Net on different datasets of the Well.
55+
56+
| Dataset | Learning Rate | Epochs | VRMSE |
57+
|---------|---------------|--------|-------|
58+
| [acoustic_scattering (maze)](https://huggingface.co/polymathic-ai/UNET-acoustic_scattering) | 1E-2 | 26 | 0.0395 |
59+
| [active_matter](https://huggingface.co/polymathic-ai/UNET-active_matter) | 5E-3 | 239 | 0.2609 |
60+
| [convective_envelope_rsg](https://huggingface.co/polymathic-ai/UNET-convective_envelope_rsg) | 5E-4 | 19 | 0.0701 |
61+
| [gray_scott_reaction_diffusion](https://huggingface.co/polymathic-ai/UNET-gray_scott_reaction_diffusion) | 1E-2 | 44 | 0.5870 |
62+
| [helmholtz_staircase](https://huggingface.co/polymathic-ai/UNET-helmholtz_staircase) | 1E-3 | 120 | 0.01655 |
63+
| [MHD_64](https://huggingface.co/polymathic-ai/UNET-MHD_64) | 5E-4 | 165 | 0.1988 |
64+
| [planetswe](https://huggingface.co/polymathic-ai/UNET-planetswe) | 1E-2 | 49 | 0.3498 |
65+
| [post_neutron_star_merger](https://huggingface.co/polymathic-ai/UNET-post_neutron_star_merger) | - | - | – |
66+
| [rayleigh_benard](https://huggingface.co/polymathic-ai/UNET-rayleigh_benard) | 1E-4 | 29 | 0.8448 |
67+
| [rayleigh_taylor_instability](https://huggingface.co/polymathic-ai/UNET-rayleigh_taylor_instability) | 5E-4 | 193 | 0.6140 |
68+
| [shear_flow](https://huggingface.co/polymathic-ai/UNET-shear_flow) | 5E-4 | 29 | 0.836 |
69+
| [supernova_explosion_64](https://huggingface.co/polymathic-ai/UNET-supernova_explosion_64) | 5E-4 | 46 | 0.3242 |
70+
| [turbulence_gravity_cooling](https://huggingface.co/polymathic-ai/UNET-turbulence_gravity_cooling) | 1E-3 | 14 | 0.3152 |
71+
| [turbulent_radiative_layer_2D](https://huggingface.co/polymathic-ai/UNET-turbulent_radiative_layer_2D) | 5E-3 | 500 | 0.2394 |
72+
| [viscoelastic_instability](https://huggingface.co/polymathic-ai/UNET-viscoelastic_instability) | 5E-4 | 198 | 0.3147 |
73+
5374
## Loading the model from Hugging Face

‎the_well/benchmark/models/unet_convnext/README.md‎

Lines changed: 23 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -50,4 +50,27 @@ For benchmarking on the Well, we used the following parameters.
5050
| Up/Down Blocks | 4 |
5151
| Bottleneck Blocks | 1 |
5252

53+
54+
## Trained Model Versions
55+
56+
Below is the list of checkpoints available for the training of CNextU-Net on different datasets of the Well.
57+
58+
| Dataset | Learning Rate | Epoch | VRMSE |
59+
|---------|---------------|-------|-------|
60+
| [acoustic_scattering_maze](https://huggingface.co/polymathic-ai/CNextU-Net-acoustic_scattering) | 1E-3 | 10 | 0.0196 |
61+
| [active_matter](https://huggingface.co/polymathic-ai/CNextU-Net-active_matter) | 5E-3 | 156 | 0.0953 |
62+
| [convective_envelope_rsg](https://huggingface.co/polymathic-ai/CNextU-Net-convective_envelope_rsg) | 1E-4 | 5 | 0.0663 |
63+
| [gray_scott_reaction_diffusion](https://huggingface.co/polymathic-ai/CNextU-Net-gray_scott_reaction_diffusion) | 1E-4 | 15 | 0.3596 |
64+
| [helmholtz_staircase](https://huggingface.co/polymathic-ai/CNextU-Net-helmholtz_staircase) | 5E-4 | 47 | 0.00146 |
65+
| [MHD_64](https://huggingface.co/polymathic-ai/CNextU-Net-MHD_64) | 5E-3 | 59 | 0.1487 |
66+
| [planetswe](https://huggingface.co/polymathic-ai/CNextU-Net-planetswe) | 1E-2 | 18 | 0.3268 |
67+
| [post_neutron_star_merger](https://huggingface.co/polymathic-ai/CNextU-Net-post_neutron_star_merger) | - | - | - |
68+
| [rayleigh_benard](https://huggingface.co/polymathic-ai/CNextU-Net-rayleigh_benard) | 5E-4 | 12 | 0.4807 |
69+
| [rayleigh_taylor_instability](https://huggingface.co/polymathic-ai/CNextU-Net-rayleigh_taylor_instability) | 5E-3 | 56 | 0.3771 |
70+
| [shear_flow](https://huggingface.co/polymathic-ai/CNextU-Net-shear_flow) | 5E-4 | 9 | 0.3972 |
71+
| [supernova_explosion_64](https://huggingface.co/polymathic-ai/CNextU-Net-supernova_explosion_64) | 5E-4 | 13 | 0.2801 |
72+
| [turbulence_gravity_cooling](https://huggingface.co/polymathic-ai/CNextU-Net-turbulence_gravity_cooling) | 1E-3 | 3 | 0.2093 |
73+
| [turbulent_radiative_layer_2D](https://huggingface.co/polymathic-ai/CNextU-Net-turbulent_radiative_layer_2D) | 5E-3 | 495 | 0.1247 |
74+
| [viscoelastic_instability](https://huggingface.co/polymathic-ai/CNextU-Net-viscoelastic_instability) | 5E-4 | 114 | 0.1966 |
75+
5376
## Loading the model from Hugging Face

0 commit comments

Comments
 (0)