Skip to content

GitLab

Commit d81fbd34 authored by Boris Bonev's avatar Boris Bonev Committed by Boris Bonev
Browse files

changing default normalization mode in DISCO

parent 96a2b546
Show whitespace changes
Inline Side-by-side
Showing with 60 additions and 78 deletions
examples/train_sfno.py
View file @ d81fbd34
......@@ -430,21 +430,20 @@ def main(train=True, load_checkpoint=False, enable_amp=False, log_grads=0):
normalization_layer="none",
)
# models[f"lsno_sc2_layers4_e32"] = partial(
# LSNO,
# spectral_transform="sht",
# img_size=(nlat, nlon),
# grid=grid,
# num_layers=4,
# scale_factor=2,
# embed_dim=32,
# operator_type="driscoll-healy",
# activation_function="gelu",
# big_skip=True,
# pos_embed=False,
# use_mlp=True,
# normalization_layer="none",
# )
models[f"lsno_sc2_layers4_e32"] = partial(
LSNO,
img_size=(nlat, nlon),
grid=grid,
num_layers=4,
scale_factor=2,
embed_dim=32,
operator_type="driscoll-healy",
activation_function="gelu",
big_skip=False,
pos_embed=False,
use_mlp=True,
normalization_layer="none",
)
# iterate over models and train each model
root_path = os.path.dirname(__file__)
......@@ -487,7 +486,7 @@ def main(train=True, load_checkpoint=False, enable_amp=False, log_grads=0):
scheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(optimizer, 'min')
gscaler = amp.GradScaler(enabled=enable_amp)
dataloader.dataset.nsteps = 2 * dt//dt_solver
train_model(model, dataloader, optimizer, gscaler, scheduler, nepochs=20, loss_fn="l2", nfuture=nfuture, enable_amp=enable_amp, log_grads=log_grads)
train_model(model, dataloader, optimizer, gscaler, scheduler, nepochs=10, loss_fn="l2", nfuture=nfuture, enable_amp=enable_amp, log_grads=log_grads)
dataloader.dataset.nsteps = 1 * dt//dt_solver
training_time = time.time() - start_time
......
notebooks/resample_sphere.ipynb
View file @ d81fbd34
This source diff could not be displayed because it is too large. You can view the blob instead.
notebooks/train_sfno.ipynb
View file @ d81fbd34
......@@ -176,7 +176,7 @@
"# activation_function = nn.ReLU,\n",
"# bias = False):\n",
"# super().__init__()\n",
" \n",
"\n",
"# current_dim = input_dim\n",
"# layers = []\n",
"# for l in range(num_layers-1):\n",
......@@ -221,7 +221,7 @@
" loss = solver.integrate_grid((prd - tar)**2, dimensionless=True).sum(dim=-1)\n",
" if relative:\n",
" loss = loss / solver.integrate_grid(tar**2, dimensionless=True).sum(dim=-1)\n",
" \n",
"\n",
" if not squared:\n",
" loss = torch.sqrt(loss)\n",
" loss = loss.mean()\n",
......@@ -515,7 +515,7 @@
],
"metadata": {
"kernelspec": {
"display_name": "Python 3 (ipykernel)",
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
......@@ -531,12 +531,7 @@
"pygments_lexer": "ipython3",
"version": "3.10.12"
},
"orig_nbformat": 4,
"vscode": {
"interpreter": {
"hash": "31f2aee4e71d21fbe5cf8b01ff0e069b9275f58929596ceb00d14d90e3e16cd6"
}
}
"orig_nbformat": 4
},
"nbformat": 4,
"nbformat_minor": 2
......
torch_harmonics/convolution.py
View file @ d81fbd34
......@@ -57,7 +57,7 @@ except ImportError as err:
def _normalize_convolution_tensor_s2(
psi_idx, psi_vals, in_shape, out_shape, kernel_size, quad_weights, transpose_normalization=False, basis_norm_mode="none", merge_quadrature=False, eps=1e-9
psi_idx, psi_vals, in_shape, out_shape, kernel_size, quad_weights, transpose_normalization=False, basis_norm_mode="mean", merge_quadrature=False, eps=1e-9
):
"""
Discretely normalizes the convolution tensor and pre-applies quadrature weights. Supports the following three normalization modes:
......@@ -135,7 +135,7 @@ def _precompute_convolution_tensor_s2(
grid_out="equiangular",
theta_cutoff=0.01 * math.pi,
transpose_normalization=False,
basis_norm_mode="none",
basis_norm_mode="mean",
merge_quadrature=False,
):
"""
......@@ -297,7 +297,7 @@ class DiscreteContinuousConvS2(DiscreteContinuousConv):
out_shape: Tuple[int],
kernel_shape: Union[int, List[int]],
basis_type: Optional[str] = "piecewise linear",
basis_norm_mode: Optional[str] = "none",
basis_norm_mode: Optional[str] = "mean",
groups: Optional[int] = 1,
grid_in: Optional[str] = "equiangular",
grid_out: Optional[str] = "equiangular",
......@@ -403,7 +403,7 @@ class DiscreteContinuousConvTransposeS2(DiscreteContinuousConv):
out_shape: Tuple[int],
kernel_shape: Union[int, List[int]],
basis_type: Optional[str] = "piecewise linear",
basis_norm_mode: Optional[str] = "none",
basis_norm_mode: Optional[str] = "mean",
groups: Optional[int] = 1,
grid_in: Optional[str] = "equiangular",
grid_out: Optional[str] = "equiangular",
......
torch_harmonics/distributed/distributed_convolution.py
View file @ d81fbd34
......@@ -76,7 +76,7 @@ def _precompute_distributed_convolution_tensor_s2(
grid_out="equiangular",
theta_cutoff=0.01 * math.pi,
transpose_normalization=False,
basis_norm_mode="none",
basis_norm_mode="mean",
merge_quadrature=False,
):
"""
......@@ -208,7 +208,7 @@ class DistributedDiscreteContinuousConvS2(DiscreteContinuousConv):
out_shape: Tuple[int],
kernel_shape: Union[int, List[int]],
basis_type: Optional[str] = "piecewise linear",
basis_norm_mode: Optional[str] = "none",
basis_norm_mode: Optional[str] = "mean",
groups: Optional[int] = 1,
grid_in: Optional[str] = "equiangular",
grid_out: Optional[str] = "equiangular",
......@@ -348,7 +348,7 @@ class DistributedDiscreteContinuousConvTransposeS2(DiscreteContinuousConv):
out_shape: Tuple[int],
kernel_shape: Union[int, List[int]],
basis_type: Optional[str] = "piecewise linear",
basis_norm_mode: Optional[str] = "none",
basis_norm_mode: Optional[str] = "mean",
groups: Optional[int] = 1,
grid_in: Optional[str] = "equiangular",
grid_out: Optional[str] = "equiangular",
......
torch_harmonics/examples/models/lsno.py
View file @ d81fbd34
......@@ -35,6 +35,7 @@ import torch.amp as amp
from torch_harmonics import RealSHT, InverseRealSHT
from torch_harmonics import DiscreteContinuousConvS2, DiscreteContinuousConvTransposeS2
from torch_harmonics import ResampleS2
from ._layers import *
......@@ -44,7 +45,7 @@ from functools import partial
class DiscreteContinuousEncoder(nn.Module):
def __init__(
self,
inp_shape=(721, 1440),
in_shape=(721, 1440),
out_shape=(480, 960),
grid_in="equiangular",
grid_out="equiangular",
......@@ -61,7 +62,7 @@ class DiscreteContinuousEncoder(nn.Module):
self.conv = DiscreteContinuousConvS2(
inp_chans,
out_chans,
in_shape=inp_shape,
in_shape=in_shape,
out_shape=out_shape,
kernel_shape=kernel_shape,
basis_type=basis_type,
......@@ -69,7 +70,7 @@ class DiscreteContinuousEncoder(nn.Module):
grid_out=grid_out,
groups=groups,
bias=bias,
theta_cutoff=math.sqrt(2) * torch.pi / float(out_shape[0] - 1),
theta_cutoff=math.sqrt(2.0) * torch.pi / float(out_shape[0] - 1),
)
def forward(self, x):
......@@ -86,7 +87,7 @@ class DiscreteContinuousEncoder(nn.Module):
class DiscreteContinuousDecoder(nn.Module):
def __init__(
self,
inp_shape=(480, 960),
in_shape=(480, 960),
out_shape=(721, 1440),
grid_in="equiangular",
grid_out="equiangular",
......@@ -99,12 +100,13 @@ class DiscreteContinuousDecoder(nn.Module):
):
super().__init__()
# set up
self.sht = RealSHT(*inp_shape, grid=grid_in).float()
# # set up
self.sht = RealSHT(*in_shape, grid=grid_in).float()
self.isht = InverseRealSHT(*out_shape, lmax=self.sht.lmax, mmax=self.sht.mmax, grid=grid_out).float()
# self.upscale = ResampleS2(*in_shape, *out_shape, grid_in=grid_in, grid_out=grid_out)
# set up DISCO convolution
self.convt = DiscreteContinuousConvTransposeS2(
self.conv = DiscreteContinuousConvS2(
inp_chans,
out_chans,
in_shape=out_shape,
......@@ -115,21 +117,22 @@ class DiscreteContinuousDecoder(nn.Module):
grid_out=grid_out,
groups=groups,
bias=False,
theta_cutoff=math.sqrt(2) * torch.pi / float(inp_shape[0] - 1),
theta_cutoff=math.sqrt(2.0) * torch.pi / float(in_shape[0] - 1),
)
# self.convt = nn.Conv2d(inp_chans, out_chans, 1, bias=False)
def _upscale_sht(self, x: torch.Tensor):
def upscale_sht(self, x: torch.Tensor):
return self.isht(self.sht(x))
def forward(self, x):
dtype = x.dtype
# x = self.upscale(x)
with amp.autocast(device_type="cuda", enabled=False):
x = x.float()
x = self._upscale_sht(x)
x = self.convt(x)
x = self.upscale_sht(x)
x = self.conv(x)
x = x.to(dtype=dtype)
return x
......@@ -182,7 +185,7 @@ class SphericalNeuralOperatorBlock(nn.Module):
grid_in=forward_transform.grid,
grid_out=inverse_transform.grid,
bias=False,
theta_cutoff=4*math.sqrt(2) * torch.pi / float(inverse_transform.nlat - 1),
theta_cutoff=4 * math.sqrt(2.0) * torch.pi / float(inverse_transform.nlat - 1),
)
elif conv_type == "global":
self.global_conv = SpectralConvS2(forward_transform, inverse_transform, input_dim, output_dim, gain=gain_factor, operator_type=operator_type, bias=False)
......@@ -272,8 +275,6 @@ class LocalSphericalNeuralOperatorNet(nn.Module):
Parameters
-----------
spectral_transform : str, optional
Type of spectral transformation to use, by default "sht"
operator_type : str, optional
Type of operator to use ('driscoll-healy', 'diagonal'), by default "driscoll-healy"
img_shape : tuple, optional
......@@ -339,7 +340,6 @@ class LocalSphericalNeuralOperatorNet(nn.Module):
def __init__(
self,
spectral_transform="sht",
operator_type="driscoll-healy",
img_size=(128, 256),
grid="equiangular",
......@@ -365,7 +365,6 @@ class LocalSphericalNeuralOperatorNet(nn.Module):
):
super().__init__()
self.spectral_transform = spectral_transform
self.operator_type = operator_type
self.img_size = img_size
self.grid = grid
......@@ -440,8 +439,7 @@ class LocalSphericalNeuralOperatorNet(nn.Module):
theta_cutoff=math.sqrt(2) * torch.pi / float(self.h - 1),
)
# prepare the spectral transform
if self.spectral_transform == "sht":
# prepare the SHT
modes_lat = int(self.h * self.hard_thresholding_fraction)
modes_lon = int(self.w // 2 * self.hard_thresholding_fraction)
modes_lat = modes_lon = min(modes_lat, modes_lon)
......@@ -449,9 +447,6 @@ class LocalSphericalNeuralOperatorNet(nn.Module):
self.trans = RealSHT(self.h, self.w, lmax=modes_lat, mmax=modes_lon, grid=grid_internal).float()
self.itrans = InverseRealSHT(self.h, self.w, lmax=modes_lat, mmax=modes_lon, grid=grid_internal).float()
else:
raise (ValueError("Unknown spectral transform"))
self.blocks = nn.ModuleList([])
for i in range(self.num_layers):
first_layer = i == 0
......@@ -490,7 +485,7 @@ class LocalSphericalNeuralOperatorNet(nn.Module):
# decoder
self.decoder = DiscreteContinuousDecoder(
inp_shape=(self.h, self.w),
in_shape=(self.h, self.w),
out_shape=self.img_size,
grid_in=grid_internal,
grid_out=grid,
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment