Instructions to use dataymeric/ArchesWeatherSR with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use dataymeric/ArchesWeatherSR with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("dataymeric/ArchesWeatherSR", 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
Add model
Browse files- checkpoints/archesweathersr.ckpt +3 -0
- config.yaml +118 -0
checkpoints/archesweathersr.ckpt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:656bcae7c2a41b325d3a707146336457ce9061a55ffc1c8505baf3b0c518d0e6
|
| 3 |
+
size 713602481
|
config.yaml
ADDED
|
@@ -0,0 +1,118 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
cluster:
|
| 2 |
+
wandb_mode: offline
|
| 3 |
+
use_custom_requeue: false
|
| 4 |
+
precision: bf16-mixed
|
| 5 |
+
batch_size: 1
|
| 6 |
+
gpus: 4
|
| 7 |
+
num_workers: 8
|
| 8 |
+
blosc_nthreads: 3
|
| 9 |
+
num_nodes: 1
|
| 10 |
+
dataloader:
|
| 11 |
+
dataset:
|
| 12 |
+
_target_: archesweathersr.dataloaders.era5_hdf5.ERA5Downscaling
|
| 13 |
+
lowres_path: era5/240x121/weatherbench2/yearly
|
| 14 |
+
highres_path: era5/1440x721/weatherbench2/yearly
|
| 15 |
+
lead_time_hours: 24
|
| 16 |
+
norm_scheme: era5
|
| 17 |
+
domain: train
|
| 18 |
+
load_prev: false
|
| 19 |
+
multistep: 0
|
| 20 |
+
validation_args:
|
| 21 |
+
domain: val_z0012
|
| 22 |
+
test_args:
|
| 23 |
+
domain: test_z0012
|
| 24 |
+
module:
|
| 25 |
+
metrics:
|
| 26 |
+
era5_ensemble_metrics:
|
| 27 |
+
_target_: geoarches.metrics.ensemble_metrics.Era5EnsembleMetrics
|
| 28 |
+
lead_time_hours: 24
|
| 29 |
+
project: super-resolution
|
| 30 |
+
module:
|
| 31 |
+
_target_: archesweathersr.lightning_modules.sr_flow_matching.DownscalingDiffusionModule
|
| 32 |
+
name: archesweathersr-1
|
| 33 |
+
cond_dim: 256
|
| 34 |
+
scheduler: euler
|
| 35 |
+
prediction_type: sample
|
| 36 |
+
conditional: ''
|
| 37 |
+
interp_args:
|
| 38 |
+
mode: bicubic
|
| 39 |
+
align_corners: false
|
| 40 |
+
state_normalization: residual
|
| 41 |
+
pow: 2.0
|
| 42 |
+
lr: 0.0003
|
| 43 |
+
betas:
|
| 44 |
+
- 0.9
|
| 45 |
+
- 0.98
|
| 46 |
+
weight_decay: 0.01
|
| 47 |
+
num_warmup_steps: 5000
|
| 48 |
+
num_training_steps: 75000
|
| 49 |
+
val:
|
| 50 |
+
sample_every_n_epochs: 5
|
| 51 |
+
num_members: 2
|
| 52 |
+
metrics:
|
| 53 |
+
- _target_: geoarches.metrics.ensemble_metrics.Era5EnsembleMetrics
|
| 54 |
+
lead_time_hours: 24
|
| 55 |
+
metrics_kwargs:
|
| 56 |
+
save_memory: true
|
| 57 |
+
inference:
|
| 58 |
+
num_steps: 25
|
| 59 |
+
num_members: 2
|
| 60 |
+
cf_guidance: 1
|
| 61 |
+
s_churn: 0.0
|
| 62 |
+
save_test_outputs: false
|
| 63 |
+
metrics:
|
| 64 |
+
era5_ensemble_metrics:
|
| 65 |
+
_target_: geoarches.metrics.ensemble_metrics.Era5EnsembleMetrics
|
| 66 |
+
lead_time_hours: 24
|
| 67 |
+
metrics_kwargs:
|
| 68 |
+
save_memory: true
|
| 69 |
+
backbone:
|
| 70 |
+
_target_: archesweathersr.backbones.archesweather.ArchesWeatherCondBackbone
|
| 71 |
+
tensor_size:
|
| 72 |
+
- 8
|
| 73 |
+
- 180
|
| 74 |
+
- 360
|
| 75 |
+
emb_dim: 192
|
| 76 |
+
cond_dim: 256
|
| 77 |
+
window_size:
|
| 78 |
+
- 1
|
| 79 |
+
- 6
|
| 80 |
+
- 12
|
| 81 |
+
droppath_coeff: 0.2
|
| 82 |
+
dropout: 0
|
| 83 |
+
depth_multiplier: 1
|
| 84 |
+
use_skip: true
|
| 85 |
+
first_interaction_layer: linear
|
| 86 |
+
axis_attn: true
|
| 87 |
+
mlp_layer: swiglu
|
| 88 |
+
mlp_ratio: 4.0
|
| 89 |
+
gradient_checkpointing: true
|
| 90 |
+
embedder:
|
| 91 |
+
_target_: archesweathersr.backbones.archesweather.WeatherEncodeDecodeLayer
|
| 92 |
+
img_size:
|
| 93 |
+
- 13
|
| 94 |
+
- 721
|
| 95 |
+
- 1440
|
| 96 |
+
emb_dim: 192
|
| 97 |
+
out_emb_dim: 384
|
| 98 |
+
patch_size:
|
| 99 |
+
- 2
|
| 100 |
+
- 4
|
| 101 |
+
- 4
|
| 102 |
+
surface_ch: 4
|
| 103 |
+
level_ch: 6
|
| 104 |
+
n_concatenated_states: 1
|
| 105 |
+
log: true
|
| 106 |
+
entity: null
|
| 107 |
+
project: super-resolution
|
| 108 |
+
name: archesweathersr-1
|
| 109 |
+
exp_dir: runs/archesweathersr-1
|
| 110 |
+
resume: true
|
| 111 |
+
seed: None
|
| 112 |
+
max_steps: 75000
|
| 113 |
+
batch_size: 1
|
| 114 |
+
save_step_frequency: 2500
|
| 115 |
+
log_freq: 100
|
| 116 |
+
limit_val_batches: 1.0
|
| 117 |
+
accumulate_grad_batches: 1
|
| 118 |
+
mode: train
|