search

amazon-science / earth-forecasting-transformer

Official implementation of Earthformer
Apache License 2.0
349 stars 58 forks source link

Questions about the target_shape of the enso dataset #64

Open fizzking opened 6 months ago

fizzking commented 6 months ago

The label of enso is Nino3.4 SST anomaly index, and the data dimension is (year,month). But why targe shape in cfg is a four-dimensional array of (12,24, 48,1), shouldn't it be two-dimensional b0a86f241b4dd828b4656e7e811b437

gaozhihan commented 6 months ago

Thank you for your question. Earthformer does not directly predict the Niño3.4 index, but rather forecasts future sea surface temperature (SST) anomalies across a certain region used to calculate the index, which has two spatial dimensions.

fizzking commented 6 months ago

Ok, I understand. But why are in_len and out_len respectively 12,14 in fig configuration, but 12 and 26 in data preprocessing? e2f4a13c2b3a7ef1ad15bab4fad6ab7 98994437c144c6424cd958f63597b48

gaozhihan commented 6 months ago

The dataset we used has a maximum support of out_len=26 (two years). For the experiments in our paper, we only explored a setting with out_len=14 (one year).

fizzking commented 6 months ago

Why I use the command: MASTER_ADDR=localhost MASTER_PORT=10001 python ./scripts/cuboid_transformer/enso/train_cuboid_enso.py --gpus 1 --cfg ./scripts/cuboid_transformer/enso/earthformer_enso_v1.yaml --ckpt_name last.ckpt --save tmp_enso Trying to run the training enso dataset results in code that automatically kills? dad6331851e289168f5333c2e6a919f

gaozhihan commented 6 months ago

This error does not appear to be related to our code. You may want to run sudo dmesg to view system log messages and help identify what happened.