| from pathlib import Path |
| import sys,os,numpy as np,torch;import torch.distributed as dist |
| from torch.nn.parallel import DistributedDataParallel as DDP |
| R=Path(__file__).resolve().parents[1];sys.path.insert(0,str(R));from model.dlesym import * |
| c=cfg(R);rank=int(os.getenv('RANK',0));world=int(os.getenv('WORLD_SIZE',1));ddp=world>1 |
| if ddp:dist.init_process_group('gloo') |
| d=np.load(R/c['data']['path']);base=DLESyM(**c['model']);m=DDP(base) if ddp else base;opt=torch.optim.AdamW(m.parameters(),lr=c['train']['learning_rate']);ls=[] |
| for i in range(rank,6,world):a,s=map(torch.tensor,(d['atmosphere'][i:i+1],d['sst'][i:i+1]));pa,ps,pr=m(a,s);loss=((pa-torch.tensor(d['target_atmosphere'][i:i+1]))**2).mean()+((ps-torch.tensor(d['target_sst'][i:i+1]))**2).mean()+pr.pow(2).mean()*.01;opt.zero_grad();loss.backward();opt.step();ls.append(float(loss)) |
| v=torch.tensor([sum(ls),len(ls)],dtype=torch.float64) |
| if ddp:dist.all_reduce(v) |
| p=R/c['paths']['checkpoint'] |
| if rank==0:p.parent.mkdir(parents=True,exist_ok=True);torch.save({'model':base.state_dict(),'model_config':c['model']},p);write(R/c['paths']['training_metrics'],{'loss':float(v[0]/v[1]),'world_size':world});print(p) |
| if ddp:dist.destroy_process_group() |
|
|
