鏂囩珷鏉ユ簮|鎭掕繙浜戠ぞ鍖哄師鍦板潃|銆愮偧閲戝ザ濡堛€戝師浣滆€厊闃垮窞鏃垛€嬧€嬮棿锛?022骞?鏈?鏃ュ績鎯咃細婵掍复宕╂簝鍘熷洜锛氬眳瀹堕殧绂讳竴涓鏈堚€︹€︿篃涓嶇煡閬撲粈涔堟椂鍊欒В灏佲€︹€︾畻浜嗭紝鎴戞憡鐗屼簡锛屾垜鏄紑濮嬬帺鍧忎簡锛佸洜涓哄績鎯呬笉濂金煉紝宸ヤ綔寰堟秷鏋侊紒鎸戜竴涓煭灏忕簿鎮嶇殑甯栧瓙鏉ュ垎浜紝涓嶇浣犲枩涓嶅枩娆煓堬紝灏辨槸杩欐牱馃暥锔忔潵锛屽睍绀轰竴涓嬸焵勶笍锛氬噯澶囧伐浣滐細importnumpyasnpimporttorchfromtorch.utils.dataimportWeightedRandomSamplerfromtorch.utils.dataimportDataLoaderfromtorch.utils.dataimportTensorDataset鐢熸垚鏁版嵁#鍋囪鏄笁鍒嗙被闂锛屾瘡涓被鐨勬牱鏈暟涓?0,1000,3000class_counts=np.array([10,1000,3000])#鎬绘暟鏍锋湰n_samples=class_counts.sum()#4010#labelslabels=[]foriinrange(len(class_counts)):labels.extend([i]*class_counts[i])Y=torch.from_numpy(np.array(labels,dtype=np.int64))#闅忔満鐢熸垚涓€浜涙暟鎹紝涓嶉噸瑕乆=torch.randn(n_samples)generateweights#缁欐瘡涓被涓€涓潈閲峜lass_weights=[n_samples/class_counts[i]foriinrange(len(class_counts))]#[401.0,4.01,1.3367]#涓烘瘡涓牱鏈敓鎴愭潈閲峸eights=[class_weights[i]foriinlabels]datapackagetrain_dataset=TensorDataset(X,Y)sampler=WeightedRandomSampler(weights,int(n_samples),replacement=True)瀹為獙A锛氫娇鐢ㄦ浛鎹㈢殑鍔犳潈鍒嗗竷锛堟牱鏈彲浠ラ噸澶嶄娇鐢級train_loader=DataLoader(train_dataset,batch_size=1024,sampler=sampler,drop_last=True)fori,(x,y)inenumerate(train_loader):print(f"batchindex{i},n_0:{(y==0).sum()},n_1:{(y==1).sum()},n_2:{(y==3).sum()}")#杈撳嚭锛?绗竴鎵癸紝姣忎釜绫诲埆鐨勪釜鏁颁负349锛?44,331#绗簩鎵癸紝姣忕被涓暟344,360,320#绗笁鎵癸紝姣忕被涓暟339,348,337瀹為獙B锛氬姞鏉冨垎甯冧笉浣跨敤replacement锛坰ampleNotreusable)sampler=WeightedRandomSampler(weights,int(num_samples),replacement=False)train_loader=DataLoader(train_dataset,batch_size=1024,sampler=sampler,drop_last=True)fori,(x,y)inenumerate(train_loader):鎵撳嵃(f"鎵归噺绱㈠紩{i},n_0:{(y==0).sum()},n_1:{(y==1).sum()},n_2:{(y==3).sum()}")#output:#绗竴鎵癸紝姣忎釜绫诲埆鐨勪釜鏁颁负10,466,548#绗簩涓壒娆★紝姣忎釜绫诲埆鐨勪釜鏁颁负0,333,691#绗笁涓壒娆★紝姣忎釜绫诲埆鐨勪釜鏁颁负0,173,851瀹為獙C锛氱畝鍗曢殢鏈哄垎閰峵rain_loader=DataLoader(train_dataset,batch_size=20,shuffle=True,drop_last=True)fori,(x,y)inenumerate(train_loader):print(f"鎵圭储寮晎i},n_0:{(y==0.sum()},n_1:{(y==1).sum()},n_2:{(y==3).sum()}")#output:#绗竴鎵癸紝姣忎竴绫荤殑涓暟涓?,227,797#绗簩鎵癸紝姣忎竴绫荤殑涓暟涓?,271,752#绗笁鎵逛腑锛屾瘡涓被鍒殑涓暟鍒嗗埆涓?銆?57銆?61銆?/p>
