ComfyUI Node

NNT Train Model

A ComfyUI node in NNT Neural Network Toolkit/Models with 27 inputs and 3 outputs.

By inventorado·Created 2 years ago·Updated 2 years ago· 69
NNT Train Model
  • MODEL
  • training_data
  • target_data
  • hyperparameters
  • model
  • training_log
  • metrics
experiment_nametraining_experiment
batch_size32
epochs10
optimizerAdam
learning_rate0.0010
weight_decay0.0001
momentum0.9
loss_functionCrossEntropyLoss
reductionmean
weight_enabledFalse
class_weights[]
margin0.0
use_lr_schedulerTrue
scheduler_typeReduceLROnPlateau
scheduler_step_size3
scheduler_gamma0.10
min_lr0.0000
use_early_stoppingTrue
patience5
min_delta0.0010
reshape_inputFalse
input_reshape_dim[-1]
flatten_inputFalse
CategoryNNT Neural Network Toolkit/Models

Inputs (27)

NameTypeDefaultDescription
MODELMODEL
training_dataTENSOR
target_dataTENSOR
experiment_nameSTRINGtraining_experiment
batch_sizeINT321–512
epochsINT101–1000
optimizerCOMBOAdam12 options: Adadelta, Adagrad, Adam, AdamW, SparseAdam, Adamax, +6
learning_rateFLOAT0.00100.000001–1
weight_decayFLOAT0.00010–0.1
momentumFLOAT0.90–1
loss_functionCOMBOCrossEntropyLoss21 options: L1Loss, MSELoss, CrossEntropyLoss, CTCLoss, NLLLoss, PoissonNLLLoss, +15
reductionCOMBOmean3 options: mean, sum, none
weight_enabledCOMBOFalse2 options: True, False
class_weightsSTRING[]
marginFLOAT0.0-1–1
use_lr_schedulerCOMBOTrue2 options: True, False
scheduler_typeCOMBOReduceLROnPlateau3 options: StepLR, ReduceLROnPlateau, CosineAnnealingLR
scheduler_step_sizeINT31–100
scheduler_gammaFLOAT0.100.01–1
min_lrFLOAT0.00001e-7–0.1
use_early_stoppingCOMBOTrue2 options: True, False
patienceINT51–50
min_deltaFLOAT0.00100.0001–0.1
reshape_inputCOMBOFalse2 options: True, False
input_reshape_dimSTRING[-1]
flatten_inputCOMBOFalse2 options: True, False
hyperparametersoptDICT

Outputs (3)

NameTypeDescription
modelMODEL
training_logSTRING
metricsDICT