clj-djl.training
add-accumulator
(add-accumulator acc key)
Adds an accumulator to the accuracy for the results of the evaluation with the given key.
add-evaluator
(add-evaluator config evaluator)
add-training-listeners
(add-training-listeners config listener)
as-consumer
macro
(as-consumer f)
backward
(backward gc target)
binary-accuracy
(binary-accuracy)(binary-accuracy threshold)(binary-accuracy acc-name threshold)(binary-accuracy acc-name threshold axis)
config
(config {:keys [loss devices data-manager initializer parameter optimizer evaluator listeners]})
fit
(fit trainer nepochs train-iter)(fit trainer nepochs train-iter test-iter)
forward
(forward trainer input)
get-accumulator
(get-accumulator acc key)
Returns the accumulated evaluator value.
get-devices
(get-devices config)
get-evaluators
(get-evaluators trainer)
get-gradient
(get-gradient ndarray)
Returns the gradient NDArray attached to this NDArray.
get-loss
(get-loss trainer)
get-manager
(get-manager trainer)
get-metrics
(get-metrics trainer)
get-model
(get-model trainer)
get-result
(get-result trainer)
gradient-collector
(gradient-collector)(gradient-collector trainer)
initialize
(initialize trainer shapes)(initialize trainer shape & shapes)
iter-seq
(iter-seq iterable)(iter-seq iterable iter)
iterate-dataset
(iterate-dataset trainer ds)
new-default-training-config
(new-default-training-config loss)
new-default-training-listeners
notify-listeners
(notify-listeners trainer callback)
opt-initializer
(opt-initializer config initializer parameter)
opt-optimizer
(opt-optimizer config optimizer)
parameter-store
(parameter-store manager copy)
progress-bar
(progress-bar)
set-metrics
(set-metrics trainer metrics)
set-requires-gradient
(set-requires-gradient ndarray requires-grad)
softmax-cross-entropy-loss
(softmax-cross-entropy-loss)
stop-gradient
(stop-gradient ndarray)
topk-accuracy
(topk-accuracy topk)(topk-accuracy index topk)(topk-accuracy name index topk)
train-batch
(train-batch trainer batch)
trainer
(trainer model config)(trainer {:keys [model loss devices data-manager initializer parameter optimizer listeners]})
training-listeners
(training-listeners)
update-accumulator
(update-accumulator acc key label-list pred-list)
Updates the accuracy with the given key based on a NDList of labels and predictions.
validate-batch
(validate-batch trainer batch)