incorrect description of num_sampled parameter in tf.nn.nce_loss() function: num_sampled is number of negative examples per 1 positive example (NOT per batch) #17949
Labels
stat:awaiting tensorflower
Status - Awaiting response from tensorflower
type:docs-bug
Document issues
Hi all,
Looks like parameter
num_sampled
intf.nn.nce_loss
function defines the number of negative examples per a positive example but not per a batch as described in tensorflow documentation (https://www.tensorflow.org/api_docs/python/tf/nn/nce_loss)(see the next code)
As a result,
_compute_sampled_logits
function generatednum_sampled
examples per 1 positive example:The text was updated successfully, but these errors were encountered: