Issues
- 1
What is meant by global conditioning only.
#25 opened by btickell - 2
Reconsider the softmax distribution
#24 opened by cheind - 2
Decouple input and output representations
#22 opened by cheind - 0
config.yaml files contain a `fit:` root element
#23 opened by cheind - 0
- 1
document sampling, quantization and encoding
#14 opened by cheind - 4
- 1
plot weight histograms
#19 opened by cheind - 3
add image support
#20 opened by cheind - 3
add positional encoding
#18 opened by cheind - 2
Add conditioning
#1 opened by cheind - 1
- 2
compare nstep unrolling and onestep unrolling
#17 opened by cheind - 2
add gumbel-noise based sampler
#13 opened by cheind - 0
use DTW in accuracy / train?
#15 opened by cheind - 3
- 4
Switch to quantizated encoding and one hot
#9 opened by cheind - 4
predictability of time series
#7 opened by cheind - 1
add conditioning
#3 opened by cheind - 1
come up with a better forecast metric
#2 opened by cheind - 0
quantizer theory
#10 opened by cheind - 0
normalization + mid-tread quantization may lead to zero being mapped to non-zero when reconstructed
#11 opened by cheind - 1
use ray.tune for hyperparameter search
#6 opened by cheind - 1
try squeeze and excite layers
#5 opened by cheind - 0
consider COT objective for quantized version
#4 opened by cheind