[WIP] LSTM in GenericHybridModel #217
Conversation
…indow = time in KeyedArrays and DimArrays
| # DimensionalData | ||
| mat = Array(Matrix(dfnot)') | ||
| da = DimArray(mat, (Dim{:col}(Symbol.(names(dfnot))), Dim{:row}(1:size(dfnot, 1)))) | ||
| da = to_dimArray(dfnot) |
There was a problem hiding this comment.
please undo the changes to this to avoid conflicts with my other PR. I'm moving this to the documentation.
|
|
||
| docstring = """ | ||
| $(name)(NN, predictors, forcing, targets$(isempty(param_syms) ? "" : ", " * join(string.(param_syms), ", "))) | ||
There was a problem hiding this comment.
your editor is somehow touching other files, without real changes.
There was a problem hiding this comment.
Not sure when it came in - I suspect the formatting script
src/utils/compute_loss.jl
Outdated
| _select_time(ŷ_t::KeyedArray, time_keys) = ŷ_t(time = time_keys) # KeyedArray: () syntax - view & differentiable | ||
| _select_time(ŷ_t::AbstractDimArray, time_keys) = ŷ_t[time = At(time_keys)] # DimArray: [] syntax - copy & differentiable | ||
|
|
||
| function assemble_loss(ŷ, y::Union{KeyedArray{T, 3}, AbstractDimArray{T, 3}}, y_nan, targets, loss_spec) where {T} |
There was a problem hiding this comment.
my original thinking for assemble_loss was to have only one, a generic function, and delegate/multidispatch on use cases to inner functions, such as _get_target_y, etc. We can come back to this later, once things work as expected.
| df = load_timeseries_netcdf("https://github.com/bask0/q10hybrid/raw/master/data/Synthetic4BookChap.nc") | ||
|
|
||
| # Select a subset of data for faster execution | ||
| df = df[1:20000, :] |
There was a problem hiding this comment.
| df = df[1:20000, :] | |
| df = df[1:20000, :] | |
| first(df, 5) |
same reason here. You get the gist 😄 .
Co-authored-by: Lazaro Alonso <lazarus.alon@gmail.com>
loss functions closer to initial
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #217 +/- ##
==========================================
- Coverage 49.27% 47.84% -1.43%
==========================================
Files 33 33
Lines 1861 2019 +158
==========================================
+ Hits 917 966 +49
- Misses 944 1053 +109 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
|
@lazarusA Oh no more checks ;-). |
|
Wanna merge? Or maybe we should wait till tomorrow? |
|
Let's wait until tomorrow - it is still very much untested and I need to understand the code coverage 😂 |
plotting = false