-
Notifications
You must be signed in to change notification settings - Fork 0
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
- Loading branch information
Showing
9 changed files
with
413 additions
and
355 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -1,64 +1,26 @@ | ||
| using DrWatson | ||
| @quickactivate "GBPQAMDecoder" | ||
| @quickactivate :GBPQAMDecoder | ||
|
|
||
| using Chain, | ||
| DataFrames, | ||
| Dates, | ||
| JLD2, | ||
| LinearAlgebra, | ||
| Parquet2, | ||
| ProgressMeter, | ||
| Random, | ||
| StatsBase, | ||
| Transducers | ||
| using DataPrep | ||
| using GBPAlgorithm | ||
|
|
||
| include(srcdir("data_utilities.jl")) | ||
| include(srcdir("decoding.jl")) | ||
|
|
||
| warm_up_results = let | ||
| run_it(; | ||
| power_vals=0:1,with_noise_vals=true:true,pilots_vals=1:1,k_vals=2:2, | ||
| T=10,w=0.25,niter=10,pilots_period=100,step=8,is_simulation=false, | ||
| load_take=200,load_skip=0,should_save=false,savedir="test_$(today())", | ||
| start_msg="Warmming up...", showprogressinfo=false | ||
| ); | ||
|
|
||
| run_it(; | ||
| power_vals=0:1,with_noise_vals=true:true,pilots_vals=1:1,k_vals=2:2, | ||
| T=10,w=0.25,niter=10,pilots_period=100,step=8,is_simulation=false, | ||
| load_take=200,load_skip=0,should_save=true,savedir="test_$(today())", | ||
| start_msg="one more, this time saving...", showprogressinfo=false | ||
| ); | ||
| end | ||
|
|
||
| println("") | ||
|
|
||
| print_help() = println(""" | ||
| Run `run_it(; | ||
| power_vals=-2:8,with_noise_vals=false:true,pilots_vals=1:1,k_vals=2:2, | ||
| T,w,niter,pilots_period,step,is_simulation=false, | ||
| load_take=1_000_000,load_skip=0,should_save=true,savedir="", | ||
| showprogressinfo=true, | ||
| start_msg="Running GBP decoder for all powers" | ||
| )` | ||
| to run the GBP decoding algorithm for all available power levels and noise conditions given: | ||
| - each windowed sequence with length `T` and overlap `step` in the beggining | ||
| - using a mixture with `k` Gaussian components (k in {1,2,3,4,5}) | ||
| - using a gradient descent step of `w` (w in 0..1) | ||
| - on signal generated with `pilots` waves every `pilots_perior` symbols (pilots in {0,1,2,3}) | ||
| - and optionally, if `should_save` saves it to a subdirectory `datadir("results",savedir)` | ||
| - `load_take` and `load_skip` refer to how many symbols to load from the signal file or generate | ||
| for a simulation | ||
| - `is_simulation` to load the model information but not the signal file and generate a signal with | ||
| the specified parameters using the respective memory and noise model | ||
| - `start_msg` can be used to identify the current task in the system log files | ||
| - `showprogressinfo` can be used to toggle the display of parameters and progress | ||
| Thw warm up results are assigned to `warm_up_results`, with a summary for all the datasets and the | ||
| decoding results split per dataset. | ||
| test_tasks = decoding_tasks_list(; | ||
| power=[-1,1],with_noise=true,pilots=0,k=2, | ||
| load_batches=2 | ||
| ) | ||
|
|
||
| Run `print_help()` to show this message again. | ||
| """) | ||
| map(test_tasks) do task_info | ||
| produce_or_load(solve_decoding_task,task_info,datadir("results","testing_script"); force=true) | ||
| end | ||
| end; | ||
|
|
||
| print_help() | ||
| println("") |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,208 @@ | ||
| module GBPQAMDecoder | ||
|
|
||
| using DrWatson | ||
| @quickactivate "GBPQAMDecoder" | ||
|
|
||
| export solve_decoding_task, load_problem_data, decoding_tasks_list, DecodingTask, AlgorithmParams | ||
|
|
||
| using Chain, | ||
| DataFrames, | ||
| Dates, | ||
| LinearAlgebra, | ||
| Parquet2, | ||
| ProgressMeter, | ||
| Random, | ||
| StatsBase, | ||
| Transducers | ||
|
|
||
| using DataPrep # local module | ||
| using GBPAlgorithm # local module | ||
|
|
||
| include(srcdir("decoding_tasks.jl")) | ||
| include(srcdir("signal_simulator.jl")) | ||
| include(srcdir("decoding.jl")) | ||
|
|
||
| # model_data_file(power) = datadir( | ||
| # "qam_data", | ||
| # "model_power_$(power)_wo_noise.parquet" | ||
| # ) | ||
|
|
||
| # signal_data_file(power::Int,with_noise::Bool,pilots::Int) = datadir( | ||
| # "qam_data", | ||
| # "points_power_$(power)_pilots_$(pilots)_$(with_noise ? "w" : "wo")_noise.parquet" | ||
| # ) | ||
|
|
||
| # noise_data_file() = datadir("noise_info.parquet") | ||
|
|
||
| # encoding_data_file() = datadir("constellation_info.parquet") | ||
|
|
||
| # dataset(fn) = Parquet2.Dataset(fn)|>DataFrame | ||
|
|
||
| # Base.@kwdef struct AlgorithmParams | ||
| # T::Int | ||
| # step::Int | ||
| # w::Float64=0.25 | ||
| # niter::Int=10 | ||
| # showprogressinfo::Bool=true | ||
| # end | ||
|
|
||
| # AlgorithmParams(d) = AlgorithmParams(;d...) | ||
|
|
||
| # Base.iterate(a::AlgorithmParams,args...) = iterate(struct2dict(a),args...) | ||
| # DrWatson.allaccess(::AlgorithmParams) = (:T,:step,:w,:niter) | ||
| # DrWatson.default_allowed(::AlgorithmParams) = (Real,) | ||
|
|
||
| # Base.@kwdef struct DecodingTask | ||
| # power::Int | ||
| # with_noise::Bool | ||
| # pilots::Int | ||
| # k::Int | ||
| # pilots_period::Int=100 | ||
| # skip_batches::Int=0 | ||
| # load_batches::Int=1_000_000 | ||
| # is_simulation::Bool=false | ||
| # alg_params::AlgorithmParams=AlgorithmParams(T=10,step=8) | ||
| # pilot_point::Tuple{Int,Int}=(1,1) | ||
| # qam_encoding::Dict{Tuple{Int,Int},Int}=DataPrep.get_qam_encoding(dataset(encoding_data_file())) | ||
| # noise_info::NamedTuple=get_noise_info(dataset(noise_data_file()),power,with_noise) | ||
| # model_file::String=model_data_file(power) | ||
| # signal_file::String=signal_data_file(power,with_noise,pilots) | ||
| # end | ||
|
|
||
| # DecodingTask(d) = DecodingTask(;d...) | ||
|
|
||
| # DrWatson.default_expand(::DecodingTask) = ["alg_params"] | ||
| # DrWatson.default_allowed(::DecodingTask) = (Real,Bool,AlgorithmParams) | ||
| # DrWatson.allaccess(::DecodingTask) = (:power,:with_noise,:pilots,:pilots_period,:k,:skip_batches,:load_batches,:alg_params) | ||
| # DrWatson.default_prefix(t::DecodingTask) = t.is_simulation ? "simulation-decoded" : "points-decoded" | ||
|
|
||
| # function decodingtask2dict(t::DecodingTask) | ||
| # d = struct2dict(t) | ||
| # d[:alg_params] = struct2dict(t.alg_params) | ||
| # return d | ||
| # end | ||
|
|
||
| # function decoding_tasks_list(; alg_params=nothing,kw...) | ||
| # d = Dict(kw) | ||
| # if !isnothing(alg_params) | ||
| # d[:alg_params] = alg_params|>pairs|>Dict|>dict_list.|>AlgorithmParams | ||
| # end | ||
| # return d|>dict_list.|>DecodingTask | ||
| # end | ||
|
|
||
| # function add_pilots_info!(signal_table,task_info) | ||
| # (;pilots,pilots_period) = task_info | ||
| # @chain signal_table begin | ||
| # transform!(eachindex=>:t) | ||
| # transform!(:t=>ByRow(t->mod(t,1:pilots_period)<=pilots)=>:is_pilot) | ||
| # end | ||
| # end | ||
|
|
||
| # function collapse_prior!(factors,sequence,T,step,part_idx) | ||
| # overlap = part_idx > 1 ? pairs(last(factors.Rs,T-step)) : () | ||
| # pilots = (r.i=>r.Ts for r in eachrow(sequence) if r.is_pilot) | ||
| # GBPAlgorithm.collapse_prior!(factors,overlap...,pilots...) | ||
| # return factors | ||
| # end | ||
|
|
||
| # function load_problem_data(task_info) | ||
| # (;k,qam_encoding) = task_info | ||
| # (;noise_var) = task_info.noise_info | ||
| # model_table = @chain task_info.model_file begin | ||
| # dataset | ||
| # DataPrep.parse_model_data(k,noise_var,qam_encoding) | ||
| # end | ||
|
|
||
| # (;load_batches,skip_batches,pilots_period) = task_info | ||
| # signal_table = if !task_info.is_simulation | ||
| # @chain task_info.signal_file begin | ||
| # Parquet2.Dataset | ||
| # Tables.rows | ||
| # _|>Drop(skip_batches*pilots_period)|>Take(load_batches*pilots_period)|>DataFrame | ||
| # DataPrep.parse_signal_data(qam_encoding) | ||
| # add_pilots_info!(task_info) | ||
| # end | ||
| # else | ||
| # @warn "Simulating signal..." | ||
| # symbol_sequence_length=min(load_batches*pilots_period,size(dataset(task_info.signal_file),1)) | ||
| # add_pilots_info!( | ||
| # simulate_signal_table( | ||
| # task_info, | ||
| # symbol_sequence_length, | ||
| # model_table, | ||
| # qam_encoding | ||
| # ), | ||
| # task_info | ||
| # ) | ||
| # end | ||
|
|
||
| # return (;model_table,signal_table) | ||
| # end | ||
|
|
||
| # progress_info_msg(prog,showinfo) = showinfo && next!(prog) | ||
|
|
||
| # function task_info_msg(task_info) | ||
| # (;T,step) = task_info.alg_params | ||
| # (;power,with_noise,pilots,k) = task_info | ||
| # @info "Running algorithm:" "Avg. Power [dBm]"=power "With 4.5dB noise"=with_noise "Pilots"=pilots "Number of mixture components"=k "Sequence length"=T "Step"=step | ||
| # end | ||
|
|
||
| # function save_info_msg(dir) | ||
| # @info "Saved decoded points" "Directory"=dir | ||
| # end | ||
|
|
||
| # function get_noise_info(noise_data,power,with_noise) | ||
| # @chain noise_data begin | ||
| # subset(:pdbm=>ByRow(p->p==power)) | ||
| # map(eachrow(_)) do r | ||
| # noise_sigma = with_noise ? r.sigma*r.scale : 0.0 | ||
| # noise_var = noise_sigma^2 | ||
| # (;noise_sigma,noise_var,noise_scale=r.scale) | ||
| # end | ||
| # only | ||
| # end | ||
| # end | ||
|
|
||
| # getresults(sequence,step,part_idx) = part_idx > 1 ? last(sequence,step) : sequence | ||
|
|
||
| # maxparts(nsyms,partlen,step) = floor(Int,1 + (nsyms - partlen)/step) | ||
| # maxparts(seqdf::DataFrame,partlen,step) = maxparts(size(seqdf,1),partlen,step) | ||
|
|
||
| # function solve_decoding_task(task_info) | ||
| # task_info_msg(task_info) | ||
| # (;model_table,signal_table) = load_problem_data(task_info) | ||
| # (;T,w,niter,step,showprogressinfo) = task_info.alg_params | ||
| # N = length(task_info.qam_encoding) | ||
| # decode_iter! = GBPAlgorithm.GBPDecoder(N,T,w) | ||
| # factors = GBPAlgorithm.Factors(N,T) | ||
| # n = maxparts(size(signal_table,1),T,step) | ||
| # prog = Progress(n; showspeed=true) | ||
| # R = withprogress(eachrow(signal_table); interval=10^-2)|> | ||
| # Partition(T,step)|> | ||
| # Enumerate()|> | ||
| # Map() do (part_idx,part) | ||
| # sequence = DataFrame(part) | ||
| # transform!(sequence,eachindex=>:i) | ||
| # GBPAlgorithm.reset_msg!(decode_iter!) | ||
| # DataPrep.memory_factor!(factors.M,model_table,sequence.Rx) | ||
| # collapse_prior!(factors,sequence,T,step,part_idx) | ||
| # decode_iter!(factors)|>Drop(niter-1)|>Take(1)|>collect|>only | ||
| # GBPAlgorithm.beliefs!(factors,decode_iter!) | ||
| # transform!(sequence,:Ts=>(x->copy(factors.Rs))=>:Rs) | ||
| # transform!(sequence,[:Ts,:Rs]=>((t,r)->t.!=r)=>:error) | ||
| # progress_info_msg(prog,showprogressinfo) | ||
| # getresults(sequence,step,part_idx) | ||
| # end|> | ||
| # Take(n)|> | ||
| # foldxl(vcat) | ||
| # results = select!(R, | ||
| # :Tx=>ByRow(((x,y),)->(;Tx_x=x,Tx_y=y))=>AsTable, | ||
| # :Rx=>ByRow(((x,y),)->(;Rx_x=x,Rx_y=y))=>AsTable, | ||
| # :Ts,:Rs,:error,:is_pilot | ||
| # ) | ||
|
|
||
| # decoding_task = decodingtask2dict(task_info) | ||
| # return @strdict decoding_task results | ||
| # end | ||
|
|
||
| end # end Decoding module |
Oops, something went wrong.