Combine PineAPPL grids and EKOs into FK tables
Project description
pineko
= PineAPPL
+ eko
pineko
converts
- interpolation grids for theory predictions ('grids' for short) in the form of
PineAPPL
grids, together with - Evolution Kernel Operators (EKO) generated by
eko
into fast-kernel (FK) tables. The collection of all FK tables constitute the theory predictions for a PDF fit and therefore is often simply called 'theory'.
pineko
replaces APFELcomb
, which was
used up to NNPDF4.0.
Prerequisites
Generating a 'theory', as defined above, requires several files which are described next.
pineko.toml
You need to provide a pineko.toml
, that provides all necessary paths to the input and output folders.
[DEBUG: Look at the debug example in this repo [1].]
ymldb
You need all files of the ymldb
[2]. [DEBUG: Look at the respective load.sh
script to load from dom.]
This defines the mapping from datasets to FK tables.
Theory Runcards
You need to provide the necessary theory runcards named with their respective theory ID inside the <paths.theory_cards>
folder [3].
Default Operator Card
You need to provide a default operator card for eko
[4].
[DEBUG: Look at the respective load.sh
script to load from dom.]
Grids
pineko
does NOT compute grids, which are instead expected input to pineko
.
There are typically two ways to obtain grids: computing them from scratch with runcards
or reusing existing ones.
Generate new Grids with rr
You need to run rr
with a given theory runcard and put the generated grid file with the same name
inside the <paths.grids>/<theory_id>
folder. The name has to match the ymldb
which is the case by default.
Inherit Grids from Existing Theory
You can reuse the grids from a different theory by running pineko theory inherit-grids SOURCE_THEORY_ID TARGET_THEORY_ID DATASET1 DATASET2 ...
.
The relation between the source theory and the target theory is non-trivial [5].
Running pineko
Running pineko
consists of two steps - each of them being potentially computationally expensive:
computing the EKO and convoluting the EKO with the grid.
Computing the EKO
Generating new EKOs
This is a two step process:
- Generate the necessary operator cards with
pineko theory opcards THEORY_ID DATASET1 DATASET2 ...
- Generate the actual EKOs with
pineko theory ekos THEORY_ID DATASET1 DATASET2 ...
Inherit EKOs from Existing Theory
You can reuse the EKOs from a different theory by running pineko theory inherit-ekos SOURCE_THEORY_ID TARGET_THEORY_ID DATASET1 DATASET2 ...
.
The relation between the source theory and the target theory is non-trivial [6].
Generating the FK Table
You need to have the EKOs computed in the previous step.
Then you can convolute the EKOs with the grids by running pineko theory fks THEORY_ID DATASET1 DATASET2 ...
[1] Actually, instead we should provide a concise description here - but let's wait to be stable first
[2] this is to be replaced by the new CommonData format
[3] this is to be replaced by a binding to the true theory DB
[4] I'm thinking how to improve this, because how could we provide a study on the interpolation accuracy? at the moment there just equal
[5] examples being SV, different evolution settings, etc.
[6] examples being SV, different DIS settings, etc.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.