Merge branch 'master' into sia_ts030425

Author: t7phy

.github/actions/prepare_environment/action.yml

@@ -14,12 +14,16 @@ name: Prepare a conda environment for NNPDF installation
 # Remember that these actions are only available after checking out the repository,
 # which should always be the first step!
 # - uses: actions/checkout@v4
+#
+# This action also removes part of the cache of the runner to avoid using too much storage
+# (see at the bottom)
+# It would probably be clever to start using a different runner though TODO
 
 inputs:
   python-version:
     required: true
     description: "Python version to use"
-    default: "3.12"
+    default: "3.13"
 
 runs:
   using: "composite"
@@ -28,12 +32,15 @@ runs:
       with:
         python-version: ${{ inputs.python-version }}
         use-mamba: true
-        channels: https://packages.nnpdf.science/public,conda-forge
+        channels: conda-forge
         show-channel-urls: true
         auto-update-conda: true
         activate-environment: nnpdf_environment
         conda-remove-defaults: true
-    - name: Add NETRC
+    - name: Add NETRC and cleanup cache
       shell: bash -l {0}
       run: |
+        sudo rm -rf /opt/hostedtoolcache/Python/3.9*
+        sudo rm -rf /opt/hostedtoolcache/CodeQL
+        sudo rm -rf /usr/local/lib/android/sdk
         echo "$NETRC_FILE" | base64 --decode > ~/.netrc

.github/workflows/all_tests_nnpdf.yml

@@ -35,7 +35,7 @@ jobs:
       - name: Test n3fit and validphys
         shell: bash -l {0}
         run: |
-          pytest --cov=${PWD} --cov-config=pyproject.toml --mpl --pyargs validphys n3fit --mpl-default-tolerance 18
+          pytest --cov=${PWD} --cov-config=pyproject.toml --mpl --pyargs validphys n3fit --mpl-default-tolerance 24
       - name: Keep coverage file
         if: startsWith(matrix.python-version, '3.13')
         uses: actions/upload-artifact@v4
@@ -100,6 +100,9 @@ jobs:
     - name: Install nnpdf without LHAPDF
       shell: bash -l {0}
       run: |
+        sudo rm -rf /opt/hostedtoolcache/Python/{3.9*}
+        sudo rm -rf /opt/hostedtoolcache/CodeQL
+        sudo rm -rf /usr/local/lib/android/sdk
         pip install .[nolha,torch]
         # Since there is no LHAPDF in the system, initialize the folder and download pdfsets.index
         lhapdf-management update --init

.github/workflows/fitbot.yml

@@ -8,10 +8,12 @@ on:
 
 # some general variables
 env:
-  N3FIT_MAXNREP: 20 # total number of replicas to fit
-  POSTFIT_NREP: 16 # requested replicas for postfit
-  REFERENCE_SET: NNBOT-955eb2bcc-2025-06-17 # reference set for exact results
-  STABLE_REFERENCE_SET: NNBOT-955eb2bcc-2025-06-17 # reference set for last tag
+  N3FIT_MAXNREP: 30 # total number of replicas to fit
+  POSTFIT_NREP: 15 # requested minimum replicas for postfit
+  # IMPORTANT
+  # WHEN CHANGING THE REFERENCE SET, THE NEW REFERENCE MUST BE MANUALLY UPLOADED TO THE SERVER
+  REFERENCE_SET: NNNBOT-1a81255f3-2026-01-27 # reference set for exact results
+  STABLE_REFERENCE_SET: NNBOT-99108504e-2025-11-22 # reference set for last tag
   PYTHONHASHSEED: "0"
 
 jobs:
@@ -55,12 +57,12 @@ jobs:
         cd $RUNFOLDER
         cp developing.yml $RUNCARD.yml
         vp-setupfit $RUNCARD.yml
-    # run n3fit replicas sequentially
+    # try running the n3fit replicas in parallel
     - name: Running n3fit
       shell: bash -l {0}
       run: |
         cd $RUNFOLDER
-        for ((i=1; i<=$N3FIT_MAXNREP; i+=1)); do n3fit $RUNCARD.yml $i ; done
+        n3fit $RUNCARD.yml 1 -r $N3FIT_MAXNREP
     # performing DGLAP
     - name: Running dglap
       shell: bash -l {0}
@@ -79,17 +81,16 @@ jobs:
       run: |
         conda activate nnpdfenv
         cd $RUNFOLDER
-        postfit $POSTFIT_NREP $RUNCARD
-        res=$(vp-upload $RUNCARD 2>&1)
-        echo ${res}
-        while echo ${res} | grep ERROR >/dev/null
-        do
-          sleep 30s
-          res=$(vp-upload $RUNCARD 2>&1)
-        done
-        url=$( echo "${res}" | grep https )
-        echo "FIT_URL=$url" >> $GITHUB_ENV
-    # running validphys report
+        postfit $POSTFIT_NREP $RUNCARD --at-least-nrep
+        ln -s ${PWD}/${RUNCARD} ${CONDA_PREFIX}/share/NNPDF/results
+        tar -czf ${RUNCARD}.tar.gz ${RUNCARD}
+        echo "PATH_TO_SAVE=${PWD}/${RUNCARD}.tar.gz" >> ${GITHUB_ENV}
+    - name: Keep the fit as an artifact
+      if: ${{ !cancelled() }}
+      uses: actions/upload-artifact@v4
+      with:
+        name: ${{ env.RUNCARD }}.tar.gz
+        path: ${{ env.PATH_TO_SAVE }}
     - name: Building and upload report
       shell: bash -l {0}
       run: |
@@ -121,6 +122,6 @@ jobs:
             - Fit Name: ${{ env.RUNCARD }}
             - Fit Report wrt master: ${{ env.REPORT_URL }}
             - Fit Report wrt latest stable reference: ${{ env.REPORT_URL_STABLE }}
-            - Fit Data: ${{ env.FIT_URL }}
+            - Fit Data: fit data is kept as an artifact. Please, remember to upload it to the server if the reference is changed.
 
           Check the report **carefully**, and please  buy me a :coffee: , or better, a GPU :wink:!

.github/workflows/redo_regressions.yml

@@ -34,7 +34,6 @@ jobs:
         echo "$NETRC_FILE" | base64 --decode > ~/.netrc
         conda config --remove channels defaults
         conda config --append channels conda-forge
-        conda config --prepend channels https://packages.nnpdf.science/public
         conda config --set show_channel_urls true
         conda install lhapdf pandoc
     - name: Install nnpdf with testing and qed extras

doc/sphinx/make_theory_csv.py

@@ -5,7 +5,7 @@
 from argparse import ArgumentParser
 from pathlib import Path
 
-from nnpdf_data import theory_cards
+from nnpdf_data import THEORY_CARDS_PATH
 from nnpdf_data.theorydbutils import fetch_all
 
 if __name__ == "__main__":
@@ -20,7 +20,7 @@
 
     args = parser.parse_args()
 
-    theory_df = fetch_all(theory_cards)
+    theory_df = fetch_all(THEORY_CARDS_PATH)
 
     # Enforce the following order in the table:
     order = ["PTO", "QED", "Comments", "IC", "Q0", "ModEv"]

doc/sphinx/source/data/data-config.rst

@@ -23,9 +23,13 @@ located in the ``nnpdf`` git repository at
 
 where a separate ``CommonData`` file is stored for each *Dataset* with the
 filename format described in :ref:`dataset-naming-convention`.
-The data is installed as part of the python package of ``nnpdf``,
+The data is installed as part of the python package of ``nnpdf`` or ``nnpdf_data``.
 all data files to be installed must have a ``.yaml`` extension.
 
+It is possible to add extra sources of data by adding a ``data_path`` variable
+to the ``nnprofile.yaml`` file (see :ref:`nnprofile`).
+The extra sources will always be searched before the default path.
+
 
 Theory lookup table
 ===================