NNUE with or without PruneAtShallowDepth?

[ChessOverflow]

No, it's an UCI option. Please look at the nodchip repository.
It's described under Generating Training Data and Training a Completely New Network.

Originally posted by @joergoster in official-stockfish#4690 (comment)

Switching the PruneAtShallowDepth off means exaxtly to disable the "Pruning at shallow depth" in search.cpp

  // Step 13. Pruning at shallow depth (~200 Elo)
      if (  !rootNode
          && (PvNode ? prune_at_shallow_depth : true)
          && pos.non_pawn_material(us)
          && bestValue > VALUE_TB_LOSS_IN_MAX_PLY)
      {
          // Skip quiet moves if movecount exceeds our FutilityMoveCount threshold
          moveCountPruning = moveCount >= futility_move_count(improving, depth);

          // Reduced depth of the next LMR search
          int lmrDepth = std::max(newDepth - reduction(improving, depth, moveCount), 0);

          if (   !captureOrPromotion
              && !givesCheck)
          {
              // Countermoves based pruning (~20 Elo)
              if (   lmrDepth < 4 + ((ss-1)->statScore > 0 || (ss-1)->moveCount == 1)
                  && (*contHist[0])[movedPiece][to_sq(move)] < CounterMovePruneThreshold
                  && (*contHist[1])[movedPiece][to_sq(move)] < CounterMovePruneThreshold)
                  continue;

              // Futility pruning: parent node (~5 Elo)
              if (   lmrDepth < 7
                  && !ss->inCheck
                  && ss->staticEval + 266 + 170 * lmrDepth <= alpha
                  &&  (*contHist[0])[movedPiece][to_sq(move)]
                    + (*contHist[1])[movedPiece][to_sq(move)]
                    + (*contHist[3])[movedPiece][to_sq(move)]
                    + (*contHist[5])[movedPiece][to_sq(move)] / 2 < 27376)
                  continue;

              // Prune moves with negative SEE (~20 Elo)
              if (!pos.see_ge(move, Value(-(30 - std::min(lmrDepth, 18)) * lmrDepth * lmrDepth)))
                  continue;
          }
          else
          {
              // Capture history based pruning when the move doesn't give check
              if (   !givesCheck
                  && lmrDepth < 1
                  && captureHistory[movedPiece][to_sq(move)][type_of(pos.piece_on(to_sq(move)))] < 0)
                  continue;

              // SEE based pruning
              if (!pos.see_ge(move, Value(-213) * depth)) // (~25 Elo)
                  continue;
          }
      }

[ChessOverflow]

[joergoster]

@ChessOverflow Who are you? Are you ginkgo20?
I was only trying to help with your training using the old nodchip trainer. Now it turns out your real intention is something different.
Please stop this! And stop opening discussions or issues on my repository! And stop pinging all those people yelling for help!

[ChessOverflow]

There is another feature: EnableTranspositionTable

Switching the EnableTranspositionTable off mean to disable the "TTEntry::save" in tt.cpp.

if (!TranspositionTable::enable_transposition_table)

void TTEntry::save(Key k, Value v, bool pv, Bound b, Depth d, Move m, Value ev) {

  if (!TranspositionTable::enable_transposition_table) {
      return;
  }
  // Preserve any existing move for the same position
  if (m || (uint16_t)k != key16)
      move16 = (uint16_t)m;

  // Overwrite less valuable entries (cheapest checks first)
  if (b == BOUND_EXACT
      || (uint16_t)k != key16
      || d - DEPTH_OFFSET > depth8 - 4)
  {
      assert(d > DEPTH_OFFSET);
      assert(d < 256 + DEPTH_OFFSET);

      key16     = (uint16_t)k;
      depth8    = (uint8_t)(d - DEPTH_OFFSET);
      genBound8 = (uint8_t)(TT.generation8 | uint8_t(pv) << 2 | b);
      value16   = (int16_t)v;
      eval16    = (int16_t)ev;
  }
}

@joergoster Can you help me for training a simple network?

[joergoster]

@ChessOverflow Sorry, no.