1.12 (#557)

* add numpy * fix rotation on new game * fix wide root zero * edit INSTALL.md * force hints on only if starting pondering * readme * edit README.md * Check that the platform is Windows, and the user32 dll has the SetProcessDpiAwarenessContext function (should exist for all versions >= Vista). Prevents blurry Kivy windows on high dpi displays. Reference: https://stackoverflow.com/questions/71704354/kivy-app-is-blurry-on-windows-with-high-resolution-screen * Prevents blurry Kivy windows on high dpi displays. Check that the platform is Windows, and the user32 dll has the SetProcessDpiAwarenessContext function (should exist for all versions >= Vista). Reference: https://stackoverflow.com/questions/71704354/kivy-app-is-blurry-on-windows-with-high-resolution-screen * Use kivy.platform instead of the platform package. * Fix bug when the board size is not square. * Set log level back to "warning" (was changed by mistake in a previous commit). * change pyinstaller icon file paths from fully qualified to relative, pointing to the ico file in the repository * add VSVersionInfo metadata to Windows executable file from PyInstaller * replace importLib modules with path modification and regular imports * Initial code for analysis in move range (not working yet). * Initial working code for move range analysis. Work in progress, i18n not done. * Initial working code for move range analysis. Work in progress, i18n not done. * Get rid of unnecessary code for getting dialog subwidgets. * Implemented greyed-out inputs when the move range checkbox is not selected. * Fix spacing which was changed by accident. * edit katrain/popups.kv, edit katrain/gui/popups.py * Fix invocation of game analysis. * Added i18n for the move range analysis. i18n done for Russian. * TODO removed for Russian. * Translated remaining Russian strings which were TODO:. * French strings for move range analysis. * Change the way analyze_extra is called when move range is specified. * scale=0 turns off * allow 3.10 * edit .github/workflows/test.yaml * Blended territory display (#556) * Check that the platform is Windows, and the user32 dll has the SetProcessDpiAwarenessContext function (should exist for all versions >= Vista). Prevents blurry Kivy windows on high dpi displays. Reference: https://stackoverflow.com/questions/71704354/kivy-app-is-blurry-on-windows-with-high-resolution-screen * Prevents blurry Kivy windows on high dpi displays. Check that the platform is Windows, and the user32 dll has the SetProcessDpiAwarenessContext function (should exist for all versions >= Vista). Reference: https://stackoverflow.com/questions/71704354/kivy-app-is-blurry-on-windows-with-high-resolution-screen * Changed territory display to be smoothly blended, for player expected territories and for loss in teaching games. Added an ownership mark to stones while displaying territory; the color of the mark indicates expected ownership, and the size of the mark is proportional to certainty. * Fixed bug: blended territory display does not work with rotation. * Revert "Prevents blurry Kivy windows on high dpi displays." This reverts commit 1cff741e3f7b6bcd8fe7d25cdc9a3fddbcaea242. * Revert "Check that the platform is Windows, and the user32 dll has the SetProcessDpiAwarenessContext function (should exist for all versions >= Vista)." This reverts commit c9ae4f6432efe566bed11eb4a9575b19df0a0db2. * Changed territory display to be smoothly blended, for player expected territories and for loss in teaching games. Added an ownership mark to stones while displaying territory; the color of the mark indicates expected ownership, and the size of the mark is proportional to certainty. Rebased blended territory display on the 1.12 branch. * Fixed bug: blended territory display does not work with rotation. Fixed for non-square board and rebased. * Revert "Prevents blurry Kivy windows on high dpi displays." This reverts commit 1cff741e3f7b6bcd8fe7d25cdc9a3fddbcaea242. * Revert "Check that the platform is Windows, and the user32 dll has the SetProcessDpiAwarenessContext function (should exist for all versions >= Vista)." This reverts commit c9ae4f6432efe566bed11eb4a9575b19df0a0db2. * Make blended territory work with non-square boards and their rotations. * Whitespace. * Make blended territory work with non-square boards and their rotations. * Change marks on stones from circles to squares. * Add Theme settings enabling different types of territory and stone marks displays. * Add Theme settings enabling different types of territory and stone marks displays. * Reformatted with black -l 120. * Documentation for territory display styles and themes. * Added acknowledgement for game used in screenshots. * spacing * Some cleanup of Theme variables. * Switch default mode back to "blended". * Added screenshot of blended style - weak stone marks. * Fix bold text. Co-authored-by: Jacob Minsky <jacob.minsky@gmail.com> * fix test yaml * edit spec/file_version.py, edit katrain/core/game.py and 3 other changes * edit katrain/core/constants.py * Fixes and enhancements for the blended territory feature. (#564) * Fix newly placed stone getting transparency before ownership is updated. * Set black and white territory colors separately. * Try dimming board when territory display is active. * Change tint of the board when in territory estimate mode and style is "blended". * Fix bug when loading SGF with initial position setup. * Revert "Change tint of the board when in territory estimate mode and style is "blended"." This reverts commit d5b46c8966d02b8660a2fe88e4c8d882af43f51f. * Marks on stone should be stone colors, not ownership colors - which look bad. * Version with new board texture (wood6.jpg from https://github.com/waltheri/wgo.js/tree/master/textures) and changed territory color parameters. * Added gamma-correction to territory coloring. * Add acknowledgement for board texture. Co-authored-by: rzcp66 <jacob.minsky@gm.com> * Remove numpy (#570) * Fix newly placed stone getting transparency before ownership is updated. * Set black and white territory colors separately. * Try dimming board when territory display is active. * Change tint of the board when in territory estimate mode and style is "blended". * Fix bug when loading SGF with initial position setup. * Revert "Change tint of the board when in territory estimate mode and style is "blended"." This reverts commit d5b46c8966d02b8660a2fe88e4c8d882af43f51f. * Marks on stone should be stone colors, not ownership colors - which look bad. * Version with new board texture (wood6.jpg from https://github.com/waltheri/wgo.js/tree/master/textures) and changed territory color parameters. * Added gamma-correction to territory coloring. * Eliminates numpy; working version with rotation without numpy. * Fixed stuff related to hover content and board rotation. * Bug fix for roi selection without numpy. * Added comment for rot90 implementation with lists. * Use reversed() instead of [::-1] for clarity. Co-authored-by: rzcp66 <jacob.minsky@gm.com> * Fix region of interest display when board is rotated. (#572) * themes Co-authored-by: Sander Land <sander@chatdesk.com> Co-authored-by: Sander Land <sander.land@cognite.com> Co-authored-by: Jacob Minsky <jacob.minsky@gmail.com> Co-authored-by: ulty4life <ulty4life@gmail.com> Co-authored-by: Jacob Minsky <35696962+jacobm-tech@users.noreply.github.com> Co-authored-by: rzcp66 <jacob.minsky@gm.com>
2022-11-20 11:01:13 +01:00
1 parent 624b3485c3
commit 695f852f2f
@@ -244,7 +244,7 @@ def request_ai_analysis(game: Game, cn: GameNode, extra_settings: Dict) -> Optio
        extra_settings=extra_settings,
    )
    while not (error or analysis):
-        time.sleep(0.01) # TODO: prevent deadlock if esc, check node in queries?
+        time.sleep(0.01)  # TODO: prevent deadlock if esc, check node in queries?
        engine.check_alive(exception_if_dead=True)
    return analysis

@@ -1,5 +1,5 @@
 PROGRAM_NAME = "KaTrain"
-VERSION = "1.11.2"
+VERSION = "1.12.0"
 HOMEPAGE = "https://github.com/sanderland/katrain"
 CONFIG_MIN_VERSION = "1.11.0"  # keep config files from this version
 ANALYSIS_FORMAT_VERSION = "1.0"
@@ -344,15 +344,11 @@ class KataGoEngine(BaseEngine):
                        if pq.get(k) != query.get(k)
                    }
                    if differences:
-                        # TODO:remove
-                        self.katrain.log(f"Found differences in ponder check: {differences}", OUTPUT_EXTRA_DEBUG)
                        self.stop_pondering()
-                        query["maxVisits"] = 1_000_000
+                        query["maxVisits"] = 10_000_000
                        query["reportDuringSearchEvery"] = PONDERING_REPORT_DT
                        self.ponder_query = query
                    else:
-                        # TODO:remove
-                        self.katrain.log("Found no differences in ponder check, skipping", OUTPUT_EXTRA_DEBUG)
                        continue

                terminate = query.get("action") == "terminate"
@@ -434,10 +430,10 @@ class KataGoEngine(BaseEngine):
            avoid = []

        settings = copy.copy(self.override_settings)
+        settings["wideRootNoise"] = self.config["wide_root_noise"]
        if time_limit:
            settings["maxTime"] = self.config["max_time"]
-        if self.config.get("wide_root_noise", 0.0) > 0.0:  # don't send if 0.0, so older versions don't error
-            settings["wideRootNoise"] = self.config["wide_root_noise"]
+
        query = {
            "rules": self.get_rules(analysis_node.ruleset),
            "priority": self.base_priority + priority,
@@ -575,7 +575,6 @@ class Game(BaseGame):
            return

        engine = self.engines[cn.next_player]
-        Clock.schedule_once(self.katrain.analysis_controls.hints.activate, 0)

        if mode == "ponder":
            cn.analyze(
@@ -602,6 +601,10 @@ class Game(BaseGame):
        if mode == "game":
            nodes = self.root.nodes_in_tree
            only_mistakes = kwargs.get("mistakes_only", False)
+            move_range = kwargs.get("move_range", None)
+            if move_range:
+                if move_range[1] < move_range[0]:
+                    move_range = reversed(move_range)
            threshold = self.katrain.config("trainer/eval_thresholds")[-4]
            if "visits" in kwargs:
                visits = kwargs["visits"]
@@ -610,9 +613,20 @@ class Game(BaseGame):
                visits = min_visits + engine.config["max_visits"]
            for node in nodes:
                max_point_loss = max(c.points_lost or 0 for c in [node] + node.children)
-                if not only_mistakes or max_point_loss > threshold:
-                    node.analyze(engine, visits=visits, priority=-1_000_000, time_limit=False, report_every=None)
-            self.katrain.controls.set_status(i18n._("game re-analysis").format(visits=visits), STATUS_ANALYSIS)
+                if only_mistakes and max_point_loss <= threshold:
+                    continue
+                if move_range and (not node.depth - 1 in range(move_range[0], move_range[1] + 1)):
+                    continue
+                node.analyze(engine, visits=visits, priority=-1_000_000, time_limit=False, report_every=None)
+            if not move_range:
+                self.katrain.controls.set_status(i18n._("game re-analysis").format(visits=visits), STATUS_ANALYSIS)
+            else:
+                self.katrain.controls.set_status(
+                    i18n._("move range analysis").format(
+                        start_move=move_range[0], end_move=move_range[1], visits=visits
+                    ),
+                    STATUS_ANALYSIS,
+                )
            return

        elif mode == "sweep":
@@ -374,7 +374,7 @@ class SGFNode:
    def place_handicap_stones(self, n_handicaps, tygem=False):
        board_size_x, board_size_y = self.board_size
        if min(board_size_x, board_size_y) < 3:
-            return # No
+            return  # No
        near_x = 3 if board_size_x >= 13 else min(2, board_size_x - 1)
        near_y = 3 if board_size_y >= 13 else min(2, board_size_y - 1)
        far_x = board_size_x - 1 - near_x