sync

.vscode/launch.json

@@ -24,14 +24,19 @@
             "program": "train.py",
             "args": [
                 //"-c",
-                //"snerf_voxels",
-                "/home/dengnc/dvs/data/__new/barbershop_fovea_r360x80_t0.6/_nets/train_t0.3/snerfadvx_voxels_x4/checkpoint_10.tar",
+                //"snerf_voxels+ls+f32",
+                "/data1/dnc/dvs/data/__nerf/room/_nets/train/snerf_voxels+ls+f32/checkpoint_1.tar",
                 "--prune",
-                "100",
+                "1",
                 "--split",
-                "100"
-                //"data/__new/barbershop_fovea_r360x80_t0.6/train_t0.3.json"
+                "1",
+                "-e",
+                "100",
+                "--views",
+                "5",
+                //"data/__nerf/room/train.json"
             ],
+            "justMyCode": false,
             "console": "integratedTerminal"
         },
         {

configs/snerf_voxels+ls+f32.json

+{
+    "model": "SNeRF",
+    "args": {
+        "color": "rgb",
+        "n_pot_encode": 10,
+        "n_dir_encode": 4,
+        "fc_params": {
+            "nf": 256,
+            "n_layers": 8,
+            "activation": "relu",
+            "skips": [ 4 ]
+        },
+        "n_featdim": 32,
+        "space": "voxels",
+        "steps": [4, 16, 8],
+        "n_samples": 16,
+        "perturb_sample": true,
+        "density_regularization_weight": 1e-4,
+        "density_regularization_scale": 1e4
+    }
+}
\ No newline at end of file

model/base.py

@@ -16,7 +16,7 @@ class BaseModelMeta(type):
 
 class BaseModel(nn.Module, metaclass=BaseModelMeta):
 
-    trainer = "Train"
+    TrainerClass = "Train"
 
     @property
     def args(self):

model/nerf.py

@@ -10,7 +10,7 @@ from utils.misc import masked_scatter
 
 class NeRF(BaseModel):
 
-    trainer = "TrainWithSpace"
+    TrainerClass = "TrainWithSpace"
     SamplerClass = Sampler
     RendererClass = VolumnRenderer
 
@@ -124,21 +124,11 @@ class NeRF(BaseModel):
         return self.pot_encoder(x)
 
     def encode_d(self, samples: Samples) -> torch.Tensor:
-        return self.dir_encoder(samples.dirs) if self.dir_encoder is not None else None
+        return self.dir_encoder(samples.dirs) if self.dir_encoder else None
 
     @torch.no_grad()
-    def get_scores(self, sampled_points: torch.Tensor, sampled_voxel_indices: torch.Tensor) -> torch.Tensor:
-        densities = self.render(Samples(sampled_points, None, None, None, sampled_voxel_indices),
-                                'density')
-        return 1 - (-densities).exp()
-
-    @torch.no_grad()
-    def pruning(self, threshold: float = 0.5, train_stats=False):
-        return self.space.pruning(self.get_scores, threshold, train_stats)
-
-    @torch.no_grad()
-    def splitting(self):
-        ret = self.space.splitting()
+    def split(self):
+        ret = self.space.split()
         if 'n_samples' in self.args0:
             self.args0['n_samples'] *= 2
         if 'voxel_size' in self.args0:
@@ -149,12 +139,10 @@ class NeRF(BaseModel):
         if 'sample_step' in self.args0:
             self.args0['sample_step'] /= 2
         self.sampler = self.SamplerClass(**self.args)
+        if self.args.get('n_featdim') and hasattr(self, "trainer"):
+            self.trainer.reset_optimizer()
         return ret
 
-    @torch.no_grad()
-    def double_samples(self):
-        pass
-
     @perf
     def forward(self, rays_o: torch.Tensor, rays_d: torch.Tensor, *,
                 extra_outputs: List[str] = [], **kwargs) -> torch.Tensor:

model/snerf_advance_x.py

@@ -40,16 +40,8 @@ class SNeRFAdvanceX(SNeRFAdvance):
         return self.cores[chunk_id](x, d, outputs, **extras)
 
     @torch.no_grad()
-    def get_scores(self, sampled_points: torch.Tensor, sampled_voxel_indices: torch.Tensor) -> torch.Tensor:
-        raise NotImplementedError()
-
-    @torch.no_grad()
-    def pruning(self, threshold: float = 0.5, train_stats=False):
-        raise NotImplementedError()
-
-    @torch.no_grad()
-    def splitting(self):
-        ret = super().splitting()
+    def split(self):
+        ret = super().split()
         k = self.args["n_samples"] // self.space.steps[0].item()
         net_samples = [val * k for val in self.space.balance_cut(0, len(self.cores))]
         if len(net_samples) != len(self.cores):

model/snerf_x.py

@@ -4,10 +4,6 @@ from .snerf import *
 
 class SNeRFX(SNeRF):
 
-    trainer = "TrainWithSpace"
-    SamplerClass = SphericalSampler
-    RendererClass = VolumnRenderer
-
     def __init__(self, args0: dict, args1: dict = {}):
         """
         Initialize a multi-sphere-layer net
@@ -42,16 +38,8 @@ class SNeRFX(SNeRF):
         return self.cores[chunk_id](x, d, outputs)
 
     @torch.no_grad()
-    def get_scores(self, sampled_points: torch.Tensor, sampled_voxel_indices: torch.Tensor) -> torch.Tensor:
-        raise NotImplementedError()
-
-    @torch.no_grad()
-    def pruning(self, threshold: float = 0.5, train_stats=False):
-        raise NotImplementedError()
-
-    @torch.no_grad()
-    def splitting(self):
-        ret = super().splitting()
+    def split(self):
+        ret = super().split()
         k = self.args["n_samples"] // self.space.steps[0].item()
         net_samples = [
             val * k for val in self.space.balance_cut(0, len(self.cores))

modules/space.py

-from math import ceil
 import torch
-import numpy as np
-from typing import List, NoReturn, Tuple, Union
+from typing import List, Tuple, Union
 from torch import nn
-from plyfile import PlyData, PlyElement
 
 from utils.geometry import *
 from utils.constants import *
@@ -73,11 +70,11 @@ class Space(nn.Module):
         return voxel_indices
 
     @torch.no_grad()
-    def pruning(self, score_fn, threshold: float = 0.5, train_stats=False):
+    def prune(self, keeps: torch.Tensor) -> Tuple[int, int]:
         raise NotImplementedError()
 
     @torch.no_grad()
-    def splitting(self):
+    def split(self):
         raise NotImplementedError()
 
 
@@ -108,7 +105,7 @@ class Voxels(Space):
         return self.voxels.size(0)
 
     @property
-    def n_corner(self) -> int:
+    def n_corners(self) -> int:
         """`int` Number of corners"""
         return self.corners.size(0)
 
@@ -145,12 +142,18 @@ class Voxels(Space):
         :param n_dims `int`: embedding dimension
         :return `Embedding(n_corners, n_dims)`: new embedding on voxel corners
         """
-        name = f'emb_{name}'
-        self.add_module(name, torch.nn.Embedding(self.n_corners.item(), n_dims))
-        return self.__getattr__(name)
+        if self.get_embedding(name) is not None:
+            raise KeyError(f"Embedding '{name}' already existed")
+        emb = torch.nn.Embedding(self.n_corners, n_dims, device=self.device)
+        setattr(self, f'emb_{name}', emb)
+        return emb
 
     def get_embedding(self, name: str = 'default') -> torch.nn.Embedding:
-        return getattr(self, f'emb_{name}')
+        return getattr(self, f'emb_{name}', None)
+
+    def set_embedding(self, weight: torch.Tensor, name: str = 'default'):
+        emb = torch.nn.Embedding(*weight.shape, _weight=weight, device=self.device)
+        setattr(self, f'emb_{name}', emb)
 
     def extract_embedding(self, pts: torch.Tensor, voxel_indices: torch.Tensor,
                           name: str = 'default') -> torch.Tensor:
@@ -167,9 +170,8 @@ class Voxels(Space):
             raise KeyError(f"Embedding '{name}' doesn't exist")
         voxels = self.voxels[voxel_indices]  # (N, 3)
         corner_indices = self.corner_indices[voxel_indices]  # (N, 8)
-        p = (pts - voxels) / self.voxel_size + 0.5  # (N, 3) normed-coords in voxel
-        features = emb(corner_indices).reshape(pts.size(0), 8, -1)  # (N, 8, X)
-        return trilinear_interp(p, features)
+        p = (pts - voxels) / self.voxel_size + .5  # (N, 3) normed-coords in voxel
+        return trilinear_interp(p, emb(corner_indices))
 
     @perf
     def ray_intersect(self, rays_o: torch.Tensor, rays_d: torch.Tensor, n_max_hits: int) -> Intersections:
@@ -220,17 +222,34 @@ class Voxels(Space):
         return voxel_indices
 
     @torch.no_grad()
-    def splitting(self) -> None:
+    def split(self) -> None:
         """
         Split voxels into smaller voxels with half size.
         """
-        n_voxels_before = self.n_voxels
-        self.steps *= 2
-        self.voxels = split_voxels(self.voxels, self.voxel_size, 2, align_border=False)\
+        new_steps = self.steps * 2
+        new_voxels = split_voxels(self.voxels, self.voxel_size, 2, align_border=False)\
             .reshape(-1, 3)
-        self._update_corners()
+        new_corners, new_corner_indices = get_corners(new_voxels, self.bbox, new_steps)
+
+        # Calculate new embeddings through trilinear interpolation
+        grid_indices_of_new_corners = to_flat_indices(
+            to_grid_coords(new_corners, self.bbox, steps=self.steps).min(self.steps - 1),
+            self.steps)
+        voxel_indices_of_new_corners = self.voxel_indices_in_grid[grid_indices_of_new_corners]
+        for name, _ in self.named_modules():
+            if not name.startswith("emb_"):
+                continue
+            new_emb_weight = self.extract_embedding(new_corners, voxel_indices_of_new_corners,
+                                                    name=name[4:])
+            self.set_embedding(new_emb_weight, name=name[4:])
+
+        # Apply new tensors
+        self.steps = new_steps
+        self.voxels = new_voxels
+        self.corners = new_corners
+        self.corner_indices = new_corner_indices
         self._update_voxel_indices_in_grid()
-        return n_voxels_before, self.n_voxels
+        return self.n_voxels // 8, self.n_voxels
 
     @torch.no_grad()
     def prune(self, keeps: torch.Tensor) -> Tuple[int, int]:
@@ -239,11 +258,6 @@ class Voxels(Space):
         self._update_voxel_indices_in_grid()
         return keeps.size(0), keeps.sum().item()
 
-    @torch.no_grad()
-    def pruning(self, score_fn, threshold: float = 0.5) -> None:
-        scores = self._get_scores(score_fn, lambda x: torch.max(x, -1)[0])  # (M)
-        return self.prune(scores > threshold)
-
     def n_voxels_along_dim(self, dim: int) -> torch.Tensor:
         sum_dims = [val for val in range(self.dims) if val != dim]
         return self.voxel_indices_in_grid.reshape(*self.steps).ne(-1).sum(sum_dims)
@@ -261,39 +275,30 @@ class Voxels(Space):
                 part = int(cdf[i]) + 1
         return bins
 
-    def sample(self, bits: int, perturb: bool = False) -> Tuple[torch.Tensor, torch.Tensor]:
-        sampled_xyz = split_voxels(self.voxels, self.voxel_size, bits)
-        sampled_idx = torch.arange(self.n_voxels, device=self.device)[:, None].expand(
-            *sampled_xyz.shape[:2])
-        sampled_xyz, sampled_idx = sampled_xyz.reshape(-1, 3), sampled_idx.flatten()
+    def sample(self, S: int, perturb: bool = False, include_border: bool = True) -> Tuple[torch.Tensor, torch.Tensor]:
+        """
+        For each voxel, sample `S^3` points uniformly, with small perturb if `perturb` is `True`.
+
+        When `perturb` is `False`, `include_border` can specify whether to sample points from border to border or at centers of sub-voxels.
+        When `perturb` is `True`, points are sampled at centers of sub-voxels, then applying a random offset in sub-voxels.
 
-    @torch.no_grad()
-    def _get_scores(self, score_fn, reduce_fn=None, bits=16) -> torch.Tensor:
-        def get_scores_once(pts, idxs):
-            scores = score_fn(pts, idxs).reshape(-1, bits ** 3)  # (B, P)
-            if reduce_fn is not None:
-                scores = reduce_fn(scores)  # (B[, ...])
-            return scores
-
-        sampled_xyz, sampled_idx = self.sample(bits)
-        chunk_size = 64
-        return torch.cat([
-            get_scores_once(sampled_xyz[i:i + chunk_size], sampled_idx[i:i + chunk_size])
-            for i in range(0, self.voxels.size(0), chunk_size)
-        ], 0)  # (M[, ...])
+        :param S `int`: number of samples along each dim
+        :param perturb `bool?`: whether perturb samples, defaults to `False`
+        :param include_border `bool?`: whether include border, defaults to `True`
+        :return `Tensor(N*S^3, 3)`: sampled points
+        :return `Tensor(N*S^3)`: voxel indices of sampled points
+        """
+        pts = split_voxels(self.voxels, self.voxel_size, S,
+                           align_border=not perturb and include_border)  # (N, X, D)
+        voxel_indices = torch.arange(self.n_voxels, device=self.device)[:, None]\
+            .expand(*pts.shape[:-1])  # (N) -> (N, X)
+        if perturb:
+            pts += (torch.rand_like(pts) - .5) * self.voxel_size / S
+        return pts.reshape(-1, 3), voxel_indices.flatten()
 
     def _ray_intersect(self, rays_o: torch.Tensor, rays_d: torch.Tensor, n_max_hits: int) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor]:
         return aabb_ray_intersect(self.voxel_size, n_max_hits, self.voxels, rays_o, rays_d)
 
-
-    def _update_corners(self):
-        """
-        Update voxel corners.
-        """
-        corners, corner_indices = get_corners(self.voxels, self.bbox, self.steps)
-        self.register_buffer("corners", corners)
-        self.register_buffer("corner_indices", corner_indices)
-
     def _update_voxel_indices_in_grid(self):
         """
         Update voxel indices in grid.
@@ -314,7 +319,7 @@ class Voxels(Space):
         # Handle embeddings
         for name, module in self.named_modules():
             if name.startswith('emb_'):
-                setattr(self, name, torch.nn.Embedding(self.n_corners.item(), module.embedding_dim))
+                setattr(self, name, torch.nn.Embedding(self.n_corners, module.embedding_dim))
 
 
 class Octree(Voxels):
@@ -339,8 +344,8 @@ class Octree(Voxels):
         return octree_ray_intersect(self.voxel_size, n_max_hits, nodes, tree, rays_o, rays_d)
 
     @torch.no_grad()
-    def splitting(self):
-        ret = super().splitting()
+    def split(self):
+        ret = super().split()
         self.clear()
         return ret
 

test.ipynb

+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 26,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import torch\n",
+    "from utils.voxels import *\n",
+    "\n",
+    "bbox, steps = torch.tensor([[-2, -3.14159, 1], [2, 3.14159, 0]]), torch.tensor([2, 3, 3])\n",
+    "voxel_size = (bbox[1] - bbox[0]) / steps\n",
+    "voxels = init_voxels(bbox, steps)\n",
+    "corners, corner_indices = get_corners(voxels, bbox, steps)\n",
+    "voxel_indices_in_grid = torch.arange(voxels.shape[0])\n",
+    "emb = torch.nn.Embedding(corners.shape[0], 3, _weight=corners)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 27,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "torch.Size([11, 3]) tensor([ 0, -1, -1,  1, -1, -1,  2,  3,  4, -1,  5,  6, -1,  7,  8, -1,  9, 10])\n"
+     ]
+    }
+   ],
+   "source": [
+    "keeps = torch.tensor([True]*18)\n",
+    "keeps[torch.tensor([1,2,4,5,9,12,15])] = False\n",
+    "voxels = voxels[keeps]\n",
+    "corner_indices = corner_indices[keeps]\n",
+    "grid_indices, _ = to_grid_indices(voxels, bbox, steps=steps)\n",
+    "voxel_indices_in_grid = grid_indices.new_full([steps.prod().item()], -1)\n",
+    "voxel_indices_in_grid[grid_indices] = torch.arange(voxels.shape[0])\n",
+    "print(voxels.shape, voxel_indices_in_grid)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 28,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "torch.Size([88, 3]) torch.Size([185, 3]) torch.Size([88, 8])\n"
+     ]
+    }
+   ],
+   "source": [
+    "new_voxels = split_voxels(voxels, (bbox[1] - bbox[0]) / steps, 2, align_border=False).reshape(-1, 3)\n",
+    "new_corners, new_corner_indices = get_corners(new_voxels, bbox, steps * 2)\n",
+    "print(new_voxels.shape, new_corners.shape, new_corner_indices.shape)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 30,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "tensor([ 0,  0, -1,  0,  0, -1,  1,  1, -1,  1,  1, -1,  2,  2,  3,  3,  4,  4,\n",
+      "         4,  2,  2,  3,  3,  4,  4,  4,  2,  2,  3,  3,  4,  4,  4,  0,  0, -1,\n",
+      "         0,  0, -1,  1,  1, -1,  1,  1, -1,  2,  2,  3,  3,  4,  4,  4,  2,  2,\n",
+      "         3,  3,  4,  4,  4,  2,  2,  3,  3,  4,  4,  4, -1, -1,  5,  5,  6,  6,\n",
+      "         6, -1, -1,  5,  5,  6,  6,  6, -1, -1,  7,  7,  8,  8,  8, -1, -1,  7,\n",
+      "         7,  8,  8,  8, -1, -1,  9,  9, 10, 10, 10, -1, -1,  9,  9, 10, 10, 10,\n",
+      "        -1, -1,  9,  9, 10, 10, 10,  5,  5,  6,  6,  6,  5,  5,  6,  6,  6,  7,\n",
+      "         7,  8,  8,  8,  7,  7,  8,  8,  8,  9,  9, 10, 10, 10,  9,  9, 10, 10,\n",
+      "        10,  9,  9, 10, 10, 10,  5,  5,  6,  6,  6,  5,  5,  6,  6,  6,  7,  7,\n",
+      "         8,  8,  8,  7,  7,  8,  8,  8,  9,  9, 10, 10, 10,  9,  9, 10, 10, 10,\n",
+      "         9,  9, 10, 10, 10])\n",
+      "tensor(0)\n"
+     ]
+    }
+   ],
+   "source": [
+    "voxel_indices_of_new_corner = voxel_indices_in_grid[to_flat_indices(to_grid_coords(new_corners, bbox, steps=steps).min(steps - 1), steps)]\n",
+    "print(voxel_indices_of_new_corner)\n",
+    "p_of_new_corners = (new_corners - voxels[voxel_indices_of_new_corner]) / voxel_size + .5\n",
+    "print(((new_corners - trilinear_interp(p_of_new_corners, emb(corner_indices[voxel_indices_of_new_corner]))) > 1e-6).sum())"
+   ]
+  }
+ ],
+ "metadata": {
+  "interpreter": {
+   "hash": "08b118544df3cb8970a671e5837a88fd458f4d4c799ef1fb2709465a22a45b92"
+  },
+  "kernelspec": {
+   "display_name": "Python 3.9.5 64-bit ('base': conda)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.9.5"
+  },
+  "orig_nbformat": 4
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}

test.py

@@ -38,7 +38,7 @@ from data.loader import DataLoader
 from utils.constants import HUGE_FLOAT
 
 
-RAYS_PER_BATCH = 2 ** 14
+RAYS_PER_BATCH = 2 ** 12
 DATA_LOADER_CHUNK_SIZE = 1e8
 
 

test1.txt

+AT4G28520
+AT2G39730
+AT5G02500
+AT2G07560
+AT4G33010
+AT5G17920
+AT5G04140
+AT2G36880
+AT2G26080
+AT1G56070
+AT3G08530
+AT3G09440
+AT5G53460
+AT3G11130
+AT3G17390
+AT1G52570
+AT3G03780
+AT2G36530
+AT5G61780
+AT2G18960
+AT5G56000
+AT2G21390
+AT4G37930
+AT5G28540
+AT3G09840
+AT5G50920
+AT2G41740
+AT4G31490
+AT4G30190
+AT4G29060
+AT2G33150
+AT5G08680
+AT4G31480
+AT1G62020
+AT3G13470
+AT5G65690
+AT3G52930
+AT3G02090
+AT1G09780
+AT1G67290
+AT3G12580
+AT3G08590
+AT3G43300
+AT4G34450
+AT1G23190
+AT3G15730
+AT2G16730
+AT5G66760
+AT3G19170
+AT3G23810
+AT3G55410
+AT5G02490
+AT1G70730
+AT2G05710
+AT3G13390
+AT2G04030
+AT1G55490
+AT3G48870
+AT5G15490
+AT1G79930
+AT1G07890
+AT1G02500
+AT2G28000
+AT3G62170
+AT5G57320
+AT5G20890
+AT2G36460
+AT1G06950
+AT1G59610
+AT3G11710
+AT3G09630
+AT1G23310
+AT3G04120
+AT5G49910
+AT4G24280
+AT1G10290
+AT3G28820
+AT5G42270
+AT5G26360
+AT1G53500
+AT1G13440
+AT3G46520
+AT3G45140
+AT5G39320
+AT1G53310
+AT3G12780
+AT1G78900
+AT5G03340
+AT2G47510
+AT3G59970
+AT1G32060
+AT1G03880
+AT4G37910
+AT5G09810
+AT1G07810
+AT4G27500
+AT1G68750
+AT5G13990
+AT1G74470
+AT3G12110
+AT4G14360
+AT5G66570
+AT1G80070
+AT5G58050
+AT4G38970
+AT1G56410
+AT5G07410
+AT3G57290
+AT3G62030
+AT2G18020
+AT4G36130
+AT1G55570
+AT2G36580
+AT4G20360
+AT1G14610
+AT1G57720
+AT5G57350
+AT1G14830
+AT4G35830
+AT3G46750
+AT1G68010
+AT5G13630
+AT3G18190
+AT1G07670
+AT5G20920
+AT5G19580
+AT3G06860
+AT3G17240
+AT5G35630
+AT3G46970
+AT2G27880
+AT3G26650
+AT3G62830
+AT2G21330
+AT2G20580
+AT3G62360
+AT3G51070
+AT3G58610
+AT1G33170
+AT4G35010
+AT3G28790
+AT5G58170
+AT1G69940
+AT5G11560
+AT3G58510
+AT3G13290
+AT4G10440
+AT5G09550
+AT1G78570
+AT3G21180
+AT1G26630
+AT5G21150
+AT3G44310
+AT2G16600
+AT1G06680
+AT2G28190
+AT3G05420
+AT1G54560
+AT2G20420
+AT3G52600
+AT5G43330
+AT1G56190
+AT4G16760
+AT3G18780
+AT3G13460
+AT3G01500
+AT2G27040
+AT3G57410
+AT5G08280
+AT3G23300
+AT1G20950
+AT4G29130
+AT1G04410
+AT1G80410
+AT5G34850
+AT2G39990
+AT3G02230
+AT4G15980
+AT5G48140
+AT1G26850
+AT3G46780
+AT5G58290
+AT5G11170
+AT5G11200
+AT1G03630
+AT4G04910
+AT1G22530
+AT1G70770
+AT1G47260
+AT1G76030
+AT3G27240
+AT1G12270
+AT3G53230
+AT3G03960
+AT2G40660
+AT5G15090
+AT5G19990
+AT1G18070
+AT4G34980
+AT5G62530
+AT2G40840
+AT4G26530
+AT4G09000
+AT5G19510
+AT1G18080
+AT3G55800
+AT1G65930
+AT2G32730
+AT5G37830
+AT1G66410
+AT5G37780
+AT1G29900
+AT1G79550
+AT3G18130
+AT5G38430
+AT1G67090
+AT5G56500
+AT4G24800
+AT5G53620
+AT3G07830
+AT5G35360
+AT1G48630
+AT5G08300
+AT3G02530
+AT5G24940
+AT4G38510
+AT3G11830
+AT1G23800
+AT4G27440
+AT3G17820
+AT1G54870
+AT5G08530
+AT5G19820
+AT5G62690
+AT5G62700
+AT5G06600
+AT1G15690
+AT5G26570
+AT4G38740
+AT2G46280
+AT5G07420
+AT4G14880
+AT2G35690
+AT3G58500
+AT3G02360
+AT3G17940
+AT3G59920
+AT3G60240
+AT3G25860
+AT5G61720
+AT3G14210
+AT2G45290
+AT3G01280
+AT1G25350
+AT5G21274
+AT5G26742
+AT1G78300
+AT1G04170
+AT2G30970
+AT3G09820
+AT1G18500
+AT1G72960
+AT4G21150
+AT1G35720
+AT3G54050
+AT5G20720
+AT4G34870
+AT1G72730
+AT2G13360
+AT2G27860
+AT2G43750
+AT5G50600
+AT5G50700
+AT1G53750
+AT3G04840
+AT5G58420
+AT4G04640
+AT2G35840
+AT4G10260
+AT3G55200
+AT3G55220
+AT5G23060
+AT1G13930
+AT1G69740
+AT2G17980
+AT3G06650
+AT3G47950
+AT3G16000
+AT3G11964
+AT5G22800
+AT1G32470
+AT5G28840
+AT4G19120
+AT2G30490
+AT2G47650
+AT4G14960
+AT5G05010
+AT3G51160
+AT2G42500
+AT5G66020
+AT1G35160
+AT3G51800
+AT1G09640
+AT5G48230
+AT3G15640
+AT2G30050
+AT1G04820
+AT1G50010
+AT1G50480
+AT4G34670
+AT1G35670
+AT3G12290
+AT3G28770
+AT1G48090
+AT1G18210
+AT4G34200
+AT5G23860
+AT3G48560
+AT2G31390
+AT2G41220
+AT1G11660
+AT5G15650
+AT3G05970
+AT5G49460
+AT2G19520
+AT3G56190
+AT3G46440
+AT2G44160
+AT2G29550
+AT3G12915
+AT1G20020
+AT3G11630
+AT5G11110
+AT1G29150
+AT5G15450
+AT1G79340
+AT5G20010
+AT4G18480
+AT5G60980
+AT5G40810
+AT5G57870
+AT4G39330
+AT1G13950
+AT5G36700
+AT5G36790
+AT3G20050
+AT1G62390
+AT2G47030
+AT1G64190
+AT3G55440
+AT1G24510
+AT3G01310
+AT5G65720
+AT4G39800
+AT4G23850
+AT5G52640
+AT4G25630
+AT3G13860
+AT5G24710
+AT5G02240
+AT1G02790
+AT5G20290
+AT5G42740
+AT3G12145
+AT5G61410
+AT5G56680
+AT2G21130
+AT4G23670
+AT2G21170
+AT5G11880
+AT4G18100
+AT1G24620
+AT5G60790
+AT4G29840
+AT1G74050
+AT5G54770
+AT5G06290
+AT1G29880
+AT4G31700
+AT2G41680
+AT3G53110
+AT3G10370
+AT2G22560
+AT5G47210
+AT3G63490
+AT1G69410
+AT3G05530
+AT5G52920
+AT3G22310
+AT2G25110
+AT2G24270
+AT3G52880
+AT4G34860
+AT1G37130
+AT5G50850
+AT5G64690
+AT2G21870
+AT1G30350
+AT1G75990
+AT4G24620
+AT2G14740
+AT5G42790
+AT4G25900
+AT5G35530
+AT1G04810
+AT3G47520
+AT5G38480
+AT3G02520
+AT5G13650
+AT3G19450
+AT1G32780
+AT2G41840
+AT3G01640
+AT5G43940
+AT5G49650
+AT4G02770
+AT5G41670
+AT1G07370
+AT3G62870
+AT3G47930
+AT1G65960
+AT2G28680
+AT1G64790
+AT3G58570
+AT2G38770
+AT4G20890
+AT1G26570
+AT4G24680
+AT5G16050
+AT5G60160
+AT1G17220
+AT1G49670
+AT1G30630
+AT3G18000
+AT3G63410
+AT5G48480
+AT5G66510
+AT1G78820
+AT1G04710
+AT5G43060
+AT1G30580
+AT5G04430
+AT3G28500
+AT1G17100
+AT2G01250
+AT4G02150
+AT5G25880
+AT1G23730
+AT1G68560
+AT3G07100
+AT3G13870
+AT1G65730
+AT2G41790
+AT1G07750
+AT5G63680
+AT5G04420
+AT2G31610
+AT2G05920
+AT4G17520
+AT1G78630
+AT3G07110
+AT1G69830
+AT1G56450
+AT5G19220
+AT1G12000
+AT2G07050
+AT1G11770
+AT3G14310
+AT4G31180
+AT5G20980
+AT3G53870
+AT4G11030
+AT1G78830
+AT1G26480
+AT4G20980
+AT4G12650
+AT3G26060
+AT3G24830
+AT2G33340
+AT3G26450
+AT1G03890
+AT1G08360
+AT1G50370
+AT5G10360
+AT3G17060
+AT3G06580
+AT2G43770
+AT5G65430
+AT3G56940
+AT5G16840
+AT1G49630
+AT1G56110
+AT1G06410
+AT3G08943
+AT2G40490
+AT2G29570
+AT5G35590
+AT4G33070
+AT5G03630
+AT5G48760
+AT4G39080
+AT3G52960
+AT2G45800
+AT1G18540
+AT5G61970
+AT1G15750
+AT1G04430
+AT1G50310
+AT4G26270
+AT2G04842
+AT5G48650
+AT3G19820
+AT4G18920
+AT3G61760
+AT5G47180
+AT4G19410
+AT5G39310
+AT3G61240
+AT4G01310
+AT5G01410
+AT4G11150
+AT1G74000
+AT5G07440
+AT5G44340
+AT3G22200
+AT2G47730
+AT1G43670
+AT4G34110
+AT5G65020
+AT5G63570
+AT2G19730
+AT3G05060
+AT3G06720
+AT5G25100
+AT5G39410
+AT1G77490
+AT3G25800
+AT1G64520
+AT2G27530
+AT3G62250
+AT3G02970
+AT5G04180
+AT3G56090
+AT2G33730
+AT5G06760
+AT3G23940
+AT4G39280
+AT2G40730
+AT4G20460
+AT1G70310
+AT1G13020
+AT3G09200
+AT5G06060
+AT4G35860
+AT5G22060
+AT1G66680
+AT5G45390
+AT4G05530
+AT1G72150
+AT3G08560
+AT1G77940
+AT4G24190
+AT1G22780
+AT1G34030
+AT4G09800
+AT1G13320
+AT1G35580
+AT5G27470
+AT2G42910
+AT5G06460
+AT1G06220
+AT2G38230
+AT1G56500
+AT1G32200
+AT1G15500
+AT4G01660
+AT5G22440
+AT1G16350
+AT1G31330
+AT1G34430
+AT2G44640
+AT3G27190
+AT5G08160
+AT2G36390
+AT1G09100
+AT3G57150
+AT1G14650
+AT4G27585
+AT1G54220
+AT1G31690
+AT2G37270
+AT2G04390
+AT1G67430
+AT4G33650
+AT2G17190
+AT1G16470
+AT4G33030
+AT4G26780
+AT3G58750
+AT2G44120
+AT3G19980
+AT5G43010
+AT1G53850
+AT2G42490
+AT3G63520
+AT5G46290
+AT4G35790
+AT5G30510
+AT3G55610
+AT1G20010
+AT1G63000
+AT3G22890
+AT1G02780
+AT1G43710
+AT1G54450
+AT1G01960
+AT5G49020
+AT1G13060
+AT3G46830
+AT4G39980
+AT3G13330
+AT1G61580
+AT4G17170
+AT3G11940
+AT2G46860
+AT5G12250
+AT5G46430
+AT3G01120
+AT2G18230
+AT2G27710
+AT3G18080
+AT1G10270
+AT1G02080
+AT4G01690
+AT4G09570
+AT3G49910
+AT5G37600
+AT3G25660
+AT1G70410
+AT1G20130
+AT5G16130
+AT2G22475
+AT4G10340
+AT2G30200
+AT5G36260
+AT1G75350
+AT4G02080
+AT3G20000
+AT4G38680
+AT3G48420
+AT2G37770
+AT1G31780
+AT3G13580
+AT2G33040
+AT2G03020
+AT3G62600
+AT5G64460
+AT3G60600
+AT1G56330
+AT4G14570
+AT4G36750
+AT3G12800
+AT1G23740
+AT1G48860
+AT4G17090
+AT5G23900
+AT1G14320
+AT3G15090
+AT5G27850
+AT3G54470
+AT4G33680
+AT2G45810
+AT1G78060
+AT1G45000
+AT5G18230
+AT3G01340
+AT2G05990
+AT4G32260
+AT1G30360
+AT2G21060
+AT5G40040
+AT3G11250
+AT5G40650
+AT5G09660
+AT5G63510
+AT2G47470
+AT2G37040
+AT5G25780
+AT5G16970
+AT3G49720
+AT4G16480
+AT1G19570
+AT3G02560
+AT4G35630
+AT5G64440
+AT5G67030
+AT5G08570
+AT2G37190
+AT2G35210
+AT1G66580
+AT1G78850
+AT1G79040
+AT2G43710
+AT1G32440
+AT1G14850
+AT5G03940
+AT2G20360
+AT3G11950
+AT1G48470
+AT2G38280
+AT5G65010
+AT4G24820
+AT1G48900
+AT3G53430
+AT5G28500
+AT4G11380
+AT3G18060
+AT5G61170
+AT2G19950
+AT1G80480
+AT3G06350
+AT1G22920
+AT1G13890
+AT4G27070
+AT5G40870
+AT3G01480
+AT1G09870
+AT4G09320
+AT4G25080
+AT5G47200
+AT5G66680
+AT2G41560
+AT1G29910
+AT1G29920
+AT1G07770
+AT5G59850
+AT1G29965
+AT1G19580
+AT3G27380
+AT3G53580
+AT1G80670
+AT2G33800
+AT1G75930
+AT2G45790
+AT2G01140
+AT3G55280
+AT5G47110
+AT5G52840
+AT4G30160
+AT1G59900
+AT3G14930
+AT3G14067
+AT1G78940
+AT5G54430
+AT4G09720
+AT5G63860
+AT3G04790
+AT3G23400
+AT1G03475
+AT5G38830
+AT2G22240
+AT4G27700
+AT4G35260
+AT5G26000
+AT2G34430
+AT5G56350
+AT3G55190
+AT1G09750
+AT1G17290
+AT1G20330
+AT2G45030
+AT5G54810
+AT3G16380
+AT3G08940
+AT1G60070
+AT5G19760
+AT5G41950
+AT4G18360
+AT4G16210
+AT2G33845
+AT1G02150
+AT5G64760
+AT4G01480
+AT1G05190
+AT3G60570
+AT4G16720
+AT2G32920
+AT3G54540
+AT3G09810
+AT1G26830
+AT2G47680
+AT3G04870
+AT3G10540
+AT5G20080
+AT3G07660
+AT4G34350
+AT4G37840
+AT5G27980
+AT5G51430
+AT5G20710
+AT5G16300
+AT5G53400
+AT4G14030
+AT4G23650
+AT5G13710
+AT5G23540
+AT2G43030
+AT2G38700
+AT1G29670
+AT1G80560
+AT5G14040
+AT1G78920
+AT5G13120
+AT5G66140
+AT4G16390
+AT3G03060
+AT2G37550
+AT5G27030
+AT3G10380
+AT3G25920
+AT3G26400
+AT2G27600
+AT4G00430
+AT5G46580
+AT2G45300
+AT5G03650
+AT5G12110
+AT1G79500
+AT4G05420
+AT3G28715
+AT5G14780
+AT1G58290
+AT5G50370
+AT4G09670
+AT2G35040
+AT3G15670
+AT5G14030
+AT4G00110
+AT2G36060
+AT5G19600
+AT3G23700
+AT3G20320
+AT1G25490
+AT3G24430
+AT3G55330
+AT1G07140
+AT2G45060
+AT5G17020
+AT5G45620
+AT3G22110
+AT3G15610
+AT2G34420
+AT2G41210
+AT3G08580
+AT3G10410
+AT4G37990
+AT5G27120
+AT2G37760
+AT3G04620
+AT4G12980
+AT5G65110
+AT5G06140
+AT4G11010
+AT5G58140
+AT5G67630
+AT1G01620
+AT5G66280
+AT1G21630
+AT5G17770
+AT4G08870
+AT4G32720
+AT2G24050
+AT1G52280
+AT1G55480
+AT3G06830
+AT1G31730
+AT5G26667
+AT5G47010
+AT1G14670
+AT4G11260
+AT1G31230
+AT4G04040
+AT5G11710
+AT1G16720
+AT5G60670
+AT5G01600
+AT3G27740
+AT5G65950
+AT1G20340
+AT3G26340
+AT2G15430
+AT1G64980
+AT1G30620
+AT1G12840
+AT1G01050
+AT5G20830
+AT1G78380
+AT3G62010
+AT4G05160
+AT4G24640
+AT5G42220
+AT4G21100
+AT4G22890
+AT1G54100
+AT5G20070
+AT5G27600
+AT1G74960
+AT4G29510
+AT2G36500
+AT5G20990
+AT1G16760
+AT2G33100
+AT1G71500
+AT1G79250
+AT1G69960
+AT1G30050
+AT2G43090
+AT2G43040
+AT1G12920
+AT5G13490
+AT1G47200
+AT1G05010
+AT3G17630
+AT4G09730
+AT2G27680
+AT2G21490
+AT4G18810
+AT1G18450
+AT4G30950
+AT1G07410
+AT3G10050
+AT4G26500
+AT1G47980
+AT3G10670
+AT5G38470
+AT2G17870
+AT2G34460
+AT3G61430
+AT3G56460
+AT4G03520
+AT2G21250
+AT4G19210
+AT3G48680
+AT1G10670
+AT4G20260
+AT1G71860
+AT4G28440
+AT1G48420
+AT5G63400
+AT2G01720
+AT1G79750
+AT3G19020
+AT1G14950
+AT3G12490
+AT1G55540
+AT3G05560
+AT5G58230
+AT2G25140
+AT5G59150
+AT3G16640
+AT4G17890
+AT4G11860
+AT1G17720
+AT5G20630
+AT2G43790
+AT2G45070
+AT1G79860
+AT1G15730
+AT1G11250
+AT5G45280
+AT4G08900
+AT3G27925
+AT5G63190
+AT4G23900
+AT5G57020
+AT1G27680
+AT1G74010
+AT5G27770
+AT5G08650
+AT4G35700
+AT1G60770
+AT1G56360
+AT2G19860
+AT3G61200
+AT1G22730
+AT1G71780
+AT2G09990
+AT5G58070
+AT4G13560
+AT5G18380
+AT1G54280
+AT2G38670
+AT2G01470
+AT5G15550
+AT3G61820
+AT3G45030
+AT5G62300
+AT5G22480
+AT1G09270
+AT2G25290
+AT2G37220
+AT5G55230
+AT1G11870
+AT3G53710
+AT5G19360
+AT3G53900
+AT5G41520
+AT4G38230
+AT2G13540
+AT2G44530
+AT5G44070
+AT3G59980
+AT5G20230
+AT4G26510
+AT2G28490
+AT3G16780
+AT1G71820
+AT3G16100
+AT3G48930
+AT5G51970
+AT2G31980
+AT3G27300
+AT5G20280
+AT1G61790
+AT2G45240
+AT4G26410
+AT5G23140
+AT3G63170
+AT4G39260
+AT5G11670
+AT4G23630
+AT3G48170
+AT4G32840
+AT1G63180
+AT1G06530
+AT5G27630
+AT5G10170
+AT5G12180
+AT3G16910
+AT1G65030
+AT3G13065
+AT3G60820
+AT5G60010
+AT4G23460
+AT2G39130
+AT5G01530
+AT3G62310
+AT4G18030
+AT4G18596
+AT4G31340
+AT3G20680
+AT1G76080
+AT4G27090
+AT1G02280
+AT2G17280
+AT3G02720
+AT1G69200
+AT3G63150
+AT2G26890
+AT1G47490
+AT3G47370
+AT3G01540
+AT4G30910
+AT1G59830
+AT2G06850
+AT5G57330
+AT3G51470
+AT2G26230
+AT5G57490
+AT1G26910
+AT2G23070
+AT3G51810
+AT5G06160
+AT1G17470
+AT5G10500
+AT1G60780
+AT5G58440
+AT5G53480
+AT5G38640
+AT3G60770
+AT3G48860
+AT5G65260
+AT5G63890
+AT5G47770
+AT5G64030
+AT1G02560
+AT4G04350
+AT1G31910
+AT1G02000
+AT2G45820
+AT2G13620
+AT3G07960
+AT1G59820
+AT1G67930
+AT3G05500
+AT3G49080
+AT5G40770
+AT5G05780
+AT2G02130
+AT1G55150
+AT4G27320
+AT1G74850
+AT1G01100
+AT1G64770
+AT1G61150
+AT5G19620
+AT2G43945
+AT4G15560
+AT3G53620
+AT5G51660
+AT3G54110
+AT4G03620
+AT3G05090
+AT1G75680
+AT2G17390
+AT5G16290
+AT3G55750
+AT5G14220
+AT5G53560
+AT5G04740
+AT2G33840
+AT4G16830
+AT4G35890
+AT5G53140
+AT3G18740
+AT3G22850
+AT2G31670
+AT3G05590
+AT1G76400
+AT5G05570
+AT2G22450
+AT4G33760
+AT3G63190
+AT1G79570
+AT2G25600
+AT4G08630
+AT1G04690
+AT1G65540
+AT5G11770
+AT2G23120
+AT1G16780
+AT5G46750
+AT5G13150
+AT4G14300
+AT4G36020
+AT2G41530
+AT1G12310
+AT1G20760
+AT2G40300
+AT4G04770
+AT3G44590
+AT4G31750
+AT1G20810
+AT3G53970
+AT5G05000
+AT3G03110
+AT5G10260
+AT5G03740
+AT1G03860
+AT3G61140
+AT1G41880
+AT3G13227
+AT5G24420
+AT2G33870
+AT1G26460
+AT1G24020
+AT5G14660
+AT5G23880
+AT3G56340
+AT3G07630
+AT5G09650
+AT4G35100
+AT4G29120
+AT4G14230
+AT1G15130
+AT2G30720
+AT2G26060
+AT1G71880
+AT3G50000
+AT5G20490
+AT5G64050
+AT1G32990
+AT1G76450
+AT2G02100
+AT4G20780
+AT3G21200
+AT3G16830
+AT1G08980
+AT1G48950
+AT3G61230
+AT5G09500
+AT1G07660
+AT1G07820
+AT2G28740
+AT3G45930
+AT3G46320
+AT3G53730
+AT5G59690
+AT5G59970
+AT2G27510
+AT1G61690
+AT5G20690
+AT4G16160
+AT5G66470
+AT2G02560
+AT4G30440
+AT1G10510
+AT2G44610
+AT5G59880
+AT3G58700
+AT4G18730
+AT5G45775
+AT3G52300
+AT2G43910
+AT1G62780
+AT5G18900
+AT3G07160
+AT5G52340
+AT2G02790
+AT1G23290
+AT1G44970
+AT2G17800
+AT4G33870
+AT5G66420
+AT3G26520
+AT1G15140
+AT1G11840
+AT3G12690
+AT5G19440
+AT2G47400
+AT3G03950
+AT1G21670
+AT3G10460
+AT5G18100
+AT4G15545
+AT5G47930
+AT1G49300
+AT1G75200
+AT3G49540
+AT3G13490
+AT3G46940
+AT1G61770
+AT4G32605
+AT1G20575
+AT5G19370
+AT4G21860
+AT3G58730
+AT3G61260
+AT1G78370
+AT5G52360
+AT2G47970
+AT4G22930
+AT1G48160
+AT3G27820
+AT3G23780
+AT2G19010
+AT4G20410
+AT4G23400
+AT2G21410
+AT1G32220
+AT3G18820
+AT5G56280
+AT4G02230
+AT4G35850
+AT3G55620
+AT5G51030
+AT2G22250
+AT3G53180
+AT1G50940
+AT5G06110
+AT2G39770
+AT1G76510
+AT1G52400
+AT1G24400
+AT3G15280
+AT4G15000
+AT1G16270
+AT1G68200
+AT4G36690
+AT1G61520
+AT1G54780
+AT4G30530
+AT1G48520
+AT3G03080
+AT1G76850
+AT1G53580
+AT5G15980
+AT5G65940
+AT3G56070
+AT3G08970
+AT3G51240
+AT5G66030
+AT3G66654
+AT4G09650
+AT1G21380
+AT5G40370
+AT2G35120
+AT1G67700
+AT1G64440
+AT5G14320
+AT2G33793
+AT3G46040
+AT5G46630
+AT3G13750
+AT4G36390
+AT4G15093
+AT5G18520
+AT3G02830
+AT5G05670
+AT4G10840
+AT2G46470
+AT4G34700
+AT5G57460
+AT5G02770
+AT1G05350
+AT5G15200
+AT1G31817
+AT5G07470
+AT5G47840
+AT5G55280
+AT5G59840
+AT2G16440
+AT3G19930
+AT1G04160
+AT3G49120
+AT5G09840
+AT4G12420
+AT5G62270
+AT3G16520
+AT1G26110
+AT4G25590
+AT4G34640
+AT4G37510
+AT4G25890
+AT4G25780
+AT3G50110
+AT1G58215
+AT5G19290
+AT3G13160
+AT5G37850
+AT5G16590
+AT2G33220
+AT3G06483
+AT2G04700
+AT2G16430
+AT3G18490
+AT3G48690
+AT4G11220
+AT3G07880
+AT5G23740
+AT2G22780
+AT2G13560
+AT3G27280
+AT4G09520
+AT3G53500
+AT1G18280
+AT4G35950
+AT4G33945
+AT3G28940
+AT3G26618
+AT5G27410
+AT1G04760
+AT4G29410
+AT1G73990
+AT5G16780
+AT5G61500
+AT1G07210
+AT3G03920
+AT3G20390
+AT1G44900
+AT4G14800
+AT4G20530
+AT4G20540
+AT4G20550
+AT4G20560
+AT4G20570
+AT4G20645
+AT4G20670
+AT3G15180
+AT1G04190
+AT3G51780
+AT1G74970
+AT3G03980
+AT1G62820
+AT4G26840
+AT3G19230
+AT3G20410
+AT1G60810
+AT1G16340
+AT5G24850
+AT1G63060
+AT3G24090
+AT1G03120
+AT1G15340
+AT1G34220
+AT1G64040
+AT1G35620
+AT3G24350
+AT4G13850
+AT4G37000
+AT2G21280
+AT5G35620
+AT5G54180
+AT4G16180
+AT2G37370
+AT5G07490
+AT2G33470
+AT3G18230
+AT5G57290
+AT1G32380
+AT1G08880
+AT1G34130
+AT3G10060
+AT5G48180
+AT2G32060
+AT3G44750
+AT2G28900
+AT4G30800
+AT3G55360
+AT3G05280
+AT3G16370
+AT1G09830
+AT3G23920
+AT5G42960
+AT1G23900
+AT2G35500
+AT1G09760
+AT1G09580
+AT5G24690
+AT5G66740
+AT4G31500
+AT5G48570
+AT2G07707
+AT5G16440
+AT4G24550
+AT1G04980
+AT1G74640
+AT2G14750
+AT3G13772
+AT5G38840
+AT3G26720
+AT2G32600
+AT4G13840
+AT3G51300
+AT3G53610
+AT4G10790
+AT5G11980
+AT5G52240
+AT5G09390
+AT5G05730
+AT4G34740
+AT2G26490
+AT1G51110
+AT1G12570
+AT3G20790
+AT5G25540
+AT1G16880
+AT2G20610
+AT1G29140
+AT4G37800
+AT5G39500
+AT4G38710
+AT1G12050
+AT1G11750
+AT4G14110
+AT5G56640
+AT1G67730
+AT1G11890
+AT3G46000
+AT2G23080
+AT1G53280
+AT1G15290
+AT1G79560
+AT3G18580
+AT5G22510
+AT5G62350
+AT2G20450
+AT3G10700
+AT4G22670
+AT5G48810
+AT1G16240
+AT1G31850
+AT1G71230
+AT3G54240
+AT3G20330
+AT5G19940
+AT4G04950
+AT1G15415
+AT5G58590
+AT5G05170
+AT4G39520
+AT1G04600
+AT3G05910
+AT5G16450
+AT3G61110
+AT1G30300
+AT4G34830
+AT2G47250
+AT3G47810
+AT1G10590
+AT4G14870
+AT5G05520
+AT4G20020
+AT5G14170
+AT4G04720
+AT2G02740
+AT2G15400
+AT3G13222
+AT1G61870
+AT3G49240
+AT1G48920
+AT3G02750
+AT1G72230
+AT5G53530
+AT5G60960
+AT3G22230
+AT2G38730
+AT4G32150
+AT4G00620
+AT1G17070
+AT1G56050
+AT4G02520
+AT3G14990
+AT2G20270
+AT1G73230
+AT2G44650
+AT1G07040
+AT2G19670
+AT5G58060
+AT1G22450
+AT2G27200
+AT3G09900
+AT1G04450
+AT5G47880
+AT4G23680
+AT5G12860
+AT1G28306
+AT2G30740
+AT4G17040
+AT4G12800
+AT5G51280
+AT1G01610
+AT1G66070
+AT4G00740
+AT3G54700
+AT5G66120
+AT2G37500
+AT5G54270
+AT1G14010
+AT4G12590
+AT2G37010
+AT5G60730
+AT5G23390
+AT1G08130
+AT5G14580
+AT5G51830
+AT2G03480
+AT1G75950
+AT4G36480
+AT1G15440
+AT3G12160
+AT2G32230
+AT5G57300
+AT4G23600
+AT1G20140
+AT2G40620
+AT5G40420
+AT4G15640
+AT5G60860
+AT2G17265
+AT1G44446
+AT5G60640
+AT1G09630
+AT1G49970
+AT5G23940
+AT3G56650
+AT3G10850
+AT1G15950
+AT4G26430
+AT5G59710
+AT4G20720
+AT1G08830
+AT2G45140
+AT5G61840
+AT3G47450
+AT2G38550
+AT1G72810
+AT2G29690
+AT4G16060
+AT5G58710
+AT3G45400
+AT5G54500
+AT5G32470
+AT2G36660
+AT2G15880
+AT5G67320
+AT2G40360
+AT3G01800
+AT3G03100
+AT1G48620
+AT1G77120
+AT4G29830
+AT2G20530
+AT4G13230
+AT3G59890
+AT3G60245
+AT1G01910
+AT3G22360
+AT1G60710
+AT5G59240
+AT5G35700
+AT2G43950
+AT5G19680
+AT4G03080
+AT1G16920
+AT2G32080
+AT5G63220
+AT5G14060
+AT1G49600
+AT2G42220
+AT2G25870
+AT3G12600
+AT5G36890
+AT4G18970
+AT2G30860
+AT2G42790
+AT5G41700
+AT3G19590
+AT3G17900
+AT4G37200
+AT2G37790
+AT5G59160
+AT2G03680
+AT2G40290
+AT5G13280
+AT5G05970
+AT2G37690
+AT3G58840
+AT3G11270
+AT3G52950
+AT2G44760
+AT1G23780
+AT2G29080
+AT5G01750
+AT3G22650
+AT1G01780
+AT2G28080
+AT5G08380
+AT5G55160
+AT1G72710
+AT1G22740
+AT1G21720
+AT3G63095
+AT5G64040
+AT5G66720
+AT1G26690
+AT5G18700
+AT5G10160
+AT2G34560
+AT1G03160
+AT2G22480
+AT2G21620
+AT5G05850
+AT5G67510
+AT5G43470
+AT3G06850
+AT3G45640
+AT4G32410
+AT3G08910
+AT1G48130
+AT5G39870
+AT3G61470
+AT3G23490
+AT1G21660
+AT4G01395
+AT3G04940
+AT5G51570
+AT2G24765
+AT4G09020
+AT4G12400
+AT3G13410
+AT2G46000
+AT4G28450
+AT1G04040
+AT4G03550
+AT5G27560
+AT5G63490
+AT1G21690
+AT3G52610
+AT5G14200
+AT2G19000
+AT3G48890
+AT3G58560
+AT1G79850
+AT5G52820
+AT5G51260
+AT2G29960
+AT4G24510
+AT5G59950
+AT1G12070
+AT3G25770
+AT4G18800
+AT4G03560
+AT2G46170
+AT5G67490
+AT4G28660
+AT2G20060
+AT5G57440
+AT1G51570
+AT1G29940
+AT4G34660
+AT2G02400
+AT5G24490
+AT3G56290
+AT3G09500
+AT5G18400
+AT4G14950
+AT1G53070
+AT4G05400
+AT3G60210
+AT5G46860
+AT3G45770
+AT1G42550
+AT4G20960
+AT3G60740
+AT3G01570
+AT1G75220
+AT1G31800
+AT3G47800
+AT3G16400
+AT4G38130
+AT4G02510
+AT4G08520
+AT3G18600
+AT3G07330
+AT3G21280
+AT2G31170
+AT3G22630
+AT3G62720
+AT3G20290
+AT3G54840
+AT4G39890
+AT1G27390
+AT3G05190
+AT3G56490
+AT2G45440
+AT5G49880
+AT1G76090
+AT4G01150
+AT5G12290
+AT5G62440
+AT1G48020
+AT5G49930
+AT1G63680
+AT1G02130
+AT1G06700
+AT3G26370
+AT2G20890
+AT5G45950
+AT5G04830
+AT2G29450
+AT1G20440
+AT2G01540
+AT1G10390
+AT3G59990
+AT3G08710
+AT1G05510
+AT4G13370
+AT3G25070
+AT4G26570
+AT5G59300
+AT2G39420
+AT4G21710
+AT3G62370
+AT1G21130
+AT4G04020
+AT4G34050
+AT5G09890
+AT2G30870
+AT4G16450
+AT3G02710
+AT1G69840
+AT3G12390
+AT4G39710
+AT4G00400
+AT1G76180
+AT4G18060
+AT3G23820
+AT4G30890
+AT5G24165
+AT4G03110
+AT2G43980
+AT1G73050
+AT1G55090
+AT1G65980
+AT1G02140
+AT3G23620
+AT1G13690
+AT2G20810
+AT1G67360
+AT5G38435
+AT1G25260
+AT4G00170
+AT4G29350
+AT3G10660
+AT4G27160
+AT5G64670
+AT3G50360
+AT5G54310
+AT3G45980
+AT2G38540
+AT3G53020
+AT4G27270
+AT1G09930
+AT1G05270
+AT2G32810
+AT5G46180
+AT5G02890
+AT5G39650
+AT1G07320
+AT4G29520
+AT3G46010
+AT1G30470
+AT1G63780
+AT1G03680
+AT1G70540
+AT2G43420
+AT3G58460
+AT5G40760
+AT5G17620
+AT4G01900
+AT4G24770
+AT1G48480
+AT2G29700
+AT5G62750
+AT5G05620
+AT2G32260
+AT2G39390
+AT1G60000
+AT4G12250
+AT5G41770
+AT5G62550
+AT3G22060
+AT4G39990
+AT3G09580
+AT4G17300
+AT5G47870
+AT5G04290
+AT1G29530
+AT4G39230
+AT5G22100
+AT2G19830
+AT1G72340
+AT5G57810
+AT5G15680
+AT5G22830
+AT1G72610
+AT5G39850
+AT1G32580
+AT5G47760
+AT1G74210
+AT2G35490
+AT3G03570
+AT1G13110
+AT1G44790
+AT1G18850
+AT5G64990
+AT3G01680
+AT1G02305
+AT5G54910
+AT3G18750
+AT5G02610
+AT2G36620
+AT4G36630
+AT1G73720
+AT5G42190
+AT1G18190
+AT5G67385
+AT1G15390
+AT4G39120
+AT2G47640
+AT1G08220
+AT2G15270
+AT2G39840
+AT1G71260
+AT4G30840
+AT1G11930
+AT3G53990
+AT1G22275
+AT5G02960
+AT1G77140
+AT1G36730
+AT1G47290
+AT1G72640
+AT1G19150
+AT2G38740
+AT4G25140
+AT1G30530
+AT1G54530
+AT1G04640
+AT3G12130
+AT3G60540
+AT1G21480
+AT1G52930
+AT4G28780
+AT3G60830
+AT3G10360
+AT3G56240
+AT5G62930
+AT5G58220
+AT4G14080
+AT1G55900
+AT1G11260
+AT5G44780
+AT1G29250
+AT3G13940
+AT2G47240
+AT1G43860
+AT2G46820
+AT1G02090
+AT5G27840
+AT1G54500
+AT4G02840
+AT4G00752
+AT2G34630
+AT2G30620
+AT1G09700
+AT1G76200
+AT4G16360
+AT1G54520
+AT4G27800
+AT3G12170
+AT3G48460
+AT5G02940
+AT4G17510
+AT3G12640
+AT4G17260
+AT1G31440
+AT2G18040
+AT3G26740
+AT4G24490
+AT5G39590
+AT3G23830
+AT1G53900
+AT3G30841
+AT3G56080
+AT4G21110
+AT5G44500
+AT2G46100
+AT3G57050
+AT1G71440
+AT1G80790
+AT2G22360
+AT5G04600
+AT5G04900
+AT2G40190
+AT5G11810
+AT2G17700
+AT5G11440
+AT5G04510
+AT4G13670
+AT4G02530
+AT3G04780
+AT2G45640
+AT2G38270
+AT1G48830
+AT1G49750
+AT4G27170
+AT3G16470
+AT1G69620
+AT5G44610
+AT5G50320
+AT1G43130
+AT3G22500
+AT2G26210
+AT3G02630
+AT5G47600
+AT5G14240
+AT4G30580
+AT3G25220
+AT4G34120
+AT2G25840
+AT1G15910
+AT5G52190
+AT1G06550
+AT3G51130
+AT5G52780
+AT1G22840
+AT1G32160
+AT3G10350
+AT4G30690
+AT1G74070
+AT4G23640
+AT3G55920
+AT4G32130
+AT1G22270
+AT2G24940
+AT1G71680
+AT3G02880
+AT3G03630
+AT1G08490
+AT5G39493
+AT4G34620
+AT3G18220
+AT4G31420
+AT5G62200
+AT3G22320
+AT1G65290
+AT5G45550
+AT4G01990
+AT5G20140
+AT1G11480
+AT3G54640
+AT3G50270
+AT3G05000
+AT3G61650
+AT1G67350
+AT2G01650
+AT2G22230
+AT2G46580
+AT5G52440
+AT1G32080
+AT5G02100
+AT2G35010
+AT1G09590
+AT1G09690
+AT1G69340
+AT1G34000
+AT5G51600
+AT2G30880
+AT2G43130
+AT1G13030
+AT3G15010
+AT3G17465
+AT2G05830
+AT3G02770
+AT5G67270
+AT1G23860
+AT5G58470
+AT4G05090
+AT5G66750
+AT3G04880
+AT5G06830
+AT4G39220
+AT4G23570
+AT1G27190
+AT2G19080
+AT3G06550
+AT4G05390
+AT3G52390
+AT3G57350
+AT1G32550
+AT2G21150
+AT3G26380
+AT1G22150
+AT5G54740
+AT4G00860
+AT4G01610
+AT1G06400
+AT5G67590
+AT3G49250
+AT5G16760
+AT2G43235
+AT2G44540
+AT2G41600
+AT3G43520
+AT5G42980
+AT2G44310
+AT3G59010
+AT4G18465
+AT2G47330
+AT5G51750
+AT1G26880
+AT5G64350
+AT3G58480
+AT1G15710
+AT2G02850
+AT3G06790
+AT2G17972
+AT1G18170
+AT3G09300
+AT3G21400
+AT1G27650
+AT5G15470
+AT5G15870
+AT3G19130
+AT4G11820
+AT5G18110
+AT2G25830
+AT1G23140
+AT3G01250
+AT4G32470
+AT1G70330
+AT5G20250
+AT5G25060
+AT5G22740
+AT4G28200
+AT4G21770
+AT2G33410
+AT4G12770
+AT2G05520
+AT1G24460
+AT1G18480
+AT4G27780
+AT4G27230
+AT3G05100
+AT1G28290
+AT2G42130
+AT5G06910
+AT1G66130
+AT4G35230
+AT1G08700
+AT5G22030
+AT1G70790
+AT5G05270
+AT4G23920
+AT3G51420
+AT1G02410
+AT3G01520
+AT1G31070
+AT5G67530
+AT3G11730
+AT5G16390
+AT5G12410
+AT5G13610
+AT4G39150
+AT5G61760
+AT4G17560
+AT1G80360
+AT2G06530
+AT4G20830
+AT3G58580
+AT5G02160
+AT1G50900
+AT1G77840
+AT1G35220
+AT4G32760
+AT3G45300
+AT4G02140
+AT3G06730
+AT2G39930
+AT1G49820
+AT1G20580
+AT1G08640
+AT4G02340
+AT1G79870
+AT5G42820
+AT1G63160
+AT1G07250
+AT1G71310
+AT1G12910
+AT3G59650
+AT3G28630
+AT1G51100
+AT2G47580
+AT5G23040
+AT5G46940
+AT2G31970
+AT1G27130
+AT2G45740
+AT5G64960
+AT2G42120
+AT5G08330
+AT2G47710
+AT3G48250
+AT3G25585
+AT4G26720
+AT3G02180
+AT1G24610
+AT5G25265
+AT1G24050
+AT1G76680
+AT1G31660
+AT2G39010
+AT1G06820
+AT2G25280
+AT1G64750
+AT2G19330
+AT3G28150
+AT2G24450
+AT3G60810
+AT4G25840
+AT2G16640
+AT5G18200
+AT1G10522
+AT1G04680
+AT5G14530
+AT5G23535
+AT4G31140
+AT2G29340
+AT2G20840
+AT4G28360
+AT4G35610
+AT3G15790
+AT1G09300
+AT2G28120
+AT4G03430
+AT2G34590
+AT4G38690
+AT1G65150
+AT4G39670
+AT2G39140
+AT1G50670
+AT3G03120
+AT5G48030
+AT5G47520
+AT1G80750
+AT3G16170
+AT3G06610
+AT3G48820
+AT1G73980
+AT3G29075
+AT1G10500
+AT1G08520
+AT1G60550
+AT1G02170
+AT4G28700
+AT3G29200
+AT1G18630
+AT4G17870
+AT1G18640
+AT4G39730
+AT2G21960
+AT1G07645
+AT2G18030
+AT1G70670
+AT3G23390
+AT4G14320
+AT1G19880
+AT1G07170
+AT2G30000
+AT2G25800
+AT1G26550
+AT1G07330
+AT2G42680
+AT1G58360
+AT2G33070
+AT5G40580
+AT1G77750
+AT3G44100
+AT2G16370
+AT3G01510
+AT3G11820
+AT4G17650
+AT1G18990
+AT5G18850
+AT5G15770
+AT2G37470
+AT2G40170
+AT3G13120
+AT5G12210
+AT2G06510
+AT5G63380
+AT3G27700
+AT2G45000
+AT1G10310
+AT1G32330
+AT1G13330
+AT5G50960
+AT3G47470
+AT4G31350
+AT4G36810
+AT4G15130
+AT3G06430
+AT5G65650
+AT5G14910
+AT3G55260
+AT4G00300
+AT1G77440
+AT5G66100
+AT2G02120
+AT2G19740
+AT2G41040
+AT1G50240
+AT1G13090
+AT1G26761
+AT1G29810
+AT2G43400
+AT5G03280
+AT3G27020
+AT2G36835
+AT1G08410
+AT1G04110
+AT3G15660
+AT1G63500
+AT2G26930
+AT5G19150
+AT2G35800
+AT1G03900
+AT5G18480
+AT5G64370
+AT3G51010
+AT3G54300
+AT1G05460
+AT2G35100
+AT4G10040
+AT3G08010
+AT2G34840
+AT1G74310
+AT5G64290
+AT5G03050
+AT5G58870
+AT5G55130
+AT3G06540
+AT2G30930
+AT5G27450
+AT2G20930
+AT3G01590
+AT1G11680
+AT1G12640
+AT5G47320
+AT5G56710
+AT2G44230
+AT3G07320
+AT2G31960
+AT1G54320
+AT2G41490
+AT4G38200
+AT1G67440
+AT1G01370
+AT3G22480
+AT4G33350
+AT4G04320
+AT1G28060
+AT5G04870
+AT1G76690
+AT1G07960
+AT4G29810
+AT5G19460
+AT4G23430
+AT5G33320
+AT1G51160
+AT5G09420
+AT1G51590
+AT2G35410
+AT3G10940
+AT1G05850
+AT2G21470
+AT4G08790
+AT3G02900
+AT1G56580
+AT1G54570
+AT1G22620
+AT2G43020
+AT5G58260
+AT3G21220
+AT4G24530
+AT4G13050
+AT5G27670
+AT1G50510
+AT2G15900
+AT4G25340
+AT1G67540
+AT1G79720
+AT4G02290
+AT5G58240
+AT1G27435
+AT3G61560
+AT4G00585
+AT5G38880
+AT2G16460
+AT5G03910
+AT3G18590
+AT4G16700
+AT2G13290
+AT4G31930
+AT1G14380
+AT3G53700
+AT3G25545
+AT1G24095
+AT3G61070
+AT3G54690
+AT2G15790
+AT5G58030
+AT2G24490
+AT2G34700
+AT3G27890
+AT3G02730
+AT3G16310
+AT4G16800
+AT5G50100
+AT5G10780
+AT4G30610
+AT3G54560
+AT1G29700
+AT5G12190
+AT5G52210
+AT1G07440
+AT1G23100
+AT4G35760
+AT1G47380
+AT2G21600
+AT5G42570
+AT5G15530
+AT5G47780
+AT5G27700
+AT3G08950
+AT1G17350
+AT4G26760
+AT2G31305
+AT1G73250
+AT3G18570
+AT1G06130
+AT4G34180
+AT1G65650
+AT5G30495
+AT2G44050
+AT3G09880
+AT3G16410
+AT4G32400
+AT2G32090
+AT1G08750
+AT4G33625
+AT2G43520
+AT3G50520
+AT5G08740
+AT5G09260
+AT5G37590
+AT3G01380
+AT1G17745
+AT1G77090
+AT3G27120
+AT3G53890
+AT2G07725
+AT2G46560
+AT2G24180
+AT3G12380
+AT1G66900
+AT3G22845
+AT4G22000
+AT4G15930
+AT3G26600
+AT3G42790
+AT4G36580
+AT3G51550
+AT2G33620
+AT3G18940
+AT5G24260
+AT5G51010
+AT3G55180
+AT3G15710
+AT5G10090
+AT1G60650
+AT2G36020
+AT5G18430
+AT3G17910
+AT2G29290
+AT4G21445
+AT1G75660
+AT2G31410
+AT5G26030
+AT2G17010
+AT1G13730
+AT2G17410
+AT4G15520
+AT4G31860
+AT1G72460
+AT1G78870
+AT5G12980
+AT2G46900
+AT2G20120
+AT1G20560
+AT2G40100
+AT5G56760
+AT1G77420
+AT3G26860
+AT2G16280
+AT2G30390
+AT3G01290
+AT3G28900
+AT1G50710
+AT5G46250
+AT5G07510
+AT5G59810
+AT5G41600
+AT2G35605
+AT3G52720
+AT3G02920
+AT1G08780
+AT2G48120
+AT4G00380
+AT3G12140
+AT2G03640
+AT3G08600
+AT1G11400
+AT3G60360
+AT1G14900
+AT3G20550
+AT5G40200
+AT1G29590
+AT3G60340
+AT1G78240
+AT3G45280
+AT5G15350
+AT3G25290
+AT2G26460
+AT1G12970
+AT5G47390
+AT4G11240
+AT5G20520
+AT5G17590
+AT4G15940
+AT5G20590
+AT5G22220
+AT3G05030
+AT3G21370
+AT1G08660
+AT3G02220
+AT2G17560
+AT3G28730
+AT5G39800
+AT2G01350
+AT3G12010
+AT5G08180
+AT3G54890
+AT3G05130
+AT1G68310
+AT4G28980
+AT1G51510
+AT1G29060
+AT1G73620
+AT1G26940
+AT3G02320
+AT5G22470
+AT3G09970
+AT4G21940
+AT3G51510
+AT3G12760
+AT4G29470
+AT1G47820
+AT1G73030
+AT5G10560
+AT1G16890
+AT5G58110
+AT3G28700
+AT2G44200
+AT5G54110
+AT2G21160
+AT4G24805
+AT4G34920
+AT3G50960
+AT5G50930
+AT1G35470
+AT4G22310
+AT4G27140
+AT2G39900
+AT5G16010
+AT1G04645
+AT3G25680
+AT3G13640
+AT3G13230
+AT1G74560
+AT1G17050
+AT3G27050
+AT1G74320
+AT2G31750
+AT5G37770
+AT4G26750
+AT4G26470
+AT4G15550
+AT3G12410
+AT3G27660
+AT4G31130
+AT5G06530
+AT5G63830
+AT3G08630
+AT1G20850
+AT1G67830
+AT3G17410
+AT3G62130
+AT3G14920
+AT1G52260
+AT1G17270
+AT1G75460
+AT2G36830
+AT3G09570
+AT2G35830
+AT1G04900
+AT1G68020
+AT3G19770
+AT4G28860
+AT2G44860
+AT3G16060
+AT5G43810
+AT1G65090
+AT3G03590
+AT2G15690
+AT1G60950
+AT5G45010
+AT2G33600
+AT3G12020
+AT3G04290
+AT1G16280
+AT4G01130
+AT3G03600
+AT5G49720
+AT3G12260
+AT3G10330
+AT2G26910
+AT1G01820
+AT1G11475
+AT1G10200
+AT2G38910
+AT5G60620
+AT5G41940
+AT1G03050
+AT1G79090
+AT2G36310
+AT5G08050
+AT2G16930
+AT5G50870
+AT1G17210
+AT5G43500
+AT2G42210
+AT5G35660
+AT1G49430
+AT4G10120
+AT2G02040
+AT2G32900
+AT3G07640
+AT3G15820
+AT3G11740
+AT1G75850
+AT1G50400
+AT4G38350
+AT4G24930
+AT3G47940
+AT1G80500
+AT3G10530
+AT1G14510
+AT5G67560
+AT5G27520
+AT5G56260
+AT5G66410
+AT1G04650
+AT3G57530
+AT5G37810
+AT5G25530
+AT2G38710
+AT1G51660
+AT5G14310
+AT4G34260
+AT5G65900
+AT3G07480
+AT4G30010
+AT1G11765
+AT5G04280
+AT4G32960
+AT2G28355
+AT1G45145
+AT5G49510
+AT4G36800
+AT2G31035
+AT3G25480
+AT2G30710
+AT3G57220
+AT1G29260
+AT1G32250
+AT5G11480
+AT5G24460
+AT5G11680
+AT1G78510
+AT2G18700
+AT5G47630
+AT4G25050
+AT3G11070
+AT4G00560
+AT4G10750
+AT5G61310
+AT5G66090
+AT5G40480
+AT5G63910
+AT4G31200
+AT2G26000
+AT3G60500
+AT4G39910
+AT5G07860
+AT1G17840
+AT2G35340
+AT2G24090
+AT4G21895
+AT2G41950
+AT4G19003
+AT1G49480
+AT5G39950
+AT5G25900
+AT3G18410
+AT1G30890
+AT2G01490
+AT1G29850
+AT4G16444
+AT4G15955
+AT1G35340
+AT3G18500
+AT4G04850
+AT4G17920
+AT4G24860
+AT4G29960
+AT5G56950
+AT2G44690
+AT3G16090
+AT1G17190
+AT5G24130
+AT1G72170
+AT5G16400
+AT5G47435
+AT1G70520
+AT1G21850
+AT4G22140
+AT2G23610
+AT1G74730
+AT2G18600
+AT1G17540
+AT2G18390
+AT1G16740
+AT3G50980
+AT4G32120
+AT4G17760
+AT4G17770
+AT5G52110
+AT1G68590
+AT5G66055
+AT3G02480
+AT1G03870
+AT3G09980
+AT4G27560
+AT2G02160
+AT5G48335
+AT3G47610
+AT4G17420
+AT1G06230
+AT4G27130
+AT5G50580
+AT5G50680
+AT5G56660
+AT4G37270
+AT1G62010
+AT1G77390
+AT3G23580
+AT2G40400
+AT1G13190
+AT4G13510
+AT2G48150
+AT1G80910
+AT4G22570
+AT5G46230
+AT5G59845
+AT5G47890
+AT5G26600
+AT2G33775
+AT3G54490
+AT5G63800
+AT2G26430
+AT3G21820
+AT1G13870
+AT2G40890
+AT5G57950
+AT1G54410
+AT5G22950
+AT5G02280
+AT3G21670
+AT1G52300
+AT4G11980
+AT4G32070
+AT5G42130
+AT3G57280
+AT3G02110
+AT3G10220
+AT1G19360
+AT4G02620
+AT3G28490
+AT1G21190
+AT2G22570
+AT2G24310
+AT4G16990
+AT2G23940
+AT3G11770
+AT3G01440
+AT5G48540
+AT1G24490
+AT1G21600
+AT1G03150
+AT3G28460
+AT2G22900
+AT1G53840
+AT3G61980
+AT3G19810
+AT4G30780
+AT3G01450
+AT5G22370
+AT4G25370
+AT2G26100
+AT1G24120
+AT1G80510
+AT5G55140
+AT1G26170
+AT1G04970
+AT2G33830
+AT1G51610
+AT1G73940
+AT3G46230
+AT4G24130
+AT1G44820
+AT2G41670
+AT3G45010
+AT3G18760
+AT2G22610
+AT2G39850
+AT5G37130
+AT3G56010
+AT4G33670
+AT5G04910
+AT4G29680
+AT1G06670
+AT5G11780
+AT2G44790
+AT4G08230
+AT1G62600
+AT3G01740
+AT1G08860
+AT1G24267
+AT5G24350
+AT3G03490
+AT1G11040
+AT5G63870
+AT4G25440
+AT1G31860
+AT2G07727
+AT1G03400
+AT5G51150
+AT1G48170
+AT3G24320
+AT1G50450
+AT5G13020
+AT2G36130
+AT2G03630
+AT5G03440
+AT5G20040
+AT3G15850
+AT1G08160
+AT2G30530
+AT5G23395
+AT2G19790
+AT5G17250
+AT4G29870
+AT3G27670
+AT4G33640
+AT1G11572
+AT1G52780
+AT1G28340
+AT1G43900
+AT5G54390
+AT3G26030
+AT5G39080
+AT2G19385
+AT1G12140
+AT3G52730
+AT5G59140
+AT2G19750
+AT4G29390
+AT5G56670
+AT1G62850
+AT5G62310
+AT1G48440
+AT1G30480
+AT3G04950
+AT1G65320
+AT4G16440
+AT5G09250
+AT5G12970
+AT1G30460
+AT4G17340
+AT2G16860
+AT2G41475
+AT5G07430
+AT5G45810
+AT1G47830
+AT2G16505
+AT1G80350
+AT4G12640
+AT3G14960
+AT1G13820
+AT5G24314
+AT1G20840
+AT1G68540
+AT4G26455
+AT2G03980
+AT2G34640
+AT2G02760
+AT5G05370
+AT3G61620
+AT5G20090
+AT5G47810
+AT1G25480
+AT5G03080
+AT2G44730
+AT3G16980
+AT3G54630
+AT4G21590
+AT4G34510
+AT4G27180
+AT5G25310
+AT1G78140
+AT5G57550
+AT2G35320
+AT1G27760
+AT5G52960
+AT1G54070
+AT5G16150
+AT3G04020
+AT4G31770
+AT2G10950
+AT1G52530
+AT3G60730
+AT4G25100
+AT3G21055
+AT2G31570
+AT1G02830
+AT5G51720
+AT5G20930
+AT3G21380
+AT3G18140
+AT1G76260
+AT1G16380
+AT5G15880
+AT5G67470
+AT5G15802
+AT1G76405
+AT2G47120
+AT1G72010
+AT5G41880
+AT3G58680
+AT4G08350
+AT4G38500
+AT1G48270
+AT5G49550
+AT2G46090
+AT4G17486
+AT1G65410
+AT5G15860
+AT2G01460
+AT5G47030
+AT3G27080
+AT1G03550
+AT1G15700
+AT3G12280
+AT5G23670
+AT5G07460
+AT5G54010
+AT3G49670
+AT2G27385
+AT3G01660
+AT2G46800
+AT3G49100
+AT4G34940
+AT2G26280
+AT5G40260
+AT5G54580
+AT1G17330
+AT2G38010
+AT5G10300
+AT4G25680
+AT3G25170
+AT5G64090
+AT5G58800
+AT5G38980
+AT5G62410
+AT3G13224
+AT1G20270
+AT5G43970
+AT5G06660
+AT5G55470
+AT1G24100
+AT4G36100
+AT3G15380
+AT5G54800
+AT2G19450
+AT5G49570
+AT5G46020
+AT5G59870
+AT5G09640
+AT2G28600
+AT3G11200
+AT1G76490
+AT3G17780
+AT2G21640
+AT5G64140
+AT1G49140
+AT5G18620
+AT3G05070
+AT5G65520
+AT5G64540
+AT1G26220
+AT2G03820
+AT1G67840
+AT2G17720
+AT4G05450
+AT1G17520
+AT2G26900
+AT3G60960
+AT1G78180
+AT5G48840
+AT3G12930
+AT2G47420
+AT5G62290
+AT1G60170
+AT3G07010
+AT2G14530
+AT5G14050
+AT2G15240
+AT1G28270
+AT1G51540
+AT1G48240
+AT1G70710
+AT3G25165
+AT1G79490
+AT3G62550
+AT2G27260
+AT4G05150
+AT3G18050
+AT3G11750
+AT1G49330
+AT3G25400
+AT3G07070
+AT1G52980
+AT5G51130
+AT1G51770
+AT4G14145
+AT4G09160
+AT3G22930
+AT1G11390
+AT4G36720
+AT1G51400
+AT5G23440
+AT3G55830
+AT2G34810
+AT1G64710
+AT4G31160
+AT5G03470
+AT1G74530
+AT2G41430
+AT4G01560
+AT5G14010
+AT1G50610
+AT4G33400
+AT1G31220
+AT1G11530
+AT3G51600
+AT5G56150
+AT5G21100
+AT3G14090
+AT1G11760
+AT1G30960
+AT5G27720
+AT2G18630
+AT1G73190
+AT5G15050
+AT2G42800
+AT4G24970
+AT5G50030
+AT4G17010
+AT1G15270
+AT1G22460
+AT3G45050
+AT1G33680
+AT4G27450
+AT5G41685
+AT5G57980
+AT4G19040
+AT2G26680
+AT3G28450
+AT1G26670
+AT3G08640
+AT5G13310
+AT4G02610
+AT1G50320
+AT3G55270
+AT1G77930
+AT5G12130
+AT4G26770
+AT5G55980
+AT1G67230
+AT5G11760
+AT5G55920
+AT2G18170
+AT1G63310
+AT2G01060
+AT2G20700
+AT3G56720
+AT5G17790
+AT1G02880
+AT3G18430
+AT1G24110
+AT5G39790
+AT3G15110
+AT1G78670
+AT2G26570
+AT3G01370
+AT1G77860
+AT2G42400
+AT5G58090
+AT1G22170
+AT3G13080
+AT2G36180
+AT2G42070
+AT1G67460
+AT4G20330
+AT1G78340
+AT1G54830
+AT1G70220
+AT5G20650
+AT4G33060
+AT1G43620
+AT1G70890
+AT2G43240
+AT2G39970
+AT4G26020
+AT5G07560
+AT1G16860
+AT5G51640
+AT3G15810
+AT5G19980
+AT5G40740
+AT3G27000
+AT4G34020
+AT5G12390
+AT2G01070
+AT3G05410
+AT4G26190
+AT5G16120
+AT4G04670
+AT5G65770
+AT5G48470
+AT1G15240
+AT3G59140
+AT4G14270
+AT1G27460
+AT5G10050
+AT5G11280
+AT1G47750
+AT5G38110
+AT1G74740
+AT5G12030
+AT3G02950
+AT1G52630
+AT5G59250
+AT1G01470
+AT2G33700
+AT3G60620
+AT5G63140
+AT1G43700
+AT5G47100
+AT4G37760
+AT3G63160
+AT5G67290
+AT1G02335
+AT1G51650
+AT1G07970
+AT1G11780
+AT4G27420
+AT5G64330
+AT2G16920
+AT1G29990
+AT5G05070
+AT2G26300
+AT1G68000
+AT5G01700
+AT3G61600
+AT1G64170
+AT3G16990
+AT1G65700
+AT4G32390
+AT4G31850
+AT5G45490
+AT4G37880
+AT1G10230
+AT3G12620
+AT1G50520
+AT5G03070
+AT1G04010
+AT4G32714
+AT5G36950
+AT1G78020
+AT1G62930
+AT1G54610
+AT5G53370
+AT1G65010
+AT4G04570
+AT1G79820
+AT1G29790
+AT3G29280
+AT5G17160
+AT1G71170
+AT5G39530
+AT3G57420
+AT5G13740
+AT4G37330
+AT5G58430
+AT4G13630
+AT1G76990
+AT1G15720
+AT1G01970
+AT4G04925
+AT4G15010
+AT5G14140
+AT3G12030
+AT5G56360
+AT4G13800
+AT5G02050
+AT1G55360
+AT3G56700
+AT1G17140
+AT5G14210
+AT4G06599
+AT5G46570
+AT3G52210
+AT4G21960
+AT1G69380
+AT5G56090
+AT4G34138
+AT5G50900
+AT3G01650
+AT2G29730
+AT5G14450
+AT1G16010
+AT4G21540
+AT3G15356
+AT5G42670
+AT4G16650
+AT2G18730
+AT5G03860
+AT1G76670
+AT1G65840
+AT2G25880
+AT3G07570
+AT5G12100
+AT3G10260
+AT1G71040
+AT1G64680
+AT4G38380
+AT1G17130
+AT3G47430
+AT4G38400
+AT1G08030
+AT3G05020
+AT2G26660
+AT1G03250
+AT1G18650
+AT3G46510
+AT5G64910
+AT3G54870
+AT5G47640
+AT2G04550
+AT5G13970
+AT5G20130
+AT1G61630
+AT4G26000
+AT1G78280
+AT4G33430
+AT5G05660
+AT4G29380
+AT5G06340
+AT1G79350
+AT5G23820
+AT3G47833
+AT1G45688
+AT5G22050
+AT2G15820
+AT3G54960
+AT1G23440
+AT3G26990
+AT1G67900
+AT3G09250
+AT3G03990
+AT3G07390
+AT4G27640
+AT5G46390
+AT2G35470
+AT3G54080
+AT2G01590
+AT3G55520
+AT3G19800
+AT1G09160
+AT3G16175
+AT5G59730
+AT3G55960
+AT3G51450
+AT2G40980
+AT1G22610
+AT3G11620
+AT2G26380
+AT1G34340
+AT4G31840
+AT2G31800
+AT5G02830
+AT2G26970
+AT5G42870
+AT5G03730
+AT1G58060
+AT1G59760
+AT4G20280
+AT1G27535
+AT3G03040
+AT2G03670
+AT2G20280
+AT1G22690
+AT4G01700
+AT1G57990
+AT3G17030
+AT2G35350
+AT1G48760
+AT2G35960
+AT3G61580
+AT1G68910
+AT5G39990
+AT4G29650
+AT2G38760
+AT4G01920
+AT5G09920
+AT1G48880
+AT5G42340
+AT1G47640
+AT1G03910
+AT5G04540
+AT1G63290
+AT3G08850
+AT3G03160
+AT5G41010
+AT1G01580
+AT5G64620
+AT3G42860
+AT3G60840
+AT5G47020
+AT2G32540
+AT1G01730
+AT2G47600
+AT3G49590
+AT4G24730
+AT1G47240
+AT3G58170
+AT1G54850
+AT4G10280
+AT5G60190
+AT1G55805
+AT2G41160
+AT4G30400
+AT4G15820
+AT5G07770
+AT5G16890
+AT1G10280
+AT1G28670
+AT1G73410
+AT5G32440
+AT3G16565
+AT1G51730
+AT2G23560
+AT3G58180
+AT2G38530
+AT3G23190
+AT2G41900
+AT5G05760
+AT5G40090
+AT2G30150
+AT5G42890
+AT2G36300
+AT5G27395
+AT3G21630
+AT3G12610
+AT5G48870
+AT1G62960
+AT2G22370
+AT2G24390
+AT2G03290
+AT1G13790
+AT3G08490
+AT2G47900
+AT4G08320
+AT1G49740
+AT1G07840
+AT2G01680
+AT1G15540
+AT1G72820
+AT3G04660
+AT2G45520
+AT5G51020
+AT1G69430
+AT4G29240
+AT4G37630
+AT5G38670
+AT2G22600
+AT2G04360
+AT3G53170
+AT1G33900
+AT2G47980
+AT5G23920
+AT3G52290
+AT1G13600
+AT1G20230
+AT3G01550
+AT4G15070
+AT3G14650
+AT3G55080
+AT1G16520
+AT5G22500
+AT3G46260
+AT5G59320
+AT5G08720
+AT5G55640
+AT3G04430
+AT5G40060
+AT1G01940
+AT1G16640
+AT2G25520
+AT1G43640
+AT3G54950
+AT5G47800
+AT5G18690
+AT1G53550
+AT5G25430
+AT4G08691
+AT1G73100
+AT1G19690
+AT1G45063
+AT2G17120
+AT1G72175
+AT1G75710
+AT2G38440
+AT3G18295
+AT4G21260
+AT1G55290
+AT5G27710
+AT1G12280
+AT1G73080
+AT4G31640
+AT3G50100
+AT5G61700
+AT4G27510
+AT1G59600
+AT5G04070
+AT1G76920
+AT4G29720
+AT1G32750
+AT3G27180
+AT4G37100
+AT5G08490
+AT4G23530
+AT1G24290
+AT2G18850
+AT5G20580
+AT2G28890
+AT2G25770
+AT5G57815
+AT5G11960
+AT2G36360
+AT2G33610

train.py

@@ -13,8 +13,9 @@ from data.loader import DataLoader
 from utils.misc import list_epochs, print_and_log
 
 
-RAYS_PER_BATCH = 2 ** 16
+RAYS_PER_BATCH = 2 ** 12
 DATA_LOADER_CHUNK_SIZE = 1e8
+root_dir = Path.cwd()
 
 
 parser = argparse.ArgumentParser()
@@ -68,7 +69,7 @@ if args.mdl_path:
     model_args = model.args
 else:
     # Create model from specified configuration
-    with Path(f'{sys.path[0]}/configs/{args.config}.json').open() as fp:
+    with Path(f'{root_dir}/configs/{args.config}.json').open() as fp:
         config = json.load(fp)
     model_name = args.config
     model_class = config['model']
@@ -76,7 +77,7 @@ else:
     model_args['bbox'] = dataset.bbox
     model_args['depth_range'] = dataset.depth_range
     model, states = mdl.create(model_class, model_args), None
-model.to(device.default()).train()
+model.to(device.default())
 
 run_dir = Path(f"_nets/{dataset.name}/{model_name}")
 run_dir.mkdir(parents=True, exist_ok=True)

train/__init__.py

@@ -22,5 +22,5 @@ def get_class(class_name: str) -> type:
 
 
 def get_trainer(model: BaseModel, **kwargs) -> base.Train:
-    train_class = get_class(model.trainer)
+    train_class = get_class(model.TrainerClass)
     return train_class(model, **kwargs)

train/base.py

@@ -42,8 +42,9 @@ class Train(object, metaclass=BaseTrainMeta):
         self.iters = 0
         self.run_dir = run_dir
 
+        self.model.trainer = self
         self.model.train()
-        self.optimizer = torch.optim.Adam(self.model.parameters(), lr=5e-4)
+        self.reset_optimizer()
 
         if states:
             if 'epoch' in states:
@@ -58,6 +59,9 @@ class Train(object, metaclass=BaseTrainMeta):
         if self.perf_mode:
             enable_perf()
 
+    def reset_optimizer(self):
+        self.optimizer = torch.optim.Adam(self.model.parameters(), lr=5e-4)
+    
     def train(self, data_loader: DataLoader, max_epochs: int):
         self.data_loader = data_loader
         self.iters_per_epoch = self.perf_frames or len(data_loader)

train/train_with_space.py

@@ -20,18 +20,15 @@ class TrainWithSpace(Train):
                 if self.splitting_loop == 1 or self.epoch % self.splitting_loop == 1:
                     try:
                         with torch.no_grad():
-                            before, after = self.model.splitting()
-                        print_and_log(
-                            f"Splitting done. # of voxels before: {before}, after: {after}")
+                            before, after = self.model.split()
+                        print_and_log(f"Splitting done: {before} -> {after}")
                     except NotImplementedError:
                         print_and_log(
                             "Note: The space does not support splitting operation. Just skip it.")
                 if self.pruning_loop == 1 or self.epoch % self.pruning_loop == 1:
                     try:
                         with torch.no_grad():
-                            #before, after = self.model.pruning()
-                            # print(f"Pruning by voxel densities done. # of voxels before: {before}, after: {after}")
-                            # self._prune_inner_voxels()
+                            # self._prune_voxels_by_densities()
                             self._prune_voxels_by_weights()
                     except NotImplementedError:
                         print_and_log(
@@ -39,26 +36,26 @@ class TrainWithSpace(Train):
 
         super()._train_epoch()
 
-    def _prune_inner_voxels(self):
+    def _prune_voxels_by_densities(self):
         space: Voxels = self.model.space
-        voxel_access_counts = torch.zeros(space.n_voxels, dtype=torch.long,
-                                          device=space.voxels.device)
-        iters_in_epoch = 0
-        batch_size = self.data_loader.batch_size
-        self.data_loader.batch_size = 2 ** 14
-        for _, rays_o, rays_d, _ in self.data_loader:
-            self.model(rays_o, rays_d,
-                       raymarching_early_stop_tolerance=0.01,
-                       raymarching_chunk_size_or_sections=[1],
-                       perturb_sample=False,
-                       voxel_access_counts=voxel_access_counts,
-                       voxel_access_tolerance=0)
-            iters_in_epoch += 1
-            percent = iters_in_epoch / len(self.data_loader) * 100
-            sys.stdout.write(f'Pruning inner voxels...{percent:.1f}%   \r')
-        self.data_loader.batch_size = batch_size
-        before, after = space.prune(voxel_access_counts > 0)
-        print(f"Prune inner voxels: {before} -> {after}")
+        threshold = .5
+        bits = 16
+
+        @torch.no_grad()
+        def get_scores(sampled_points: torch.Tensor, sampled_voxel_indices: torch.Tensor) -> torch.Tensor:
+            densities = self.model.render(
+                Samples(sampled_points, None, None, None, sampled_voxel_indices),
+                'density')
+            return 1 - (-densities).exp()
+
+        sampled_xyz, sampled_idx = space.sample(bits)
+        chunk_size = 64
+        scores = torch.cat([
+            torch.max(get_scores(sampled_xyz[i:i + chunk_size], sampled_idx[i:i + chunk_size])
+                      .reshape(-1, bits ** 3), -1)[0]
+            for i in range(0, self.voxels.size(0), chunk_size)
+        ], 0)  # (M[, ...])
+        return space.prune(scores > threshold)
 
     def _prune_voxels_by_weights(self):
         space: Voxels = self.model.space

utils/misc.py

@@ -57,10 +57,11 @@ def meshgrid(*size: int, normalize: bool = False, swap_dim: bool = False) -> tor
     """
     if len(size) == 1:
         size = (size[0], size[0])
-    y, x = torch.meshgrid(torch.arange(0, size[0]), torch.arange(0, size[1]))
-    if swap_dim:
-        return torch.stack([y / (size[0] - 1.), x / (size[1] - 1.)], 2) if normalize else torch.stack([y, x], 2)
-    return torch.stack([x / (size[1] - 1.), y / (size[0] - 1.)], 2) if normalize else torch.stack([x, y], 2)
+    y, x = torch.meshgrid(torch.arange(size[0]), torch.arange(size[1]), indexing='ij')
+    if normalize:
+        x.div_(size[1] - 1.)
+        y.div_(size[0] - 1.)
+    return torch.stack([y, x], 2) if swap_dim else torch.stack([x, y], 2)
 
 
 def get_angle(x: torch.Tensor, y: torch.Tensor) -> torch.Tensor:

utils/voxels.py

@@ -13,6 +13,13 @@ def get_grid_steps(bbox: torch.Tensor, step_size: Union[torch.Tensor, float]) ->
     return ((bbox[1] - bbox[0]) / step_size).ceil().long()
 
 
+def to_flat_indices(grid_coords: torch.Tensor, steps: torch.Tensor) -> torch.Tensor:
+    indices = grid_coords[..., 0]
+    for i in range(1, grid_coords.shape[-1]):
+        indices = indices * steps[i] + grid_coords[..., i]
+    return indices
+
+
 def to_grid_coords(pts: torch.Tensor, bbox: torch.Tensor, *,
                    step_size: Union[torch.Tensor, float] = None,
                    steps: torch.Tensor = None) -> torch.Tensor:
@@ -55,20 +62,7 @@ def to_grid_indices(pts: torch.Tensor, bbox: torch.Tensor, *,
         steps = get_grid_steps(bbox, step_size)  # (D)
     grid_coords = to_grid_coords(pts, bbox, step_size=step_size, steps=steps)  # (N..., D)
     outside_mask = torch.logical_or(grid_coords < 0, grid_coords >= steps).any(-1)  # (N...)
-    if pts.size(-1) == 1:
-        grid_indices = grid_coords[..., 0]
-    elif pts.size(-1) == 2:
-        grid_indices = grid_coords[..., 0] * steps[1] + grid_coords[..., 1]
-    elif pts.size(-1) == 3:
-        grid_indices = grid_coords[..., 0] * steps[1] * steps[2] \
-            + grid_coords[..., 1] * steps[2] + grid_coords[..., 2]
-    elif pts.size(-1) == 4:
-        grid_indices = grid_coords[..., 0] * steps[1] * steps[2] * steps[3] \
-            + grid_coords[..., 1] * steps[2] * steps[3] \
-            + grid_coords[..., 2] * steps[3] \
-            + grid_coords[..., 3]
-    else:
-        raise NotImplementedError("The function does not support D>4")
+    grid_indices = to_flat_indices(grid_coords, steps)
     return grid_indices, outside_mask
 
 
@@ -76,7 +70,7 @@ def init_voxels(bbox: torch.Tensor, steps: torch.Tensor):
     """
     Initialize voxels.
     """
-    x, y, z = torch.meshgrid(*[torch.arange(steps[i]) for i in range(3)])
+    x, y, z = torch.meshgrid(*[torch.arange(steps[i]) for i in range(3)], indexing="ij")
     return to_voxel_centers(torch.stack([x, y, z], -1).reshape(-1, 3), bbox, steps=steps)
 
 
@@ -96,7 +90,7 @@ def to_voxel_centers(grid_coords: torch.Tensor, bbox: torch.Tensor, *,
     :param steps `Tensor(1|D)`: (optional) steps alone every dim
     :return `Tensor(N..., D)`: discretized grid coordinates
     """
-    grid_coords = grid_coords.float() + 0.5
+    grid_coords = grid_coords.float() + .5
     if step_size is not None:
         return grid_coords * step_size + bbox[0]
     return grid_coords / steps * (bbox[1] - bbox[0]) + bbox[0]
@@ -121,8 +115,8 @@ def split_voxels_local(voxel_size: Union[torch.Tensor, float], n: int, align_bor
         dtype = like.dtype
         device = like.device
     c = torch.arange(1 - n, n, 2, dtype=dtype, device=device)
-    offset = torch.stack(torch.meshgrid([c] * dims), -1).flatten(0, -2) * voxel_size / 2 /\
-        (n - 1 if align_border else n)
+    offset = torch.stack(torch.meshgrid([c] * dims, indexing='ij'), -1).flatten(0, -2)\
+        * voxel_size * .5 / (n - 1 if align_border else n)
     return offset
 
 
@@ -144,7 +138,7 @@ def split_voxels(voxel_centers: torch.Tensor, voxel_size: Union[torch.Tensor, fl
 
 def get_corners(voxel_centers: torch.Tensor, bbox: torch.Tensor, steps: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]:
     half_voxel_size = (bbox[1] - bbox[0]) / steps * 0.5
-    expand_bbox = bbox
+    expand_bbox = bbox.clone()
     expand_bbox[0] -= 0.5 * half_voxel_size
     expand_bbox[1] += 0.5 * half_voxel_size
     double_grid_coords = to_grid_coords(voxel_centers, expand_bbox, step_size=half_voxel_size)