-
Notifications
You must be signed in to change notification settings - Fork 1
/
arguments.py
248 lines (236 loc) · 19.1 KB
/
arguments.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
import argparse
import numpy as np
import os
import torch
parser = argparse.ArgumentParser()
parser.add_argument("--seed", type=int, default=39, help="Random seed")
parser.add_argument("--mode", type=str, help="mode to run, see main.py")
parser.add_argument("--verbose", action="store_true", default=False, help="print out actions + other logs during task")
parser.add_argument("--set_name", type=str, default="test00", help="experiment name")
parser.add_argument("--split", type=str, help="eval split")
parser.add_argument("--add_back_objs_progresscheck_v2", action="store_true", default=False, help="add back in objects to object tracker for progress check")
parser.add_argument("--use_openai", action="store_true", default=False, help="")
parser.add_argument("--gpt_model", type=str, default="gpt-4", help="options: gpt-3.5-turbo, text-davinci-003, gpt-4, gpt-4-32k")
parser.add_argument('--skip_if_exists', default=False, action='store_true', help='skip if file exists in teach metrics')
parser.add_argument("--dpi", type=int, default=100, help="DPI for plotting")
parser.add_argument("--max_traj_steps", type=int, default=1000, help="maximum trajectory steps")
parser.add_argument('--remove_map_vis', default=False, action='store_true', help='remove map visual from movies')
parser.add_argument('--always_check_constraints', default=False, action='store_true', help='')
parser.add_argument("--root", type=str, default="", help="root folder")
parser.add_argument("--tag", type=str, default="", help="root folder tag")
parser.add_argument("--teleport_to_objs", action="store_true", default=False, help="teleport to objects instead of navigating")
parser.add_argument("--render", action="store_true", default=False, help="render video and logs")
parser.add_argument("--use_gt_objecttrack", action="store_true", default=False, help="if navigating, use GT object masks for getting object detections + centroids?")
parser.add_argument("--use_gt_depth", action="store_true", default=False, help="if navigating, use GT depth maps? ")
parser.add_argument("--use_gt_seg", action="store_true", default=False, help="use GT segmentation?")
parser.add_argument("--use_gt_success_checker", action="store_true", default=False, help="use GT segmentation?")
parser.add_argument("--use_gt_centroids", action="store_true", default=False, help="use GT centroids?")
parser.add_argument("--do_masks", action="store_true", default=False, help="use masks?")
parser.add_argument("--use_solq", action="store_true", default=False, help="use SOLQ?")
parser.add_argument("--use_gt_subgoals", action="store_true", default=False, help="use GT subgoals?")
parser.add_argument("--sample_every_other", action="store_true", default=False, help="run every other episode in the split")
parser.add_argument("--episode_in_try_except", action="store_true", default=False, help="Continue to next episode if assertion error occurs? ")
parser.add_argument("--log_every", type=int, default=1, help="log every X episodes")
parser.add_argument("--on_aws", action="store_true", default=False, help="on AWS?")
parser.add_argument("--new_parser", action="store_true", default=False, help="on AWS?")
parser.add_argument('--load_explore', action='store_true', default=False, help="load explore for full task from path?")
parser.add_argument("--movie_dir", type=str, default="./output/movies", help="where to output rendered movies")
parser.add_argument("--precompute_map_path", default='./output/precomputed_maps', type=str, help="load trajectory list from file?")
parser.add_argument("--create_movie", action="store_true", default=False, help="create mp4 movie")
parser.add_argument("--visualize_masks", action="store_true", default=False, help="visualize masks in object tracker visuals")
parser.add_argument("--metrics_dir", type=str, default="./output/metrics", help="where to output rendered movies")
parser.add_argument("--llm_output_dir", type=str, default="./output/llm", help="where to output rendered movies")
parser.add_argument("--gpt_embedding_dir", type=str, default="./data", help="where to output rendered movies")
parser.add_argument("--run_error_correction_llm", action="store_true", default=False, help="run error correction for LLM")
parser.add_argument("--run_error_correction_basic", action="store_true", default=False, help="run error correction - manual correction")
parser.add_argument("--use_progress_check", action="store_true", default=False, help="run progress check at the end to replan")
parser.add_argument("--remove_unusable_slice", action="store_true", default=False, help="remove the unusable slice from the environment after slicing")
parser.add_argument("--add_back_objs_progresscheck", action="store_true", default=False, help="add back in objects to object tracker for progress check")
parser.add_argument("--teach_data_dir", type=str, default="./data", help="data directory where teach data is held")
parser.add_argument("--data_path", type=str, default="./data", help="data directory where teach data is held")
parser.add_argument("--dont_use_controller", action="store_true", default=False, help="dont init controller")
parser.add_argument("--use_constraint_check", action="store_true", default=False, help="use constraint check")
parser.add_argument("--num_continual_iter", type=int, default=3, help="how many continual learning iterations?")
parser.add_argument("--mod_api_continual", action="store_true", default=False, help="setting to modify the api instead of example retrieval for continual learning experiments")
parser.add_argument("--ablate_example_retrieval", action="store_true", default=False, help="ablate example retrieval for GPT?")
parser.add_argument("--increased_explore", action="store_true", default=False, help="increase explore for GT depth")
parser.add_argument("--max_episodes", type=int, default=None, help="maximum episodes to evaluate")
###########%%%%%%% agent parameters %%%%%%%###########
parser.add_argument("--start_startx", action="store_true", default=False, help="start x server upon calling main")
parser.add_argument("--server_port", type=int, default=1, help="server port for x server")
parser.add_argument("--do_headless_rendering", action="store_true", default=False, help="render in headless mode with new Ai2thor version")
parser.add_argument("--HORIZON_DT", type=int, default=30, help="pitch movement delta")
parser.add_argument("--DT", type=int, default=90, help="yaw movement delta")
parser.add_argument("--STEP_SIZE", type=int, default=0.25, help="yaw movement delta")
parser.add_argument("--pitch_range", type=list, default=[-30,60], help="pitch allowable range for the agent. positive is 'down'")
parser.add_argument("--fov", type=int, default=90, help="field of view")
parser.add_argument("--W", type=int, default=480, help="image width")
parser.add_argument("--H", type=int, default=480, help="image height")
parser.add_argument("--visibilityDistance", type=float, default=1.5, help="visibility NOTE: this will not change rearrangement visibility")
parser.add_argument('--debug', default=False, action='store_true')
parser.add_argument("--eval_split", type=str, default="test", help="evaluation mode: combined (rearrange), train, test, val")
parser.add_argument('--use_estimated_depth', action='store_true', help="use estimated depth?")
parser.add_argument('--num_search_locs_object', type=int, default=20, help='number of search locations for searching for object')
parser.add_argument("--dist_thresh", type=float, default=0.5, help="navigation distance threshold to point goal")
parser.add_argument('--use_GT_constraint_checks', default=False, action='store_true', help='use GT constraint checks?')
parser.add_argument('--use_GT_error_feedback', default=False, action='store_true', help='use GT error feedback?')
parser.add_argument("--max_api_fails", type=int, default=30, help="maximum allowable api failures")
parser.add_argument('--use_llm_search', default=False, action='store_true', help='use llm search')
parser.add_argument('--use_mask_rcnn_pred', default=False, action='store_true', help='use maskrcnn')
parser.add_argument("--episode_file", type=str, default=None, help="specify an episode file name to run")
###########%%%%%%% object tracker %%%%%%%###########
parser.add_argument("--OT_dist_thresh", type=float, default=0.5, help="distance threshold for NMS for object tracker")
parser.add_argument("--OT_dist_thresh_searching", type=float, default=0.5, help="distance threshold for NMS for object tracker")
parser.add_argument("--confidence_threshold", type=float, default=0.4, help="confidence threshold for detections [0, 0.1]")
parser.add_argument("--confidence_threshold_interm", type=float, default=0.4, help="intermediate object score threshold")
parser.add_argument("--confidence_threshold_searching", type=float, default=0.4, help="confidence threshold for detections when searching for a target object class [0, 0.1]")
parser.add_argument("--nms_threshold", type=float, default=0.5, help="NMS threshold for object tracker")
parser.add_argument("--use_GT_centroids", action="store_true", default=False, help="use GT centroids for object tracker")
parser.add_argument('--only_one_obj_per_cat', action='store_true', default=False, help="only one object per cateogry in the object tracker?")
parser.add_argument('--env_frame_width_FILM', type=int, default=300, help='Frame width (default:84)')
parser.add_argument('--env_frame_height_FILM', type=int, default=300, help='Frame height (default:84)')
# Depth network
parser.add_argument("--randomize_object_placements", action="store_true", default=False, help="")
parser.add_argument("--randomize_scene_lighting_and_material", action="store_true", default=False, help="")
parser.add_argument("--randomize_agent_pickup", action="store_true", default=False, help="")
parser.add_argument("--randomize_object_state", action="store_true", default=False, help="")
parser.add_argument("--load_model", action="store_true", default=False, help="")
parser.add_argument("--load_model_path", type=str, default="", help="load checkpoint path")
parser.add_argument("--lr_scheduler_from_scratch", action="store_true", default=False, help="")
parser.add_argument("--optimizer_from_scratch", action="store_true", default=False, help="")
parser.add_argument("--start_one", action="store_true", default=False, help="")
parser.add_argument('--max_iters', type=int, default=100000, help='max train iterations')
parser.add_argument('--log_freq', type=int, default=500, help='log frequency')
parser.add_argument('--val_freq', type=int, default=500, help='validation frequency')
parser.add_argument('--save_freq', type=int, default=2500, help='save checkpoint frequency')
parser.add_argument("--load_train_agent", action="store_true", default=False, help="")
parser.add_argument('--batch_size', default=2, type=int, help="batch size for model training")
parser.add_argument("--S", type=int, default=2, help="Number of views per trajectory")
parser.add_argument("--radius_min", type=float, default=0.0, help="radius min to spawn near target object")
parser.add_argument("--radius_max", type=float, default=7.0, help="radius max to spawn near target object")
parser.add_argument("--views_to_attempt", type=int, default=8, help="max views to attempt for getting trajectory")
parser.add_argument("--movement_mode", type=str, default="random", help="movement mode for action sampling for getting trajectory (forward_first, random); forward_first: always try to move forward")
parser.add_argument("--fail_if_no_objects", type=bool, default=True, help="fail view if no objects in view")
parser.add_argument("--torch_checkpoint_path", type=str, default="", help="torch hub checkpoint path")
parser.add_argument('--lr_scheduler_freq', type=int, default=20000, help='lr frequency for step')
parser.add_argument('--keep_latest', default=5, type=int, help="number of checkpoints to keep at one time")
parser.add_argument("--val_load_dir", type=str, default="./dataset/val", help="val load dir")
parser.add_argument("--run_val", action="store_true", default=False, help="")
parser.add_argument("--load_val_agent", action="store_true", default=False, help="")
parser.add_argument('--n_val', type=int, default=50, help='number of validation iters')
parser.add_argument("--zoedepth_checkpoint", type=str, default="./checkpoints/ZOEDEPTH-model-00015000.pth", help="zoe depth checkpoint to load for teach")
parser.add_argument("--solq_checkpoint", type=str, default="./checkpoints/SOLQ-model-00023000.pth", help="SOLQ checkpoint to load for teach")
# SOLQ hyperparams
parser.add_argument('--lr', default=2e-5, type=float)
parser.add_argument('--lr_backbone_names', default=["backbone.0"], type=str, nargs='+')
parser.add_argument('--lr_backbone_mult', default=0.1, type=float)
parser.add_argument('--lr_linear_proj_names', default=['reference_points', 'sampling_offsets'], type=str, nargs='+')
parser.add_argument('--lr_linear_proj_mult', default=0.1, type=float)
parser.add_argument('--lr_text_encoder_mult', default=0.05, type=float)
parser.add_argument('--lr_text_encoder_names', default=["text_encoder"], type=str, nargs='+')
parser.add_argument('--weight_decay', default=1e-4, type=float)
parser.add_argument('--epochs', default=50, type=int)
parser.add_argument('--lr_drop', default=4000, type=int)
parser.add_argument('--save_period', default=10, type=int)
parser.add_argument('--lr_drop_epochs', default=None, type=int, nargs='+')
parser.add_argument('--clip_max_norm', default=1.0, type=float,
help='gradient clipping max norm')
parser.add_argument('--meta_arch', default='solq', type=str)
parser.add_argument('--sgd', action='store_true')
# Variants of Deformable DETR
parser.add_argument('--with_box_refine', default=True, action='store_true')
parser.add_argument('--two_stage', default=True)
# VecInst
parser.add_argument('--with_vector', default=True, action='store_true')
parser.add_argument('--n_keep', default=256, type=int,
help="Number of coeffs to be remained")
parser.add_argument('--gt_mask_len', default=128, type=int,
help="Size of target mask")
parser.add_argument('--vector_loss_coef', default=3, type=float)
parser.add_argument('--vector_hidden_dim', default=1024, type=int,
help="Size of the vector embeddings (dimension of the transformer)")
parser.add_argument('--no_vector_loss_norm', default=False, action='store_true')
parser.add_argument('--activation', default='relu', type=str, help="Activation function to use")
parser.add_argument('--checkpoint', default=False, action='store_true')
parser.add_argument('--vector_start_stage', default=0, type=int)
parser.add_argument('--num_machines', default=1, type=int)
parser.add_argument('--loss_type', default='l1', type=str)
parser.add_argument('--dcn', default=False, action='store_true')
# Model parameters
parser.add_argument('--frozen_weights', type=str, default=None,
help="Path to the pretrained model. If set, only the mask head will be trained")
parser.add_argument('--pretrained', default=None, help='resume from checkpoint')
# * Backbone
parser.add_argument('--backbone', default='resnet50', type=str,
help="Name of the convolutional backbone to use")
parser.add_argument('--dilation', action='store_true',
help="If true, we replace stride with dilation in the last convolutional block (DC5)")
parser.add_argument('--position_embedding', default='sine', type=str, choices=('sine', 'learned', 'rel'),
help="Type of positional embedding to use on top of the image features")
parser.add_argument('--position_embedding_scale', default=2 * np.pi, type=float,
help="position / size * scale")
parser.add_argument('--num_feature_levels', default=4, type=int, help='number of feature levels')
# * Transformer
parser.add_argument('--enc_layers', default=6, type=int,
help="Number of encoding layers in the transformer")
parser.add_argument('--dec_layers', default=6, type=int,
help="Number of decoding layers in the transformer")
parser.add_argument('--dim_feedforward', default=1024, type=int,
help="Intermediate size of the feedforward layers in the transformer blocks")
parser.add_argument('--hidden_dim', default=384, type=int,
help="Size of the embeddings (dimension of the transformer)")
parser.add_argument('--dropout', default=0.1, type=float,
help="Dropout applied in the transformer")
parser.add_argument('--nheads', default=8, type=int,
help="Number of attention heads inside the transformer's attentions")
parser.add_argument('--num_queries', default=300, type=int,
help="Number of query slots")
parser.add_argument('--dec_n_points', default=4, type=int)
parser.add_argument('--enc_n_points', default=4, type=int)
# * Segmentation
parser.add_argument('--masks', type=bool, default=True,
help="Train segmentation head if the flag is provided")
# Loss
parser.add_argument('--no_aux_loss', dest='aux_loss', action='store_false',
help="Disables auxiliary decoding losses (loss at each layer)")
# * Matcher
parser.add_argument('--set_cost_class', default=2, type=float,
help="Class coefficient in the matching cost")
parser.add_argument('--set_cost_bbox', default=5, type=float,
help="L1 box coefficient in the matching cost")
parser.add_argument('--set_cost_giou', default=2, type=float,
help="giou box coefficient in the matching cost")
parser.add_argument('--dataset_file', default='coco')
parser.add_argument('--coco_path', default='./data/coco', type=str)
parser.add_argument('--coco_panoptic_path', type=str)
parser.add_argument('--remove_difficult', action='store_true')
parser.add_argument('--alg', default='instformer', type=str)
parser.add_argument('--output_dir', default='',
help='path where to save, empty for no saving')
parser.add_argument('--device', default='cuda',
help='device to use for training / testing')
parser.add_argument('--resume', default='', help='resume from checkpoint')
parser.add_argument('--start_epoch', default=0, type=int, metavar='N',
help='start epoch')
parser.add_argument('--eval', action='store_true')
parser.add_argument('--test', action='store_true')
parser.add_argument('--num_workers', default=2, type=int)
parser.add_argument('--cache_mode', default=False, action='store_true', help='whether to cache images on memory')
# * Loss coefficients
parser.add_argument('--mask_loss_coef', default=1, type=float)
parser.add_argument('--dice_loss_coef', default=1, type=float)
parser.add_argument('--cls_loss_coef', default=2, type=float)
parser.add_argument('--bbox_loss_coef', default=5, type=float)
parser.add_argument('--giou_loss_coef', default=2, type=float)
parser.add_argument('--focal_alpha', default=0.25, type=float)
### WANDB
parser.add_argument("--group", type=str, default="default", help="group name")
parser.add_argument("--wandb_directory", type=str, default='./wandb', help="Path to wandb metadata")
args = parser.parse_args()
args.use_estimated_depth = True
if args.use_gt_depth:
args.use_estimated_depth = False
args.metrics_dir = os.path.join(args.metrics_dir, args.set_name)
args.llm_output_dir = os.path.join(args.llm_output_dir, args.set_name)
args.movie_dir = os.path.join(args.movie_dir, args.set_name)
args.cuda = torch.device('cuda' if torch.cuda.is_available() else 'cpu')