File size: 7,441 Bytes
987ed1b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import torch
import torch.nn.functional as F
from diffusion_policy.sampler.metric import euclidean_distance, coverage_distance

torch.set_printoptions(precision=1, sci_mode=False)

def coherence_sampler(policy, prior, obs_dict, num_sample=10, beta=0.5):
    """
    Sample an action from a policy that preserves coherence with a prior.

    Args:
        policy: a policy network to predict sequences of actions
        prior: the prediction made in the previous time step
        obs_dict: dictionary containing observations at the current time step
        num_sample (int, optional): number of samples to generate
        beta (float, optional): weight decay factor for coherence

    Returns:
        dict: a selected dictionary of actions 
    """
    if prior is None:
        return policy.predict_action(obs_dict)

    # pre-process
    B, OH, OD = obs_dict['obs'].shape
    obs_dict_batch = dict()
    for key in obs_dict.keys():
        if key == 'prior':
            continue
        obs_dict_batch[key] = obs_dict[key].unsqueeze(1).repeat(1, num_sample, 1, 1).view(B * num_sample, OH, OD)

    # predict
    action_dict_batch = policy.predict_action(obs_dict_batch)

    # post-process
    AH, PH, AD = action_dict_batch['action'].shape[1], action_dict_batch['action_pred'].shape[1], action_dict_batch['action_pred'].shape[2]
    action_dict_batch['action'] = action_dict_batch['action'].view(B, num_sample, AH, AD)
    action_dict_batch['action_pred'] = action_dict_batch['action_pred'].view(B, num_sample, PH, AD)
    if 'action_obs_pred' in action_dict_batch:
        action_dict_batch['action_obs_pred'] = action_dict_batch['action_obs_pred'].view(B, num_sample, AH, OD)
    if 'obs_pred' in action_dict_batch:
        action_dict_batch['obs_pred'] = action_dict_batch['obs_pred'].view(B, num_sample, PH, OD)

    # distance measure
    start_overlap = policy.n_obs_steps - 1
    end_overlap = prior.shape[1]
    dist_raw = euclidean_distance(action_dict_batch['action_pred'][:, :, start_overlap:end_overlap], prior.unsqueeze(1)[:, :, start_overlap:], reduction='none')

    weights = torch.tensor([beta**i for i in range(end_overlap-start_overlap)]).to(dist_raw.device)
    weights = weights / weights.sum()
    dist_weighted = dist_raw * weights.view(1, 1, end_overlap-start_overlap)
    dist = dist_weighted.sum(dim=2)

    # sample selection
    _, cross_index = dist.sort(descending=False)
    index = cross_index[:, 0]

    # slicing
    action_dict = dict()
    range_tensor = torch.arange(B, device=index.device)
    for key in action_dict_batch.keys():
        action_dict[key] = action_dict_batch[key][range_tensor, index]

    return action_dict

def ema_sampler(policy, prior, obs_dict, beta):
    action_dict = policy.predict_action(obs_dict)
    if prior is not None:
        # frame matching
        if policy.oa_step_convention:
            start = policy.n_obs_steps - 1
        else:
            start = policy.n_obs_steps
        end = start + policy.n_action_steps
        assert (action_dict['action'] == action_dict['action_pred'][:,start:end]).all().item()
        # ema update
        CH = prior.shape[1]
        action_dict['action_pred'][:,:CH] = prior * beta + action_dict['action_pred'][:,:CH] * (1. - beta)
        action_dict['action'] = action_dict['action_pred'][:,start:end]
    return action_dict

def cma_sampler(policy, prior, obs_dict, num_sample=10, beta1=0.75, beta2=0.95):
    if prior is None:
        return policy.predict_action(obs_dict)

    # pre-process
    B, OH, OD = obs_dict['obs'].shape
    obs_dict_batch = dict()
    for key in obs_dict.keys():
        obs_dict_batch[key] = obs_dict[key].unsqueeze(1).repeat(1, num_sample, 1, 1).reshape(B * num_sample, OH, OD)

    # predict
    action_dict_batch = policy.predict_action(obs_dict_batch)

    # post-process
    AH, PH, AD = action_dict_batch['action'].shape[1], action_dict_batch['action_pred'].shape[1], action_dict_batch['action_pred'].shape[2]
    action_dict_batch['action'] = action_dict_batch['action'].reshape(B, num_sample, AH, AD)
    action_dict_batch['action_pred'] = action_dict_batch['action_pred'].reshape(B, num_sample, PH, AD)
    if 'action_obs_pred' in action_dict_batch:
        action_dict_batch['action_obs_pred'] = action_dict_batch['action_obs_pred'].reshape(B, num_sample, AH, OD)
    if 'obs_pred' in action_dict_batch:
        action_dict_batch['obs_pred'] = action_dict_batch['obs_pred'].reshape(B, num_sample, PH, OD)

    # distance measure
    CH = prior.shape[1]
    dist_raw = euclidean_distance(action_dict_batch['action_pred'][:, :, :CH], prior.unsqueeze(1), reduction='none')

    weights = torch.tensor([beta2**i for i in range(CH)]).to(dist_raw.device)
    weights = weights / weights.sum()
    dist_weighted = dist_raw * weights.view(1, 1, CH)
    dist = dist_weighted.sum(dim=2)

    # sample selection
    _, cross_index = dist.sort(descending=False)
    index = cross_index[:, 0]

    # slicing
    action_dict = dict()
    range_tensor = torch.arange(B, device=index.device)
    for key in action_dict_batch.keys():
        action_dict[key] = action_dict_batch[key][range_tensor, index]

    # frame matching
    if policy.oa_step_convention:
        start = policy.n_obs_steps - 1
    else:
        start = policy.n_obs_steps
    end = start + policy.n_action_steps
    assert (action_dict['action'] == action_dict['action_pred'][:,start:end]).all().item()

    # ema update
    action_dict['action_pred'][:,:CH] = prior * beta1 + action_dict['action_pred'][:,:CH] * (1. - beta1)
    action_dict['action'] = action_dict['action_pred'][:,start:end]

    return action_dict


def ac_sampler(policy, prior, obs_dict, tau):
    # Essential, this sampler is sgac sampler with: previous_obs_dict=None
    action_dict = policy.predict_action(obs_dict)
    if prior is not None:
        # frame matching
        start = policy.n_obs_steps - 1
        end = start + policy.n_action_steps
        assert (action_dict['action'] == action_dict['action_pred'][:, start:end]).all().item()
        CH = prior.shape[1]

        new = policy.normalizer['action'].normalize(action_dict['action_pred'][:, :CH])[:, start:end]
        old = policy.normalizer['action'].normalize(prior)[:, start:end]
        cos_sim = F.cosine_similarity(new, old, dim=2, eps=1e-8)

        has_negative = (cos_sim < tau).any(dim=1)  # [B]
        mask = ~has_negative

        action_dict['action_pred'][:, :CH][mask] = prior[mask]
        action_dict['action'] = action_dict['action_pred'][:, start:end]

    return action_dict


def sgac_sampler(policy, prior, obs_dict, previous_obs_dict, tau):
    action_dict = policy.predict_action(obs_dict, previous_obs_dict)
    if prior is not None:
        # frame matching
        start = policy.n_obs_steps - 1
        end = start + policy.n_action_steps
        assert (action_dict['action'] == action_dict['action_pred'][:, start:end]).all().item()
        CH = prior.shape[1]

        new = policy.normalizer['action'].normalize(action_dict['action_pred'][:, :CH])[:, start:end]
        old = policy.normalizer['action'].normalize(prior)[:, start:end]
        cos_sim = F.cosine_similarity(new, old, dim=2, eps=1e-8)

        has_negative = (cos_sim < tau).any(dim=1)  # [B]
        mask = ~has_negative

        action_dict['action_pred'][:, :CH][mask] = prior[mask]
        action_dict['action'] = action_dict['action_pred'][:, start:end]

    return action_dict