knowledge_induced_decoder/training_functions.py at master · watermeleon/knowledge_induced_decoder · GitHub

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


from zmq import device


def evaluate_loss(model, dataloader, loss_fn, spec, vocab_size):
    # Validation loss
    model.eval()
    running_loss = .0
    print("now doing eval loss")
    i = 0
    with tqdm(desc='Epoch %d - validation' % e, unit='it', total=len(dataloader),  disable=spec['tdqm_disable']) as pbar:
        with torch.no_grad():
            for it, (detections, captions, context_feats) in enumerate(dataloader):
                detections, captions, context_feats = detections.to(device), captions.to(device), context_feats.to(device)
                out = model(detections, captions, context_feats)
                captions = captions[:, 1:].contiguous()
                out = out[:, :-1].contiguous()
                loss = loss_fn(out.view(-1, vocab_size), captions.view(-1))
                this_loss = loss.item()
                running_loss += this_loss

                pbar.set_postfix(loss=running_loss / (it + 1))
                pbar.update()

    val_loss = running_loss / len(dataloader)
    return val_loss


def evaluate_metrics_standard(model, dataloader, spec, transform_tok = None):
    model.eval()
    gen = {}
    gts = {}
    seq_len = 20

    print("now doing eval metrics")
    with tqdm(desc='Epoch %d - evaluation' % e, unit='it', total=len(dataloader), disable=spec['tdqm_disable']) as pbar:
        for it, (images, caps_gt) in enumerate(iter(dataloader)):
            caps_gt, context_feats = caps_gt[0], torch.stack(caps_gt[1])
            context_feats = context_feats[:,0,:,:]
            images, context_feats = images.to(device), context_feats.to(device)
            with torch.no_grad():
                out, _ = model.beam_search(images, context_feats, seq_len, spec['eos_tokenid'], 5, out_size=1)

            caps_gen = [transform_tok.decode(sent) for sent in out]
            caps_gen = [sent.split("<|endoftext|>")[0] for sent in caps_gen]

            for i, (gts_i, gen_i) in enumerate(zip(caps_gt, caps_gen)):
                gen['%d_%d' % (it, i)] = [gen_i, ]
                gts['%d_%d' % (it, i)] = gts_i
            pbar.update()

    gts = evaluation.PTBTokenizer.tokenize(gts)
    gen = evaluation.PTBTokenizer.tokenize(gen)
    scores, _ = evaluation.compute_scores(gts, gen)
    return scores

def evaluate_metrics_gpt2(model, dataloader, spec, transform_tok = None):
    model.eval()
    gen = {}
    gts = {}

    print("now doing eval metrics")
    with tqdm(desc='Epoch %d - evaluation' % e, unit='it', total=len(dataloader), disable=spec['tdqm_disable']) as pbar:
        for it, (images, caps_gt) in enumerate(iter(dataloader)):
            caps_gt, context_feats = caps_gt[0], torch.stack(caps_gt[1])
            context_feats = context_feats[:,0,:,:]
            images, context_feats = images.to(device), context_feats.to(device)
            caps_gen = []
            with torch.no_grad():
                enc_output, mask_enc = model.encoder(images)
                prefix_embed = model.decoder(torch.ones((len(caps_gt), 6), dtype=int, device=device).to(device), enc_output, mask_enc, context_feats, gen_sent=True)
                for prefix_i in prefix_embed:
                    out = generate2(model.decoder, transform_tok, embed=prefix_i[None,:])
                    caps_gen.append(out)

            for i, (gts_i, gen_i) in enumerate(zip(caps_gt, caps_gen)):
                gen['%d_%d' % (it, i)] = [gen_i, ]
                gts['%d_%d' % (it, i)] = gts_i
            pbar.update()


    gts = evaluation.PTBTokenizer.tokenize(gts)
    gen = evaluation.PTBTokenizer.tokenize(gen)
    scores, _ = evaluation.compute_scores(gts, gen)
    return scores


def train_xe(model, dataloader, optim, spec, vocab_size):
    # Training with cross-entropy
    model.train()
    running_loss = .0
    i = 0
    print("training XE")

    with tqdm(desc='Epoch %d - train' % e, unit='it', total=len(dataloader),  disable=spec['tdqm_disable']) as pbar:
        for it, (detections, captions, context_feats) in enumerate(dataloader):
            detections, captions, context_feats = detections.to(device), captions.to(device), context_feats.to(device)
            out = model(detections, captions, context_feats)
            optim.zero_grad()
            captions_gt = captions[:, 1:].contiguous()

            out = out[:, :-1].contiguous()
            loss = loss_fn(out.view(-1, vocab_size), captions_gt.view(-1))
            loss.backward()

            optim.step()
            this_loss = loss.item()
            running_loss += this_loss

            pbar.set_postfix(loss=running_loss / (it + 1))
            pbar.update()
            scheduler.step()

    loss = running_loss / len(dataloader)
    return loss


def train_scst(model, dataloader, optim, cider, spec, transform_tok):

    tokenizer_pool = Pool()
    running_reward = .0
    running_reward_baseline = .0
    model.train()
    running_loss = .0
    seq_len = 20
    beam_size = 5
    print("trainin SCTS")
    with tqdm(desc='Epoch %d - train' % e, unit='it', total=len(dataloader),  disable=spec['tdqm_disable']) as pbar:
        for it, (detections, caps_gt) in enumerate(dataloader):
            caps_gt, context_feats = caps_gt[0], torch.stack(caps_gt[1])
            context_feats = context_feats[:,0,:,:]
            detections, context_feats = detections.to(device) , context_feats.to(device)
            outs, log_probs = model.beam_search(detections, context_feats, seq_len, spec["eos_tokenid"],
                                                beam_size, out_size=beam_size)
            optim.zero_grad()

            # Rewards
            caps_gt = list(itertools.chain(*([c, ] * beam_size for c in caps_gt)))
            caps_gen = [transform_tok.decode(sent) for sent in outs.view(-1, seq_len)]
            caps_gen = [sent.split("<|endoftext|>")[0] for sent in caps_gen]
            caps_gen, caps_gt = tokenizer_pool.map(evaluation.PTBTokenizer.tokenize, [caps_gen, caps_gt])

            reward = cider.compute_score(caps_gt, caps_gen)[1].astype(np.float32)
            reward = torch.from_numpy(reward).to(device).view(detections.shape[0], beam_size)
            reward_baseline = torch.mean(reward, -1, keepdim=True)
            loss = -torch.mean(log_probs, -1) * (reward - reward_baseline)

            loss = loss.mean()
            loss.backward()
            optim.step()

            running_loss += loss.item()
            running_reward += reward.mean().item()
            running_reward_baseline += reward_baseline.mean().item()
            pbar.set_postfix(loss=running_loss / (it + 1), reward=running_reward / (it + 1),
                             reward_baseline=running_reward_baseline / (it + 1))
            pbar.update()

    loss = running_loss / len(dataloader)
    reward = running_reward / len(dataloader)
    reward_baseline = running_reward_baseline / len(dataloader)

    return loss, reward, reward_baseline