onnx

游雁

2023-02-25 14a1b5eb20c951b1fe23ca7ea389778a6899332a

onnx

 funasr/export/models/predictor/cif.py

@@ -16,6 +16,11 @@
    

    return mask.type(dtype).to(device) if device is not None else mask.type(dtype)



def sequence_mask_scripts(lengths, maxlen:int):

    row_vector = torch.arange(0, maxlen, 1).type(lengths.dtype).to(lengths.device)

    matrix = torch.unsqueeze(lengths, dim=-1)

    mask = row_vector < matrix

    return mask.type(torch.float32).to(lengths.device)



class CifPredictorV2(nn.Module):

    def __init__(self, model):

@@ -71,28 +76,131 @@
        

        return hidden, alphas, token_num_floor



# @torch.jit.script

# def cif(hidden, alphas, threshold: float):

#     batch_size, len_time, hidden_size = hidden.size()

#     threshold = torch.tensor([threshold], dtype=alphas.dtype).to(alphas.device)

#

#     # loop varss

#     integrate = torch.zeros([batch_size], device=hidden.device)

#     frame = torch.zeros([batch_size, hidden_size], device=hidden.device)

#     # intermediate vars along time

#     list_fires = []

#     list_frames = []

#

#     for t in range(len_time):

#         alpha = alphas[:, t]

#         distribution_completion = torch.ones([batch_size], device=hidden.device) - integrate

#

#         integrate += alpha

#         list_fires.append(integrate)

#

#         fire_place = integrate >= threshold

#         integrate = torch.where(fire_place,

#                                 integrate - torch.ones([batch_size], device=hidden.device),

#                                 integrate)

#         cur = torch.where(fire_place,

#                           distribution_completion,

#                           alpha)

#         remainds = alpha - cur

#

#         frame += cur[:, None] * hidden[:, t, :]

#         list_frames.append(frame)

#         frame = torch.where(fire_place[:, None].repeat(1, hidden_size),

#                             remainds[:, None] * hidden[:, t, :],

#                             frame)

#

#     fires = torch.stack(list_fires, 1)

#     frames = torch.stack(list_frames, 1)

#     list_ls = []

#     len_labels = torch.round(alphas.sum(-1)).int()

#     max_label_len = len_labels.max().item()

#     # print("type: {}".format(type(max_label_len)))

#     for b in range(batch_size):

#         fire = fires[b, :]

#         l = torch.index_select(frames[b, :, :], 0, torch.nonzero(fire >= threshold).squeeze())

#         pad_l = torch.zeros([int(max_label_len - l.size(0)), int(hidden_size)], dtype=l.dtype, device=hidden.device)

#         list_ls.append(torch.cat([l, pad_l], 0))

#     return torch.stack(list_ls, 0), fires



# @torch.jit.script

# def cif(hidden, alphas, threshold: float):

#     batch_size, len_time, hidden_size = hidden.size()

#     threshold = torch.tensor([threshold], dtype=alphas.dtype).to(alphas.device)

#

#     # loop varss

#     integrate = torch.zeros([batch_size], dtype=alphas.dtype, device=hidden.device)

#     frame = torch.zeros([batch_size, hidden_size], dtype=hidden.dtype, device=hidden.device)

#     # intermediate vars along time

#     list_fires = []

#     list_frames = []

#

#     for t in range(len_time):

#         alpha = alphas[:, t]

#         distribution_completion = torch.ones([batch_size], dtype=alphas.dtype, device=hidden.device) - integrate

#

#         integrate += alpha

#         list_fires.append(integrate)

#

#         fire_place = integrate >= threshold

#         integrate = torch.where(fire_place,

#                                 integrate - torch.ones([batch_size], dtype=alphas.dtype, device=hidden.device),

#                                 integrate)

#         cur = torch.where(fire_place,

#                           distribution_completion,

#                           alpha)

#         remainds = alpha - cur

#

#         frame += cur[:, None] * hidden[:, t, :]

#         list_frames.append(frame)

#         frame = torch.where(fire_place[:, None].repeat(1, hidden_size),

#                             remainds[:, None] * hidden[:, t, :],

#                             frame)

#

#     fires = torch.stack(list_fires, 1)

#     frames = torch.stack(list_frames, 1)

#     len_labels = torch.floor(torch.sum(alphas, dim=1)).int()

#     max_label_len = torch.max(len_labels)

#     pad_num = max_label_len - len_labels

#     pad_num_max = torch.max(pad_num).item()

#     frames_pad_tensor = torch.zeros([int(batch_size), int(pad_num_max), int(hidden_size)], dtype=frames.dtype,

#                                     device=frames.device)

#     fires_pad_tensor = torch.ones([int(batch_size), int(pad_num_max)], dtype=fires.dtype, device=fires.device)

#     fires_pad_tensor_mask = sequence_mask_scripts(pad_num, maxlen=int(pad_num_max))

#     fires_pad_tensor *= fires_pad_tensor_mask

#     frames_pad = torch.cat([frames, frames_pad_tensor], dim=1)

#     fires_pad = torch.cat([fires, fires_pad_tensor], dim=1)

#     index_bool = fires_pad >= threshold

#     frames_fire = frames_pad[index_bool]

#     frames_fire = torch.reshape(frames_fire, (int(batch_size), -1, int(hidden_size)))

#     frames_fire_mask = sequence_mask_scripts(len_labels, maxlen=int(max_label_len))

#     frames_fire *= frames_fire_mask[:, :, None]

#

#     return frames_fire, fires





@torch.jit.script

def cif(hidden, alphas, threshold: float):

    batch_size, len_time, hidden_size = hidden.size()

    threshold = torch.tensor([threshold], dtype=alphas.dtype).to(alphas.device)

    

    # loop varss

    integrate = torch.zeros([batch_size], device=hidden.device)

    frame = torch.zeros([batch_size, hidden_size], device=hidden.device)

    integrate = torch.zeros([batch_size], dtype=alphas.dtype, device=hidden.device)

    frame = torch.zeros([batch_size, hidden_size], dtype=hidden.dtype, device=hidden.device)

    # intermediate vars along time

    list_fires = []

    list_frames = []

    

    for t in range(len_time):

        alpha = alphas[:, t]

        distribution_completion = torch.ones([batch_size], device=hidden.device) - integrate

        distribution_completion = torch.ones([batch_size], dtype=alphas.dtype, device=hidden.device) - integrate

        

        integrate += alpha

        list_fires.append(integrate)

        

        fire_place = integrate >= threshold

        integrate = torch.where(fire_place,

                                integrate - torch.ones([batch_size], device=hidden.device),

                                integrate - torch.ones([batch_size], dtype=alphas.dtype, device=hidden.device),

                                integrate)

        cur = torch.where(fire_place,

                          distribution_completion,

@@ -107,13 +215,20 @@
    

    fires = torch.stack(list_fires, 1)

    frames = torch.stack(list_frames, 1)

    list_ls = []

    len_labels = torch.round(alphas.sum(-1)).int()

    max_label_len = len_labels.max().item()

    print("type: {}".format(type(max_label_len)))

    # list_ls = []

    len_labels = torch.round(alphas.sum(-1)).type(torch.int32)

    # max_label_len = int(torch.max(len_labels).item())

    # print("type: {}".format(type(max_label_len)))

    fire_idxs = fires >= threshold

    frame_fires = torch.zeros_like(hidden)

    max_label_len = frames[0, fire_idxs[0]].size(0)

    for b in range(batch_size):

        fire = fires[b, :]

        l = torch.index_select(frames[b, :, :], 0, torch.nonzero(fire >= threshold).squeeze())

        pad_l = torch.zeros([int(max_label_len - l.size(0)), int(hidden_size)], device=hidden.device)

        list_ls.append(torch.cat([l, pad_l], 0))

    return torch.stack(list_ls, 0), fires

        # fire = fires[b, :]

        frame_fire = frames[b, fire_idxs[b]]

        frame_len = frame_fire.size(0)

        frame_fires[b, :frame_len, :] = frame_fire

	
        if frame_len >= max_label_len:

            max_label_len = frame_len

    frame_fires = frame_fires[:, :max_label_len, :]

    return frame_fires, fires