From 4ace5a95b052d338947fc88809a440ccd55cf6b4 Mon Sep 17 00:00:00 2001
From: 游雁 <zhifu.gzf@alibaba-inc.com>
Date: 星期四, 16 十一月 2023 16:39:52 +0800
Subject: [PATCH] funasr pages
---
funasr/models/decoder/sanm_decoder.py | 141 ++++++++++++++---------------------------------
1 files changed, 42 insertions(+), 99 deletions(-)
diff --git a/funasr/models/decoder/sanm_decoder.py b/funasr/models/decoder/sanm_decoder.py
index 3e4e554..ff35e46 100644
--- a/funasr/models/decoder/sanm_decoder.py
+++ b/funasr/models/decoder/sanm_decoder.py
@@ -105,48 +105,48 @@
return x, tgt_mask, memory, memory_mask, cache
- #def forward_chunk(self, tgt, tgt_mask, memory, memory_mask=None, cache=None):
- # """Compute decoded features.
+ def forward_one_step(self, tgt, tgt_mask, memory, memory_mask=None, cache=None):
+ """Compute decoded features.
- # Args:
- # tgt (torch.Tensor): Input tensor (#batch, maxlen_out, size).
- # tgt_mask (torch.Tensor): Mask for input tensor (#batch, maxlen_out).
- # memory (torch.Tensor): Encoded memory, float32 (#batch, maxlen_in, size).
- # memory_mask (torch.Tensor): Encoded memory mask (#batch, maxlen_in).
- # cache (List[torch.Tensor]): List of cached tensors.
- # Each tensor shape should be (#batch, maxlen_out - 1, size).
+ Args:
+ tgt (torch.Tensor): Input tensor (#batch, maxlen_out, size).
+ tgt_mask (torch.Tensor): Mask for input tensor (#batch, maxlen_out).
+ memory (torch.Tensor): Encoded memory, float32 (#batch, maxlen_in, size).
+ memory_mask (torch.Tensor): Encoded memory mask (#batch, maxlen_in).
+ cache (List[torch.Tensor]): List of cached tensors.
+ Each tensor shape should be (#batch, maxlen_out - 1, size).
- # Returns:
- # torch.Tensor: Output tensor(#batch, maxlen_out, size).
- # torch.Tensor: Mask for output tensor (#batch, maxlen_out).
- # torch.Tensor: Encoded memory (#batch, maxlen_in, size).
- # torch.Tensor: Encoded memory mask (#batch, maxlen_in).
+ Returns:
+ torch.Tensor: Output tensor(#batch, maxlen_out, size).
+ torch.Tensor: Mask for output tensor (#batch, maxlen_out).
+ torch.Tensor: Encoded memory (#batch, maxlen_in, size).
+ torch.Tensor: Encoded memory mask (#batch, maxlen_in).
- # """
- # # tgt = self.dropout(tgt)
- # residual = tgt
- # if self.normalize_before:
- # tgt = self.norm1(tgt)
- # tgt = self.feed_forward(tgt)
+ """
+ # tgt = self.dropout(tgt)
+ residual = tgt
+ if self.normalize_before:
+ tgt = self.norm1(tgt)
+ tgt = self.feed_forward(tgt)
- # x = tgt
- # if self.self_attn:
- # if self.normalize_before:
- # tgt = self.norm2(tgt)
- # if self.training:
- # cache = None
- # x, cache = self.self_attn(tgt, tgt_mask, cache=cache)
- # x = residual + self.dropout(x)
+ x = tgt
+ if self.self_attn:
+ if self.normalize_before:
+ tgt = self.norm2(tgt)
+ if self.training:
+ cache = None
+ x, cache = self.self_attn(tgt, tgt_mask, cache=cache)
+ x = residual + self.dropout(x)
- # if self.src_attn is not None:
- # residual = x
- # if self.normalize_before:
- # x = self.norm3(x)
+ if self.src_attn is not None:
+ residual = x
+ if self.normalize_before:
+ x = self.norm3(x)
- # x = residual + self.dropout(self.src_attn(x, memory, memory_mask))
+ x = residual + self.dropout(self.src_attn(x, memory, memory_mask))
- # return x, tgt_mask, memory, memory_mask, cache
+ return x, tgt_mask, memory, memory_mask, cache
def forward_chunk(self, tgt, memory, fsmn_cache=None, opt_cache=None, chunk_size=None, look_back=0):
"""Compute decoded features.
@@ -438,7 +438,7 @@
for i in range(self.att_layer_num):
decoder = self.decoders[i]
c = cache[i]
- x, tgt_mask, memory, memory_mask, c_ret = decoder.forward_chunk(
+ x, tgt_mask, memory, memory_mask, c_ret = decoder.forward_one_step(
x, tgt_mask, memory, memory_mask, cache=c
)
new_cache.append(c_ret)
@@ -448,13 +448,13 @@
j = i + self.att_layer_num
decoder = self.decoders2[i]
c = cache[j]
- x, tgt_mask, memory, memory_mask, c_ret = decoder.forward_chunk(
+ x, tgt_mask, memory, memory_mask, c_ret = decoder.forward_one_step(
x, tgt_mask, memory, memory_mask, cache=c
)
new_cache.append(c_ret)
for decoder in self.decoders3:
- x, tgt_mask, memory, memory_mask, _ = decoder.forward_chunk(
+ x, tgt_mask, memory, memory_mask, _ = decoder.forward_one_step(
x, tgt_mask, memory, None, cache=None
)
@@ -878,6 +878,7 @@
lora_rank: int = 8,
lora_alpha: int = 16,
lora_dropout: float = 0.1,
+ chunk_multiply_factor: tuple = (1,),
tf2torch_tensor_name_prefix_torch: str = "decoder",
tf2torch_tensor_name_prefix_tf: str = "seq2seq/decoder",
):
@@ -970,6 +971,7 @@
)
self.tf2torch_tensor_name_prefix_torch = tf2torch_tensor_name_prefix_torch
self.tf2torch_tensor_name_prefix_tf = tf2torch_tensor_name_prefix_tf
+ self.chunk_multiply_factor = chunk_multiply_factor
def forward(
self,
@@ -1032,65 +1034,6 @@
ys.unsqueeze(0), ys_mask, x.unsqueeze(0), cache=state
)
return logp.squeeze(0), state
-
- #def forward_chunk(
- # self,
- # memory: torch.Tensor,
- # tgt: torch.Tensor,
- # cache: dict = None,
- #) -> Tuple[torch.Tensor, torch.Tensor]:
- # """Forward decoder.
-
- # Args:
- # hs_pad: encoded memory, float32 (batch, maxlen_in, feat)
- # hlens: (batch)
- # ys_in_pad:
- # input token ids, int64 (batch, maxlen_out)
- # if input_layer == "embed"
- # input tensor (batch, maxlen_out, #mels) in the other cases
- # ys_in_lens: (batch)
- # Returns:
- # (tuple): tuple containing:
-
- # x: decoded token score before softmax (batch, maxlen_out, token)
- # if use_output_layer is True,
- # olens: (batch, )
- # """
- # x = tgt
- # if cache["decode_fsmn"] is None:
- # cache_layer_num = len(self.decoders)
- # if self.decoders2 is not None:
- # cache_layer_num += len(self.decoders2)
- # new_cache = [None] * cache_layer_num
- # else:
- # new_cache = cache["decode_fsmn"]
- # for i in range(self.att_layer_num):
- # decoder = self.decoders[i]
- # x, tgt_mask, memory, memory_mask, c_ret = decoder.forward_chunk(
- # x, None, memory, None, cache=new_cache[i]
- # )
- # new_cache[i] = c_ret
-
- # if self.num_blocks - self.att_layer_num > 1:
- # for i in range(self.num_blocks - self.att_layer_num):
- # j = i + self.att_layer_num
- # decoder = self.decoders2[i]
- # x, tgt_mask, memory, memory_mask, c_ret = decoder.forward_chunk(
- # x, None, memory, None, cache=new_cache[j]
- # )
- # new_cache[j] = c_ret
-
- # for decoder in self.decoders3:
-
- # x, tgt_mask, memory, memory_mask, _ = decoder.forward_chunk(
- # x, None, memory, None, cache=None
- # )
- # if self.normalize_before:
- # x = self.after_norm(x)
- # if self.output_layer is not None:
- # x = self.output_layer(x)
- # cache["decode_fsmn"] = new_cache
- # return x
def forward_chunk(
self,
@@ -1190,7 +1133,7 @@
for i in range(self.att_layer_num):
decoder = self.decoders[i]
c = cache[i]
- x, tgt_mask, memory, memory_mask, c_ret = decoder.forward_chunk(
+ x, tgt_mask, memory, memory_mask, c_ret = decoder.forward_one_step(
x, tgt_mask, memory, None, cache=c
)
new_cache.append(c_ret)
@@ -1200,14 +1143,14 @@
j = i + self.att_layer_num
decoder = self.decoders2[i]
c = cache[j]
- x, tgt_mask, memory, memory_mask, c_ret = decoder.forward_chunk(
+ x, tgt_mask, memory, memory_mask, c_ret = decoder.forward_one_step(
x, tgt_mask, memory, None, cache=c
)
new_cache.append(c_ret)
for decoder in self.decoders3:
- x, tgt_mask, memory, memory_mask, _ = decoder.forward_chunk(
+ x, tgt_mask, memory, memory_mask, _ = decoder.forward_one_step(
x, tgt_mask, memory, None, cache=None
)
--
Gitblit v1.9.1