mirrors/icefall
1
0
Fork 0
mirror of https://github.com/k2-fsa/icefall.git synced 2025-09-07 16:14:17 +00:00
Code Issues Packages Projects Releases Wiki Activity

Update decode.py

Browse Source
This commit is contained in:
jinzr 2023-09-26 14:12:20 +08:00
parent 32e807763f
commit 8467eb7c26
1 changed files with 9 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
egs/multi_zh_en/ASR/zipformer/decode.py
View File

@ -383,7 +383,7 @@ def decode_one_batch(
max_states=params.max_states, max_states=params.max_states,
) )
for hyp in sp.decode(hyp_tokens): for hyp in sp.decode(hyp_tokens):
hyps.append(hyp.split()) hyps.append(smart_byte_decode(hyp).split())
elif params.decoding_method == "fast_beam_search_nbest_LG": elif params.decoding_method == "fast_beam_search_nbest_LG":
hyp_tokens = fast_beam_search_nbest_LG( hyp_tokens = fast_beam_search_nbest_LG(
model=model, model=model,
@ -411,7 +411,7 @@ def decode_one_batch(
nbest_scale=params.nbest_scale, nbest_scale=params.nbest_scale,
) )
for hyp in sp.decode(hyp_tokens): for hyp in sp.decode(hyp_tokens):
hyps.append(hyp.split()) hyps.append(smart_byte_decode(hyp).split())
elif params.decoding_method == "fast_beam_search_nbest_oracle": elif params.decoding_method == "fast_beam_search_nbest_oracle":
hyp_tokens = fast_beam_search_nbest_oracle( hyp_tokens = fast_beam_search_nbest_oracle(
model=model, model=model,
@ -422,11 +422,13 @@ def decode_one_batch(
max_contexts=params.max_contexts, max_contexts=params.max_contexts,
max_states=params.max_states, max_states=params.max_states,
num_paths=params.num_paths, num_paths=params.num_paths,
ref_texts=sp.encode(supervisions["text"]), ref_texts=sp.encode(
byte_encode(tokenize_by_CJK_char(supervisions["text"]))
),
nbest_scale=params.nbest_scale, nbest_scale=params.nbest_scale,
) )
for hyp in sp.decode(hyp_tokens): for hyp in sp.decode(hyp_tokens):
hyps.append(hyp.split()) hyps.append(smart_byte_decode(hyp).split())
elif params.decoding_method == "greedy_search" and params.max_sym_per_frame == 1: elif params.decoding_method == "greedy_search" and params.max_sym_per_frame == 1:
hyp_tokens = greedy_search_batch( hyp_tokens = greedy_search_batch(
model=model, model=model,
@ -434,7 +436,7 @@ def decode_one_batch(
encoder_out_lens=encoder_out_lens, encoder_out_lens=encoder_out_lens,
) )
for hyp in sp.decode(hyp_tokens): for hyp in sp.decode(hyp_tokens):
hyps.append(hyp.split()) hyps.append(smart_byte_decode(hyp).split())
elif params.decoding_method == "modified_beam_search": elif params.decoding_method == "modified_beam_search":
hyp_tokens = modified_beam_search( hyp_tokens = modified_beam_search(
model=model, model=model,
@ -443,7 +445,7 @@ def decode_one_batch(
beam=params.beam_size, beam=params.beam_size,
) )
for hyp in sp.decode(hyp_tokens): for hyp in sp.decode(hyp_tokens):
hyps.append(hyp.split()) hyps.append(smart_byte_decode(hyp).split())
else: else:
batch_size = encoder_out.size(0) batch_size = encoder_out.size(0)
@ -467,7 +469,7 @@ def decode_one_batch(
raise ValueError( raise ValueError(
f"Unsupported decoding method: {params.decoding_method}" f"Unsupported decoding method: {params.decoding_method}"
) )
hyps.append(sp.decode(hyp).split()) hyps.append(smart_byte_decode(sp.decode(hyp)).split())
if params.decoding_method == "greedy_search": if params.decoding_method == "greedy_search":
return {"greedy_search": hyps} return {"greedy_search": hyps}