From 33d3d2084403fd34b79c835d2f2fe04f6cd8f738 Mon Sep 17 00:00:00 2001
From: 游雁 <zhifu.gzf@alibaba-inc.com>
Date: 星期三, 13 九月 2023 09:33:54 +0800
Subject: [PATCH] Merge branch 'main' of github.com:alibaba-damo-academy/FunASR add
---
funasr/build_utils/build_model_from_file.py | 71 +++++++++++++++++++++++++++++++++--
1 files changed, 67 insertions(+), 4 deletions(-)
diff --git a/funasr/build_utils/build_model_from_file.py b/funasr/build_utils/build_model_from_file.py
index 5488c10..65e0d5f 100644
--- a/funasr/build_utils/build_model_from_file.py
+++ b/funasr/build_utils/build_model_from_file.py
@@ -6,7 +6,6 @@
import torch
import yaml
-from typeguard import check_argument_types
from funasr.build_utils.build_model import build_model
from funasr.models.base_model import FunASRModel
@@ -30,7 +29,6 @@
device: Device type, "cpu", "cuda", or "cuda:N".
"""
- assert check_argument_types()
if config_file is None:
assert model_file is not None, (
"The argument 'model_file' must be provided "
@@ -72,7 +70,14 @@
model.load_state_dict(model_dict)
else:
model_dict = torch.load(model_file, map_location=device)
- model.load_state_dict(model_dict)
+ if task_name == "ss":
+ model_dict = model_dict['model']
+ if task_name == "diar" and mode == "sond":
+ model_dict = fileter_model_dict(model_dict, model.state_dict())
+ if task_name == "vad":
+ model.encoder.load_state_dict(model_dict)
+ else:
+ model.load_state_dict(model_dict)
if model_name_pth is not None and not os.path.exists(model_name_pth):
torch.save(model_dict, model_name_pth)
logging.info("model_file is saved to pth: {}".format(model_name_pth))
@@ -85,7 +90,7 @@
ckpt,
mode,
):
- assert mode == "paraformer" or mode == "uniasr"
+ assert mode == "paraformer" or mode == "uniasr" or mode == "sond" or mode == "sv" or mode == "tp"
logging.info("start convert tf model to torch model")
from funasr.modules.streaming_utils.load_fr_tf import load_tf_dict
var_dict_tf = load_tf_dict(ckpt)
@@ -113,6 +118,49 @@
# stride_conv
var_dict_torch_update_local = model.stride_conv.convert_tf2torch(var_dict_tf, var_dict_torch)
var_dict_torch_update.update(var_dict_torch_update_local)
+ elif mode == "paraformer":
+ # encoder
+ var_dict_torch_update_local = model.encoder.convert_tf2torch(var_dict_tf, var_dict_torch)
+ var_dict_torch_update.update(var_dict_torch_update_local)
+ # predictor
+ var_dict_torch_update_local = model.predictor.convert_tf2torch(var_dict_tf, var_dict_torch)
+ var_dict_torch_update.update(var_dict_torch_update_local)
+ # decoder
+ var_dict_torch_update_local = model.decoder.convert_tf2torch(var_dict_tf, var_dict_torch)
+ var_dict_torch_update.update(var_dict_torch_update_local)
+ # bias_encoder
+ var_dict_torch_update_local = model.clas_convert_tf2torch(var_dict_tf, var_dict_torch)
+ var_dict_torch_update.update(var_dict_torch_update_local)
+ elif "mode" == "sond":
+ if model.encoder is not None:
+ var_dict_torch_update_local = model.encoder.convert_tf2torch(var_dict_tf, var_dict_torch)
+ var_dict_torch_update.update(var_dict_torch_update_local)
+ # speaker encoder
+ if model.speaker_encoder is not None:
+ var_dict_torch_update_local = model.speaker_encoder.convert_tf2torch(var_dict_tf, var_dict_torch)
+ var_dict_torch_update.update(var_dict_torch_update_local)
+ # cd scorer
+ if model.cd_scorer is not None:
+ var_dict_torch_update_local = model.cd_scorer.convert_tf2torch(var_dict_tf, var_dict_torch)
+ var_dict_torch_update.update(var_dict_torch_update_local)
+ # ci scorer
+ if model.ci_scorer is not None:
+ var_dict_torch_update_local = model.ci_scorer.convert_tf2torch(var_dict_tf, var_dict_torch)
+ var_dict_torch_update.update(var_dict_torch_update_local)
+ # decoder
+ if model.decoder is not None:
+ var_dict_torch_update_local = model.decoder.convert_tf2torch(var_dict_tf, var_dict_torch)
+ var_dict_torch_update.update(var_dict_torch_update_local)
+ elif "mode" == "sv":
+ # speech encoder
+ var_dict_torch_update_local = model.encoder.convert_tf2torch(var_dict_tf, var_dict_torch)
+ var_dict_torch_update.update(var_dict_torch_update_local)
+ # pooling layer
+ var_dict_torch_update_local = model.pooling_layer.convert_tf2torch(var_dict_tf, var_dict_torch)
+ var_dict_torch_update.update(var_dict_torch_update_local)
+ # decoder
+ var_dict_torch_update_local = model.decoder.convert_tf2torch(var_dict_tf, var_dict_torch)
+ var_dict_torch_update.update(var_dict_torch_update_local)
else:
# encoder
var_dict_torch_update_local = model.encoder.convert_tf2torch(var_dict_tf, var_dict_torch)
@@ -126,5 +174,20 @@
# bias_encoder
var_dict_torch_update_local = model.clas_convert_tf2torch(var_dict_tf, var_dict_torch)
var_dict_torch_update.update(var_dict_torch_update_local)
+ return var_dict_torch_update
return var_dict_torch_update
+
+
+def fileter_model_dict(src_dict: dict, dest_dict: dict):
+ from collections import OrderedDict
+ new_dict = OrderedDict()
+ for key, value in src_dict.items():
+ if key in dest_dict:
+ new_dict[key] = value
+ else:
+ logging.info("{} is no longer needed in this model.".format(key))
+ for key, value in dest_dict.items():
+ if key not in new_dict:
+ logging.warning("{} is missed in checkpoint.".format(key))
+ return new_dict
--
Gitblit v1.9.1