From c2dee5e3c29eba79e591d9e9caebaef15ea4e56b Mon Sep 17 00:00:00 2001
From: hnluo <haoneng.lhn@alibaba-inc.com>
Date: 星期四, 29 六月 2023 11:09:28 +0800
Subject: [PATCH] Merge pull request #687 from alibaba-damo-academy/dev_lhn
---
funasr/train/trainer.py | 19 +++++++++++++++++++
1 files changed, 19 insertions(+), 0 deletions(-)
diff --git a/funasr/train/trainer.py b/funasr/train/trainer.py
index 4f83ace..f066909 100644
--- a/funasr/train/trainer.py
+++ b/funasr/train/trainer.py
@@ -44,6 +44,7 @@
from funasr.train.reporter import Reporter
from funasr.train.reporter import SubReporter
from funasr.utils.build_dataclass import build_dataclass
+from funasr.utils.kwargs2args import kwargs2args
if torch.distributed.is_available():
from torch.distributed import ReduceOp
@@ -620,6 +621,24 @@
all_steps_are_invalid = False
continue
+ if iiter == 1 and summary_writer is not None:
+ try:
+ args = kwargs2args(model.forward, batch)
+ except (ValueError, TypeError):
+ logging.warning(
+ "inpect.signature() is failed for the model. "
+ "The graph can't be added for tensorboard."
+ )
+ else:
+ try:
+ summary_writer.add_graph(model, args, use_strict_trace=False)
+ except Exception:
+ logging.warning(
+ "summary_writer.add_graph() is failed for the model. "
+ "The graph can't be added for tensorboard."
+ )
+ del args
+
with autocast(scaler is not None):
with reporter.measure_time("forward_time"):
retval = model(**batch)
--
Gitblit v1.9.1