From 611770b78d06f0bd5afb2f5b55da1a64026785f9 Mon Sep 17 00:00:00 2001 From: Jinzhe Zeng Date: Fri, 4 Aug 2023 17:19:01 -0400 Subject: [PATCH 1/3] change ckpt_prefix to relative path Before, `ckpt_prefix` is an absolute path and the generated symilnks link to an abosulote path. If users wants to move the whole directory, the relative path should be more user-friendly. Signed-off-by: Jinzhe Zeng --- deepmd/train/trainer.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/deepmd/train/trainer.py b/deepmd/train/trainer.py index 7145489959..4a49356592 100644 --- a/deepmd/train/trainer.py +++ b/deepmd/train/trainer.py @@ -811,7 +811,7 @@ def save_checkpoint(self, cur_batch: int): try: ckpt_prefix = self.saver.save( self.sess, - os.path.join(os.getcwd(), self.save_ckpt), + self.save_ckpt, global_step=cur_batch, ) except google.protobuf.message.DecodeError as e: From 3413c1f8dd602074bf4c0d9871567e22d4dacaff Mon Sep 17 00:00:00 2001 From: Jinzhe Zeng Date: Fri, 4 Aug 2023 17:19:52 -0400 Subject: [PATCH 2/3] Revert "change ckpt_prefix to relative path" This reverts commit 611770b78d06f0bd5afb2f5b55da1a64026785f9. --- deepmd/train/trainer.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/deepmd/train/trainer.py b/deepmd/train/trainer.py index 4a49356592..7145489959 100644 --- a/deepmd/train/trainer.py +++ b/deepmd/train/trainer.py @@ -811,7 +811,7 @@ def save_checkpoint(self, cur_batch: int): try: ckpt_prefix = self.saver.save( self.sess, - self.save_ckpt, + os.path.join(os.getcwd(), self.save_ckpt), global_step=cur_batch, ) except google.protobuf.message.DecodeError as e: From 66c14a2607c1e857e7d17041c4971bfea270442d Mon Sep 17 00:00:00 2001 From: Jinzhe Zeng Date: Fri, 4 Aug 2023 17:23:48 -0400 Subject: [PATCH 3/3] symlink `model.ckpt.*` to relative paths Before, `model.ckpt.*` is symilnks to an abosulote path. If users wants to move the whole directory, the relative path will be more user-friendly. Signed-off-by: Jinzhe Zeng --- deepmd/train/trainer.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/deepmd/train/trainer.py b/deepmd/train/trainer.py index 7145489959..a6ac96dab4 100644 --- a/deepmd/train/trainer.py +++ b/deepmd/train/trainer.py @@ -832,7 +832,7 @@ def save_checkpoint(self, cur_batch: int): pass if platform.system() != "Windows": # by default one does not have access to create symlink on Windows - os.symlink(ori_ff, new_ff) + os.symlink(os.path.relpath(ori_ff, os.path.dirname(new_ff)), new_ff) else: shutil.copyfile(ori_ff, new_ff) log.info("saved checkpoint %s" % self.save_ckpt)