From dc482e24fa1798ca910336fc5219941689ff897d Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?M=C3=A9lodie=20Boillet?= <boillet@teklia.com>
Date: Thu, 3 Aug 2023 08:37:22 +0200
Subject: [PATCH] Apply cb58891e
---
dan/manager/training.py | 3 ---
dan/ocr/document/train.py | 1 -
docs/usage/train/parameters.md | 1 -
tests/conftest.py | 1 -
4 files changed, 6 deletions(-)
diff --git a/dan/manager/training.py b/dan/manager/training.py
index f231e372..1eba8670 100644
--- a/dan/manager/training.py
+++ b/dan/manager/training.py
@@ -524,7 +524,6 @@ class GenericTrainingManager:
self.begin_time = time()
focus_metric_name = self.params["training_params"]["focus_metric"]
nb_epochs = self.params["training_params"]["max_nb_epochs"]
- interval_save_weights = self.params["training_params"]["interval_save_weights"]
metric_names = self.params["training_params"]["train_metrics"]
display_values = None
@@ -676,8 +675,6 @@ class GenericTrainingManager:
# save model weights
if self.is_master:
self.save_model(epoch=num_epoch, name="last")
- if interval_save_weights and num_epoch % interval_save_weights == 0:
- self.save_model(epoch=num_epoch, name="weights", keep_weights=True)
self.writer.flush()
def evaluate(self, set_name, mlflow_logging=False, **kwargs):
diff --git a/dan/ocr/document/train.py b/dan/ocr/document/train.py
index 97cae761..d4c8330a 100644
--- a/dan/ocr/document/train.py
+++ b/dan/ocr/document/train.py
@@ -164,7 +164,6 @@ def get_config():
* 24
* 1.9, # maximum time before to stop (in seconds)
"load_epoch": "last", # ["best", "last"]: last to continue training, best to evaluate
- "interval_save_weights": None, # None: keep best and last only
"batch_size": 2, # mini-batch size for training
"use_ddp": False, # Use DistributedDataParallel
"ddp_port": "20027",
diff --git a/docs/usage/train/parameters.md b/docs/usage/train/parameters.md
index 5dd48610..478d410f 100644
--- a/docs/usage/train/parameters.md
+++ b/docs/usage/train/parameters.md
@@ -157,7 +157,6 @@ For a detailed description of all augmentation transforms, see the [dedicated pa
| `training_params.max_nb_epochs` | Maximum number of epochs before stopping training. | `int` | `800` |
| `training_params.max_training_time` | Maximum time (in seconds) before stopping training. | `int` | `164160` |
| `training_params.load_epoch` | Model to load. Should be either `"best"` (evaluation) or `last` (training). | `str` | `"last"` |
-| `training_params.interval_save_weights` | Step to save weights. Set to `None` to keep only best and last epochs. | `int` | `None` |
| `training_params.batch_size` | Mini-batch size for the training loop. | `int` | `2` |
| `training_params.use_ddp` | Whether to use DistributedDataParallel. | `bool` | `False` |
| `training_params.ddp_port` | DDP port. | `int` | `20027` |
diff --git a/tests/conftest.py b/tests/conftest.py
index 4a365be7..d85d11b4 100644
--- a/tests/conftest.py
+++ b/tests/conftest.py
@@ -110,7 +110,6 @@ def training_config():
"max_nb_epochs": 4, # maximum number of epochs before to stop
"max_training_time": 1200, # maximum time before to stop (in seconds)
"load_epoch": "last", # ["best", "last"]: last to continue training, best to evaluate
- "interval_save_weights": None, # None: keep best and last only
"batch_size": 2, # mini-batch size for training
"use_ddp": False, # Use DistributedDataParallel
"nb_gpu": 0,
--
GitLab