dan/manager/metrics.py

@@ -16,7 +16,6 @@ class MetricManager:
         if "simara" in dataset_name and "page" in dataset_name:
             self.post_processing_module = PostProcessingModuleSIMARA
             self.matching_tokens = SIMARA_MATCHING_TOKENS
-            self.edit_and_num_edge_nodes = edit_and_num_items_for_ged_from_str_simara
         else:
             self.matching_tokens = dict()
 
@@ -246,7 +245,7 @@ class MetricManager:
                     pp_pred.append(pp_module.post_process(pred))
                     metrics["nb_pp_op_layout"].append(pp_module.num_op)
                 metrics["nb_gt_layout_token"] = [
-                    len(keep_only_tokens(str_x, self.layout_tokens))
+                    len(keep_only_ner_tokens(str_x, self.layout_tokens))
                     for str_x in values["str_x"]
                 ]
                 edit_and_num_items = [
@@ -262,16 +261,16 @@ class MetricManager:
         return self.epoch_metrics[name]
 
 
-def keep_only_tokens(str, tokens):
+def keep_only_ner_tokens(str, tokens):
     """
-    Remove all but layout tokens from string
+    Remove all but ner tokens from string
     """
     return re.sub("([^" + tokens + "])", "", str)
 
 
-def keep_all_but_tokens(str, tokens):
+def keep_all_but_ner_tokens(str, tokens):
     """
-    Remove all layout tokens from string
+    Remove all ner tokens from string
     """
     return re.sub("([" + tokens + "])", "", str)
 
@@ -310,7 +309,7 @@ def format_string_for_wer(str, layout_tokens, remove_punct=False):
             r"([\[\]{}/\\()\"'&+*=<>?.;:,!\-—_€#%°])", "", str
         )  # remove punctuation
     if layout_tokens is not None:
-        str = keep_all_but_tokens(
+        str = keep_all_but_ner_tokens(
             str, layout_tokens
         )  # remove layout tokens from metric
     str = re.sub("([ \n])+", " ", str).strip()  # keep only one space character
@@ -322,7 +321,7 @@ def format_string_for_cer(str, layout_tokens):
     Format string for CER computation: remove layout tokens and extra spaces
     """
     if layout_tokens is not None:
-        str = keep_all_but_tokens(
+        str = keep_all_but_ner_tokens(
             str, layout_tokens
         )  # remove layout tokens from metric
     str = re.sub("([\n])+", "\n", str)  # remove consecutive line breaks
@@ -378,8 +377,8 @@ def compute_layout_precision_per_threshold(
         pred, begin_token, end_token, associated_score=score, order_by_score=True
     )
     gt_list = extract_by_tokens(gt, begin_token, end_token)
-    pred_list = [keep_all_but_tokens(p, layout_tokens) for p in pred_list]
-    gt_list = [keep_all_but_tokens(gt, layout_tokens) for gt in gt_list]
+    pred_list = [keep_all_but_ner_tokens(p, layout_tokens) for p in pred_list]
+    gt_list = [keep_all_but_ner_tokens(gt, layout_tokens) for gt in gt_list]
     precision_per_threshold = [
         compute_layout_AP_for_given_threshold(gt_list, pred_list, threshold / 100)
         for threshold in range(5, 51, 5)
@@ -514,7 +513,7 @@ def str_to_graph_simara(str):
     Compute graph from string of layout tokens for the SIMARA dataset at page level
     """
     begin_layout_tokens = "".join(list(SIMARA_MATCHING_TOKENS.keys()))
-    layout_token_sequence = keep_only_tokens(str, begin_layout_tokens)
+    layout_token_sequence = keep_only_ner_tokens(str, begin_layout_tokens)
     g = nx.DiGraph()
     g.add_node("D", type="document", level=2, page=0)
     token_name_dict = {"ⓘ": "I", "ⓓ": "D", "ⓢ": "S", "ⓒ": "C", "ⓟ": "P", "ⓐ": "A"}
@@ -549,16 +548,3 @@ def graph_edit_distance(g1, g2):
     ):
         new_edit = v
     return new_edit
-
-
-def edit_and_num_items_for_ged_from_str_simara(str_gt, str_pred):
-    """
-    Compute graph edit distance and num nodes/edges for normalized graph edit distance
-    For the SIMARA dataset
-    """
-    g_gt = str_to_graph_simara(str_gt)
-    g_pred = str_to_graph_simara(str_pred)
-    return (
-        graph_edit_distance(g_gt, g_pred),
-        g_gt.number_of_nodes() + g_gt.number_of_edges(),
-    )

dan/manager/training.py

@@ -4,8 +4,6 @@ import json
 import os
 import pickle
 import random
-import sys
-from datetime import date
 from time import time
 
 import numpy as np
@@ -523,7 +521,6 @@ class GenericTrainingManager:
             return
         params = copy.deepcopy(self.params)
         params = class_to_str_dict(params)
-        params["date"] = date.today().strftime("%d/%m/%Y")
         total_params = 0
         for model_name in self.models.keys():
             current_params = compute_nb_params(self.models[model_name])
@@ -533,21 +530,6 @@ class GenericTrainingManager:
             ]
             total_params += current_params
         params["model_params"]["total_params"] = "{:,}".format(total_params)
-
-        params["hardware"] = dict()
-        if self.device != "cpu":
-            for i in range(self.params["training_params"]["nb_gpu"]):
-                params["hardware"][str(i)] = "{} {}".format(
-                    torch.cuda.get_device_name(i), torch.cuda.get_device_properties(i)
-                )
-        else:
-            params["hardware"]["0"] = "CPU"
-        params["software"] = {
-            "python_version": sys.version,
-            "pytorch_version": torch.__version__,
-            "cuda_version": torch.version.cuda,
-            "cudnn_version": torch.backends.cudnn.version(),
-        }
         with open(path, "w") as f:
             json.dump(params, f, indent=4)
 
@@ -871,8 +853,9 @@ class GenericTrainingManager:
             with open(path, "w") as f:
                 yaml.dump(metrics, stream=f)
 
-            # Log mlflow artifacts
-            mlflow.log_artifact(path, "predictions")
+            if mlflow_logging:
+                # Log mlflow artifacts
+                mlflow.log_artifact(path, "predictions")
 
     def output_pred(self, name):
         path = os.path.join(
@@ -1104,14 +1087,7 @@ class Manager(OCRManager):
         reduced_size = [s[:2] for s in batch_data["imgs_reduced_shape"]]
         y_len = batch_data["labels_len"]
 
-        # add errors in teacher forcing
-        if (
-            "teacher_forcing_error_rate" in self.params["training_params"]
-            and self.params["training_params"]["teacher_forcing_error_rate"] is not None
-        ):
-            error_rate = self.params["training_params"]["teacher_forcing_error_rate"]
-            simulated_y_pred, y_len = self.add_label_noise(y, y_len, error_rate)
-        elif "label_noise_scheduler" in self.params["training_params"]:
+        if "label_noise_scheduler" in self.params["training_params"]:
             error_rate = (
                 self.params["training_params"]["label_noise_scheduler"][
                     "min_error_rate"