chaque modèle a son dossier

mb.py

@@ -7,6 +7,7 @@ import matplotlib.pyplot as plt
 from sklearn.metrics import pairwise
 from tokenizers import Tokenizer
 from torch.utils.data import DataLoader
+from pathlib import Path
 
 
 from operator import is_not
@@ -31,10 +32,10 @@ from eval import *
 #     "lemmatized": liste de chaines de caractères
 #   }
 # }
-def get_acr_def_dict(model, tokenizer, acronyms, device="cuda"):
-    definitions_file = "output/minibert/definitions.tar"
+def get_acr_def_dict(model, tokenizer, acronyms, outdir, device="cuda"):
+    definitions_file = os.path.join(outdir, "definitions.tar")
 
-    if os.path.exists(definitions_file) and False:
+    if os.path.exists(definitions_file):
         definitions = torch.load(definitions_file)
     else:
         definitions = {}
@@ -59,7 +60,7 @@ def get_acr_def_dict(model, tokenizer, acronyms, device="cuda"):
     return definitions
 
 
-def minibert_wsd(model_path, glob):
+def minibert_wsd(model_path, glob, outdir):
     device = "cuda"
     pin_memory = device != "cpu"
 
@@ -83,7 +84,7 @@ def minibert_wsd(model_path, glob):
     model.load_state_dict(checkpoint["model_state_dict"])
     model.eval()
 
-    definitions = get_acr_def_dict(model, tokenizer, acronyms, device)
+    definitions = get_acr_def_dict(model, tokenizer, acronyms, outdir, device)
 
     if glob:
         all_defs = []
@@ -119,34 +120,35 @@ def minibert_wsd(model_path, glob):
                 idef = torch.argmax(dists).item()
                 annotated[isent]["acronymes"][iacr]["prediction"] = definitions[acr]["lemmatized"][idef]
 
-    fname = os.path.basename(os.path.dirname(model_path))
     if glob:
-        predictions_path = f"output/minibert/predictions_{fname}_glob.json"
+        predictions_path = os.path.join(outdir, "predictions_glob.json")
     else:
-        predictions_path = f"output/minibert/predictions_{fname}.json"
+        predictions_path = os.path.join(outdir, "predictions.json")
     with open(predictions_path, "w", encoding="UTF-8") as f:
         f.write(json.dumps(annotated, indent=4, ensure_ascii=False))
 
 def all_minibert_wsd(args):
     for md in os.listdir(args.path):
         cp_path = os.path.join(args.path, md, "checkpoint-00100.tar")
-        minibert_wsd(cp_path, args.glob)
+        outdir = os.path.join("output", "minibert", md)
+        Path(outdir).mkdir(exist_ok=True, parents=True)
+        minibert_wsd(cp_path, args.glob, outdir)
 
 def eval_minibert(args):
-    pred_dir = "output/minibert"
-    pred_files = [f for f in os.listdir(pred_dir) if f.startswith("predictions_")]
+    pred_files = Path("output/minibert").glob("**/predictions*.json")
 
     resd = {
         "file": [],
-        "pos": []
+        "pos": [],
+        "glob": []
     }
 
     for f in pred_files:
-        pred_path = os.path.join(pred_dir, f)
-        annot = load_annot(pred_path)
+        annot = load_annot(str(f))
         prec, rapp, prm = acc(count(annot))
-        resd["file"].append(f)
+        resd["file"].append(f.parent.name)
         resd["pos"].append(prec)
+        resd["glob"].append(str(f).endswith("_glob.json"))
     df = pd.DataFrame(resd)
     df.sort_values("pos", inplace=True, ascending=False)
     df.to_csv("output/minibert/scores.csv", index=False)