add weights_only=False to torch.load (#1984)

Author: teowenshen

docs/source/for-dummies/model-export.rst

@@ -41,7 +41,7 @@ To give you an idea of what ``tdnn/exp/pretrained.pt`` contains, we can use the
 .. code-block:: python3
 
     >>> import torch
-    >>> m = torch.load("tdnn/exp/pretrained.pt")
+    >>> m = torch.load("tdnn/exp/pretrained.pt", weights_only=False)
     >>> list(m.keys())
     ['model']
     >>> list(m["model"].keys())

egs/aidatatang_200zh/ASR/local/prepare_lang.py

@@ -28,7 +28,7 @@
 
 4. Generate L.pt, in k2 format. It can be loaded by
 
-        d = torch.load("L.pt")
+        d = torch.load("L.pt", weights_only=False)
         lexicon = k2.Fsa.from_dict(d)
 
 5. Generate L_disambig.pt, in k2 format.

egs/aidatatang_200zh/ASR/pruned_transducer_stateless2/pretrained.py

@@ -224,7 +224,7 @@ def main():
     logging.info("Creating model")
     model = get_transducer_model(params)
 
-    checkpoint = torch.load(args.checkpoint, map_location="cpu")
+    checkpoint = torch.load(args.checkpoint, map_location="cpu", weights_only=False)
     model.load_state_dict(checkpoint["model"], strict=False)
     model.to(device)
     model.eval()

egs/aishell/ASR/conformer_ctc/decode.py

@@ -503,7 +503,7 @@ def main():
     else:
         H = None
         HLG = k2.Fsa.from_dict(
-            torch.load(f"{params.lang_dir}/HLG.pt", map_location=device)
+            torch.load(f"{params.lang_dir}/HLG.pt", map_location=device, weights_only=False)
         )
         assert HLG.requires_grad is False
 

egs/aishell/ASR/conformer_ctc/pretrained.py

@@ -249,7 +249,7 @@ def main():
         use_feat_batchnorm=params.use_feat_batchnorm,
     )
 
-    checkpoint = torch.load(args.checkpoint, map_location="cpu")
+    checkpoint = torch.load(args.checkpoint, map_location="cpu", weights_only=False)
     model.load_state_dict(checkpoint["model"], strict=False)
     model.to(device)
     model.eval()
@@ -315,7 +315,7 @@ def main():
         hyps = [[token_sym_table[i] for i in ids] for ids in token_ids]
     elif params.method in ["1best", "attention-decoder"]:
         logging.info(f"Loading HLG from {params.HLG}")
-        HLG = k2.Fsa.from_dict(torch.load(params.HLG, map_location="cpu"))
+        HLG = k2.Fsa.from_dict(torch.load(params.HLG, map_location="cpu", weights_only=False))
         HLG = HLG.to(device)
         if not hasattr(HLG, "lm_scores"):
             # For whole-lattice-rescoring and attention-decoder

egs/aishell/ASR/conformer_mmi/decode.py

@@ -516,7 +516,7 @@ def main():
     else:
         H = None
         HLG = k2.Fsa.from_dict(
-            torch.load(f"{params.lang_dir}/HLG.pt", map_location=device)
+            torch.load(f"{params.lang_dir}/HLG.pt", map_location=device, weights_only=False)
         )
         assert HLG.requires_grad is False
 

egs/aishell/ASR/local/prepare_lang.py

@@ -28,7 +28,7 @@
 
 4. Generate L.pt, in k2 format. It can be loaded by
 
-        d = torch.load("L.pt")
+        d = torch.load("L.pt", weights_only=False)
         lexicon = k2.Fsa.from_dict(d)
 
 5. Generate L_disambig.pt, in k2 format.

egs/aishell/ASR/pruned_transducer_stateless2/pretrained.py

@@ -227,7 +227,7 @@ def main():
     logging.info("About to create model")
     model = get_transducer_model(params)
 
-    checkpoint = torch.load(args.checkpoint, map_location="cpu")
+    checkpoint = torch.load(args.checkpoint, map_location="cpu", weights_only=False)
     model.load_state_dict(checkpoint["model"], strict=False)
     model.to(device)
     model.eval()

egs/aishell/ASR/pruned_transducer_stateless3/pretrained.py

@@ -228,7 +228,7 @@ def main():
     logging.info("About to create model")
     model = get_transducer_model(params)
 
-    checkpoint = torch.load(args.checkpoint, map_location="cpu")
+    checkpoint = torch.load(args.checkpoint, map_location="cpu", weights_only=False)
     model.load_state_dict(checkpoint["model"], strict=False)
     model.to(device)
     model.eval()

egs/aishell/ASR/pruned_transducer_stateless7_bbpe/decode.py

@@ -773,7 +773,7 @@ def main():
             lg_filename = params.lang_dir / "LG.pt"
             logging.info(f"Loading {lg_filename}")
             decoding_graph = k2.Fsa.from_dict(
-                torch.load(lg_filename, map_location=device)
+                torch.load(lg_filename, map_location=device, weights_only=False)
             )
             decoding_graph.scores *= params.ngram_lm_scale
         else: