Fix librispeech CI test errors

.github/scripts/docker/generate_build_matrix.py

@@ -91,7 +91,7 @@ def get_matrix(min_torch_version, specified_torch_version, specified_python_vers
     matrix = []
     for p in python_version:
         for t in torch_version:
-            if min_torch_version and version_gt(min_torch_version, t):
+            if min_torch_version and version_ge(min_torch_version, t):
                 continue
 
             # torchaudio <= 1.13.x supports only python <= 3.10

.github/workflows/librispeech.yml

@@ -30,7 +30,7 @@ jobs:
         run: |
           # outputting for debugging purposes
           python ./.github/scripts/docker/generate_build_matrix.py --python-version "3.10"
-          MATRIX=$(python ./.github/scripts/docker/generate_build_matrix.py --python-version "3.10")
+          MATRIX=$(python ./.github/scripts/docker/generate_build_matrix.py --python-version "3.10" --min-torch-version "2.6.0")
           echo "::set-output name=matrix::${MATRIX}"
   librispeech:
     needs: generate_build_matrix

.github/workflows/test.yml

@@ -30,7 +30,8 @@ jobs:
         id: set-matrix
         run: |
           # outputting for debugging purposes
-          MATRIX=$(python ./.github/scripts/docker/generate_build_matrix.py --min-torch-version "2.3")
+          python ./.github/scripts/docker/generate_build_matrix.py --python-version "3.10"
+          MATRIX=$(python ./.github/scripts/docker/generate_build_matrix.py --python-version "3.10")
           echo "::set-output name=matrix::${MATRIX}"
   test:
     needs: generate_build_matrix

.github/workflows/yesno.yml

@@ -31,7 +31,8 @@ jobs:
         run: |
           # outputting for debugging purposes
           python ./.github/scripts/docker/generate_build_matrix.py --python-version "3.10"
-          MATRIX=$(python ./.github/scripts/docker/generate_build_matrix.py --python-version "3.10")
+          # MATRIX=$(python ./.github/scripts/docker/generate_build_matrix.py --python-version "3.10")
+          MATRIX=$(python ./.github/scripts/docker/generate_build_matrix.py --python-version "3.10" --min-torch-version "2.5.0")
           echo "::set-output name=matrix::${MATRIX}"
   yesno:
     needs: generate_build_matrix

egs/librispeech/ASR/pruned_transducer_stateless7_ctc/ctc_decode.py

@@ -633,7 +633,9 @@ def main():
         H = None
         bpe_model = 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
 
@@ -672,7 +674,9 @@ def main():
                 torch.save(G.as_dict(), params.lm_dir / "G_4_gram.pt")
         else:
             logging.info("Loading pre-compiled G_4_gram.pt")
-            d = torch.load(params.lm_dir / "G_4_gram.pt", map_location=device)
+            d = torch.load(
+                params.lm_dir / "G_4_gram.pt", map_location=device, weights_only=False
+            )
             G = k2.Fsa.from_dict(d)
 
         if params.decoding_method == "whole-lattice-rescoring":

egs/librispeech/ASR/pruned_transducer_stateless7_ctc/decode.py

@@ -786,7 +786,7 @@ 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:

egs/librispeech/ASR/pruned_transducer_stateless7_ctc/jit_pretrained_ctc.py

@@ -347,7 +347,9 @@ def main():
         "whole-lattice-rescoring",
     ]:
         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
@@ -358,7 +360,9 @@ def main():
             "whole-lattice-rescoring",
         ]:
             logging.info(f"Loading G from {params.G}")
-            G = k2.Fsa.from_dict(torch.load(params.G, map_location="cpu"))
+            G = k2.Fsa.from_dict(
+                torch.load(params.G, map_location="cpu", weights_only=False)
+            )
             G = G.to(device)
             if params.method == "whole-lattice-rescoring":
                 # Add epsilon self-loops to G as we will compose

egs/librispeech/ASR/pruned_transducer_stateless7_ctc/pretrained.py

@@ -247,7 +247,7 @@ def main():
     num_param = sum([p.numel() for p in model.parameters()])
     logging.info(f"Number of model parameters: {num_param}")
 
-    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/librispeech/ASR/pruned_transducer_stateless7_ctc/pretrained_ctc.py

@@ -286,7 +286,7 @@ def main():
     num_param = sum([p.numel() for p in model.parameters()])
     logging.info(f"Number of model parameters: {num_param}")
 
-    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/librispeech/ASR/pruned_transducer_stateless7_streaming/decode.py

@@ -936,7 +936,7 @@ 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:

egs/librispeech/ASR/pruned_transducer_stateless7_streaming/pretrained.py

@@ -247,7 +247,7 @@ def main():
     num_param = sum([p.numel() for p in model.parameters()])
     logging.info(f"Number of model parameters: {num_param}")
 
-    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/librispeech/ASR/zipformer/ctc_decode.py

@@ -947,7 +947,9 @@ def main():
         H = None
         bpe_model = 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
 
@@ -987,7 +989,9 @@ def main():
                 torch.save(G.as_dict(), params.lm_dir / "G_4_gram.pt")
         else:
             logging.info("Loading pre-compiled G_4_gram.pt")
-            d = torch.load(params.lm_dir / "G_4_gram.pt", map_location=device)
+            d = torch.load(
+                params.lm_dir / "G_4_gram.pt", map_location=device, weights_only=False
+            )
             G = k2.Fsa.from_dict(d)
 
         if params.decoding_method in [

egs/librispeech/ASR/zipformer/decode.py

@@ -1013,7 +1013,7 @@ 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:

egs/librispeech/ASR/zipformer/decode_gigaspeech.py

@@ -1049,7 +1049,7 @@ 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:

egs/librispeech/ASR/zipformer/finetune.py

@@ -765,7 +765,7 @@ def load_model_params(
 
     """
     logging.info(f"Loading checkpoint from {ckpt}")
-    checkpoint = torch.load(ckpt, map_location="cpu")
+    checkpoint = torch.load(ckpt, map_location="cpu", weights_only=False)
 
     # if module list is empty, load the whole model from ckpt
     if not init_modules:

egs/librispeech/ASR/zipformer/jit_pretrained_ctc.py

@@ -346,7 +346,9 @@ def main():
         "whole-lattice-rescoring",
     ]:
         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
@@ -357,7 +359,9 @@ def main():
             "whole-lattice-rescoring",
         ]:
             logging.info(f"Loading G from {params.G}")
-            G = k2.Fsa.from_dict(torch.load(params.G, map_location="cpu"))
+            G = k2.Fsa.from_dict(
+                torch.load(params.G, map_location="cpu", weights_only=False)
+            )
             G = G.to(device)
             if params.method == "whole-lattice-rescoring":
                 # Add epsilon self-loops to G as we will compose

egs/librispeech/ASR/zipformer/pretrained.py

@@ -289,7 +289,7 @@ def main():
     num_param = sum([p.numel() for p in model.parameters()])
     logging.info(f"Number of model parameters: {num_param}")
 
-    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/librispeech/ASR/zipformer/pretrained_ctc.py

@@ -305,7 +305,7 @@ def main():
     num_param = sum([p.numel() for p in model.parameters()])
     logging.info(f"Number of model parameters: {num_param}")
 
-    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()
@@ -389,7 +389,9 @@ def main():
         "whole-lattice-rescoring",
     ]:
         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
@@ -400,7 +402,9 @@ def main():
             "whole-lattice-rescoring",
         ]:
             logging.info(f"Loading G from {params.G}")
-            G = k2.Fsa.from_dict(torch.load(params.G, map_location="cpu"))
+            G = k2.Fsa.from_dict(
+                torch.load(params.G, map_location="cpu", weights_only=False)
+            )
             G = G.to(device)
             if params.method == "whole-lattice-rescoring":
                 # Add epsilon self-loops to G as we will compose

egs/librispeech/ASR/zipformer_ctc/decode.py

@@ -679,7 +679,9 @@ def main():
         H = None
         bpe_model = 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
 
@@ -719,7 +721,9 @@ def main():
                 torch.save(G.as_dict(), params.lm_dir / "G_4_gram.pt")
         else:
             logging.info("Loading pre-compiled G_4_gram.pt")
-            d = torch.load(params.lm_dir / "G_4_gram.pt", map_location=device)
+            d = torch.load(
+                params.lm_dir / "G_4_gram.pt", map_location=device, weights_only=False
+            )
             G = k2.Fsa.from_dict(d)
 
         if params.method in [

egs/yesno/ASR/local/compile_hlg.py

@@ -47,7 +47,7 @@ def compile_HLG(lang_dir: str) -> k2.Fsa:
     max_token_id = max(lexicon.tokens)
     logging.info(f"Building ctc_topo. max_token_id: {max_token_id}")
     H = k2.ctc_topo(max_token_id)
-    L = k2.Fsa.from_dict(torch.load(f"{lang_dir}/L_disambig.pt"))
+    L = k2.Fsa.from_dict(torch.load(f"{lang_dir}/L_disambig.pt", weights_only=False))
 
     logging.info("Loading G.fst.txt")
     with open("data/lm/G.fst.txt") as f:

egs/yesno/ASR/tdnn/decode.py

@@ -271,7 +271,9 @@ def main():
 
     logging.info(f"device: {device}")
 
-    HLG = k2.Fsa.from_dict(torch.load(f"{params.lang_dir}/HLG.pt", map_location="cpu"))
+    HLG = k2.Fsa.from_dict(
+        torch.load(f"{params.lang_dir}/HLG.pt", map_location="cpu", weights_only=False)
+    )
     HLG = HLG.to(device)
     assert HLG.requires_grad is False
 

egs/yesno/ASR/tdnn/jit_pretrained.py

@@ -131,7 +131,9 @@ def main():
     model.to(device)
 
     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)
 
     logging.info("Constructing Fbank computer")

egs/yesno/ASR/tdnn/onnx_pretrained.py

@@ -176,7 +176,9 @@ def main():
     model = OnnxModel(params.nn_model)
 
     logging.info(f"Loading HLG from {args.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)
 
     logging.info("Constructing Fbank computer")

egs/yesno/ASR/tdnn/pretrained.py

@@ -148,13 +148,15 @@ def main():
         num_classes=params.num_classes,
     )
 
-    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"])
     model.to(device)
     model.eval()
 
     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)
 
     logging.info("Constructing Fbank computer")