Skip to content

Commit c655239

Browse files
committed
updates now finished
1 parent 1236faf commit c655239

File tree

4 files changed

+15
-10
lines changed

4 files changed

+15
-10
lines changed

common/baselines/tedlium2/default_tools.py

Lines changed: 7 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -8,18 +8,22 @@
88
version listed here. Nevertheless, the most recent "head" should be safe to be used as well.
99
"""
1010
from sisyphus import tk
11+
from i6_experiments.common.tools.rasr import compile_rasr_binaries_i6mode
1112
from i6_experiments.common.tools.sctk import compile_sctk
1213
from i6_core.tools.git import CloneGitRepositoryJob
1314

1415
PACKAGE = __package__
1516

16-
RASR_BINARY_PATH = tk.Path("/work/tools/asr/rasr/20211217_tf23_cuda101_mkl/arch/linux-x86_64-standard")
17+
RASR_BINARY_PATH = compile_rasr_binaries_i6mode(
18+
branch="bene_apptainer_tf213",
19+
configure_options=["--apptainer-setup=2023-08-29_tensorflow-2.13_v1"],
20+
) # use most recent RASR
1721
RASR_BINARY_PATH.hash_overwrite = "TEDLIUM2_DEFAULT_RASR_BINARY_PATH"
1822

1923
SCTK_BINARY_PATH = compile_sctk() # use last published version
2024
SCTK_BINARY_PATH.hash_overwrite = "TEDLIUM2_DEFAULT_SCTK_BINARY_PATH"
2125

22-
SRILM_PATH = tk.Path("/work/tools/users/luescher/srilm-1.7.3/bin/i686-m64/")
26+
SRILM_PATH = tk.Path("/work/tools22/users/luescher/srilm-1.7.3-app-u22/bin/i686-m64")
2327
SRILM_PATH.hash_overwrite = "TEDLIUM2_DEFAULT_SRILM_PATH"
2428

2529
RETURNN_EXE = tk.Path(
@@ -29,6 +33,6 @@
2933

3034
RETURNN_RC_ROOT = CloneGitRepositoryJob(
3135
"https://github.com/rwth-i6/returnn",
32-
commit="d7689b945b2fe781b3c79fbef9d82f018c7b11e8",
36+
commit="11d33468ad56a6c254168560c29e77e65eb45b7c",
3337
).out_repository
3438
RETURNN_RC_ROOT.hash_overwrite = "TEDLIUM2_DEFAULT_RETURNN_RC_ROOT"

common/baselines/tedlium2/hybrid/baseline_config.py

Lines changed: 1 addition & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -41,10 +41,7 @@ def run_tedlium2_hybrid_baseline():
4141
)
4242
# image only, so just python3
4343
returnn_exe = tk.Path("/usr/bin/python3", hash_overwrite="GENERIC_RETURNN_LAUNCHER")
44-
blas_lib = tk.Path(
45-
"/work/tools/asr/tensorflow/2.3.4-generic+cuda10.1+mkl/bazel_out/external/mkl_linux/lib/libmklml_intel.so",
46-
hash_overwrite="TF23_MKL_BLAS",
47-
)
44+
blas_lib = tk.Path("/lib/x86_64-linux-gnu/liblapack.so.3")
4845
blas_lib.hash_overwrite = "TEDLIUM2_DEFAULT_RASR_BINARY_PATH"
4946
steps = RasrSteps()
5047
steps.add_step("extract", rasr_init_args.feature_extraction_args)

common/datasets/tedlium2/corpus.py

Lines changed: 6 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -33,8 +33,12 @@ def get_bliss_corpus_dict(audio_format: str = "wav", output_prefix: str = "datas
3333
"codec": "pcm_s16le",
3434
"recover_duration": False,
3535
},
36-
"ogg": {"output_format": "ogg", "codec": "libvorbis"},
37-
"flac": {"output_format": "flac", "codec": "flac"},
36+
"ogg": {"output_format": "ogg", "codec": "libvorbis", "recover_duration": False},
37+
"flac": {
38+
"output_format": "flac",
39+
"codec": "flac",
40+
"recover_duration": False,
41+
},
3842
}
3943

4044
converted_bliss_corpus_dict = {}

common/datasets/tedlium2/export.py

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -71,7 +71,7 @@ def _export_lexicon(output_prefix: str = "datasets"):
7171
"""
7272
lexicon_output_prefix = os.path.join(output_prefix, TEDLIUM_PREFIX, "lexicon")
7373

74-
bliss_lexicon = get_bliss_lexicon(output_prefix=output_prefix)
74+
bliss_lexicon = get_bliss_lexicon(add_unknown_phoneme_and_mapping=False, output_prefix=output_prefix)
7575
tk.register_output(os.path.join(lexicon_output_prefix, "tedlium2.lexicon.xml.gz"), bliss_lexicon)
7676

7777
g2p_bliss_lexicon = get_g2p_augmented_bliss_lexicon(

0 commit comments

Comments
 (0)