update users/raissi

rwth-i6 · Dec 18, 2024 · 75ee8ba · 75ee8ba
1 parent d9160f1
commit 75ee8ba
Show file tree

Hide file tree

Showing 6 changed files with 79 additions and 3 deletions.
diff --git a/users/raissi/experiments/domain_mismtach/medline/base_args.py b/users/raissi/experiments/domain_mismtach/medline/base_args.py
@@ -58,6 +58,16 @@ class DATASET:
     cached=True,
     hash_overwrite="GLOWTTS_V1_DEV_MED_055",
 )
+wmt22_medline_noise0551 = tk.Path(
+    ("/").join([PREPATH_CORPORA, "wmt22_medline_v1_sequiturg2p_glowtts460_noise055_seed1.xml.gz"]),
+    cached=True,
+    hash_overwrite="GLOWTTS_V1_DEV_MED_0551",
+)
+wmt22_medline_noise0552 = tk.Path(
+    ("/").join([PREPATH_CORPORA, "wmt22_medline_v1_sequiturg2p_glowtts460_noise055_seed2.xml.gz"]),
+    cached=True,
+    hash_overwrite="GLOWTTS_V1_DEV_MED_0552",
+)
 wmt22_medline_noise03 = tk.Path(
     ("/").join([PREPATH_CORPORA, "wmt22_medline_v1_sequiturg2p_glowtts460_noise03.xml.gz"]),
     cached=True,
@@ -119,6 +129,22 @@ class DATASET:
         lm=tk.Path(f"{PREPATH_ASR3}/lm/v2/medline+LBS/ufal_v1_mixlex_v2.lm.gz", cached=True, hash_overwrite="v21_lm"),
         description="lexicon uses both LBS and medline data with words repeating 3 or more",
     ),
+    0.55: DATASET(
+        lexicon_with_unk=tk.Path(
+            f"{PREPATH_ASR3}/lexicon/v2/medline+LBS/ufal_v1_mixlex_v2.rasr_with_unk.xml.gz",
+            cached=True,
+            hash_overwrite="v21_unk",
+        ),
+        lexicon_no_unk=tk.Path(
+            f"{PREPATH_ASR3}/lexicon/v2/medline+LBS/ufal_v1_mixlex_v2.rasr_without_unk.xml.gz",
+            cached=True,
+            hash_overwrite="v21_nounk",
+        ),
+        corpus=wmt22_medline_noise055,
+        lm=tk.Path(f"{PREPATH_ASR3}/lm/v2/medline+LBS/ufal_v1_mixlex_v2.lm.gz", cached=True, hash_overwrite="v21_lm"),
+        description="lexicon uses both LBS and medline data with words repeating 3 or more",
+    ),
+
     0.3: DATASET(
         lexicon_with_unk=tk.Path(
             f"{PREPATH_ASR3}/lexicon/v2/medline+LBS/ufal_v1_mixlex_v2.rasr_with_unk.xml.gz",
@@ -167,6 +193,36 @@ class DATASET:
         lm=tk.Path(f"{PREPATH_ASR3}/lm/v2/only_medline/ufal_v1_lm_3more.gz", cached=True, hash_overwrite="v22_lm"),
         description="lexicon uses only medline data with words repeating 3 or more",
     ),
+    0.551: DATASET(
+        lexicon_with_unk=tk.Path(
+            f"{PREPATH_ASR3}/lexicon/v2/only_medline/ufal_v1_3more_only.rasr_with_unk.xml.gz",
+            cached=True,
+            hash_overwrite="v22_unk",
+        ),
+        lexicon_no_unk=tk.Path(
+            f"{PREPATH_ASR3}/lexicon/v2/only_medline/ufal_v1_3more_only.rasr_without_unk.xml.gz",
+            cached=True,
+            hash_overwrite="v22_nounk",
+        ),
+        corpus=wmt22_medline_noise0551,
+        lm=tk.Path(f"{PREPATH_ASR3}/lm/v2/only_medline/ufal_v1_lm_3more.gz", cached=True, hash_overwrite="v22_lm"),
+        description="lexicon uses only medline data with words repeating 3 or more",
+    ),
+    0.552: DATASET(
+        lexicon_with_unk=tk.Path(
+            f"{PREPATH_ASR3}/lexicon/v2/only_medline/ufal_v1_3more_only.rasr_with_unk.xml.gz",
+            cached=True,
+            hash_overwrite="v22_unk",
+        ),
+        lexicon_no_unk=tk.Path(
+            f"{PREPATH_ASR3}/lexicon/v2/only_medline/ufal_v1_3more_only.rasr_without_unk.xml.gz",
+            cached=True,
+            hash_overwrite="v22_nounk",
+        ),
+        corpus=wmt22_medline_noise0552,
+        lm=tk.Path(f"{PREPATH_ASR3}/lm/v2/only_medline/ufal_v1_lm_3more.gz", cached=True, hash_overwrite="v22_lm"),
+        description="lexicon uses only medline data with words repeating 3 or more",
+    ),
     0.3: DATASET(
         lexicon_with_unk=tk.Path(
             f"{PREPATH_ASR3}/lexicon/v2/only_medline/ufal_v1_3more_only.rasr_with_unk.xml.gz",

diff --git a/users/raissi/experiments/tts_librispeech/base_args.py b/users/raissi/experiments/tts_librispeech/base_args.py
@@ -38,9 +38,11 @@
 dev_other_noise07 = tk.Path(("/").join([prepath_data, "dev-other_sequiturg2p_glowtts460_noise07.xml.gz"]),cached=True, hash_overwrite="GLOWTTS_V1_DEV_07")
 dev_other_noise03 = tk.Path(("/").join([prepath_data, "dev-other_sequiturg2p_glowtts460_noise03.xml.gz"]),cached=True, hash_overwrite="GLOWTTS_V1_DEV_03")
 dev_other_noise055 = tk.Path(("/").join([prepath_data, "dev-other_sequiturg2p_glowtts460_noise055.xml.gz"]),cached=True, hash_overwrite="GLOWTTS_V1_DEV_055")
+dev_other_noise0551 = tk.Path(("/").join([prepath_data, "dev-other_sequiturg2p_glowtts460_noise055_seed1.xml.gz"]),cached=True, hash_overwrite="GLOWTTS_V1_DEV_0551")
+dev_other_noise0552 = tk.Path(("/").join([prepath_data, "dev-other_sequiturg2p_glowtts460_noise055_seed2.xml.gz"]),cached=True, hash_overwrite="GLOWTTS_V1_DEV_0552")
 
 TTS_DEVOTHER = {
-    0.3: dev_other_noise03, 0.7: dev_other_noise07, 0.55: dev_other_noise055
+    0.3: dev_other_noise03, 0.7: dev_other_noise07, 0.55: dev_other_noise055, 0.551: dev_other_noise0551, 0.552: dev_other_noise0552,
 }
 
 

diff --git a/users/raissi/setups/common/TF_factored_hybrid_system.py b/users/raissi/setups/common/TF_factored_hybrid_system.py
@@ -1531,6 +1531,14 @@ def get_best_recog_scales_and_transition_values(
                             [v for v in np.arange(0.1, 0.6, 0.1).round(1)],
                         )
                     )
+                elif context_type == PhoneticContext.diphone:
+                    prior_scales = list(
+                        itertools.product(
+                            [v for v in np.arange(0.1, 0.7, 0.1).round(1)],
+                            [v for v in np.arange(0.1, 0.7, 0.1).round(1)],
+                        )
+                    )
+
                 else:
                     raise NotImplementedError("You were not supposed to run monophone decoding with factored decoder")
             else:
@@ -1551,6 +1559,7 @@ def get_best_recog_scales_and_transition_values(
             sil_tdp = (11.0, 0.0, "infinity", 20.0)
             sp_tdp = (8.0, 0.0, "infinity", 0.0)
 
+
         best_config_scales = recognizer.recognize_optimize_scales_v2(
             label_info=self.label_info,
             search_parameters=tune_args,

diff --git a/users/raissi/setups/common/decoder/BASE_factored_hybrid_search.py b/users/raissi/setups/common/decoder/BASE_factored_hybrid_search.py
@@ -982,6 +982,7 @@ def recognize(
             )
 
         lat2ctm_extra_config = rasr.RasrConfig()
+
         lat2ctm_extra_config.flf_lattice_tool.network.to_lemma.links = "best"
         lat2ctm = recog.LatticeToCtmJob(
             crp=search_crp,

diff --git a/users/raissi/setups/common/helpers/network/modify.py b/users/raissi/setups/common/helpers/network/modify.py
@@ -0,0 +1,8 @@
+import copy
+from typing import Dict
+
+
+def delete_right_context_branch(network: Dict):
+    for k in ["currentState", "linear1-triphone", "linear2-triphone", "right-output"]:
+        network.pop(k, None)
+    return network
diff --git a/users/raissi/setups/librispeech/decoder/LBS_factored_hybrid_search.py b/users/raissi/setups/librispeech/decoder/LBS_factored_hybrid_search.py
@@ -93,8 +93,8 @@ def get_kazuki_lstm_config(
 
         #model and graph info
         trafo_config.loader.type = "meta"
-        trafo_config.loader.meta_graph_file = tk.Path("/u/raissi/Desktop/debug/lstm_lm/mini.graph.meta", cached=True)
-        trafo_config.loader.saved_model_file = DelayedFormat("/work/asr4/rossenbach/custom_projects/kazuki_replicate_lm_training/net-model/network.029")
+        trafo_config.loader.meta_graph_file = tk.Path("/u/raissi/Desktop/debug/lstm_lm/mini2.3.graph.meta", cached=True)
+        trafo_config.loader.saved_model_file = DelayedFormat("//u/raissi/Desktop/debug/lstm_lm/models/net-model-mini2.2/network.030")
         trafo_config.loader.required_libraries = self.library_path
 
         trafo_config.type = "tfrnn"