Downgraded requirements for datasets and sacremoses, updated Thot

sillsdev · Oct 25, 2024 · 11c60ef · 11c60ef
1 parent 81c52e7
commit 11c60ef
Show file tree

Hide file tree

Showing 5 changed files with 33 additions and 41 deletions.
diff --git a/machine/translation/thot/simplex_model_weight_tuner.py b/machine/translation/thot/simplex_model_weight_tuner.py
@@ -83,11 +83,8 @@ def _generate_translations(
         try:
             model = load_smt_model(self._word_alignment_model_type, parameters)
             decoder = load_smt_decoder(model, parameters)
-            if decoder is not None:
-                translations = decoder.translate_batch([to_sentence(s) for s in source_corpus])
-                return [to_target_tokens(t.target) for t in translations]
-            else:
-                raise ValueError("Decoder could not be loaded.")
+            translations = decoder.translate_batch([to_sentence(s) for s in source_corpus])
+            return [to_target_tokens(t.target) for t in translations]
         finally:
             if decoder is not None:
                 decoder.clear()

diff --git a/machine/translation/thot/thot_smt_model_trainer.py b/machine/translation/thot/thot_smt_model_trainer.py
@@ -489,11 +489,7 @@ def _train_tune_corpus(
             for i in range(len(tune_source_corpus)):
                 if i > 0:
                     progress(ProgressStatus.from_step(i, len(tune_source_corpus)))
-                if decoder is None or smt_model is None:
-                    raise RuntimeError("Decoder or SMT model is None")
                 decoder.train_sentence_pair(to_sentence(tune_source_corpus[i]), to_sentence(tune_target_corpus[i]))
-            if smt_model is None:
-                raise RuntimeError("SMT model is None")
             smt_model.print_translation_model(parameters.translation_model_filename_prefix)
             smt_model.print_language_model(parameters.language_model_filename_prefix)
             progress(ProgressStatus.from_step(len(tune_source_corpus), len(tune_source_corpus)))

diff --git a/machine/translation/thot/thot_word_alignment_model_trainer.py b/machine/translation/thot/thot_word_alignment_model_trainer.py
@@ -80,8 +80,6 @@ def __init__(
             if model_type >= ThotWordAlignmentModelType.IBM2:
                 if parameters.get_hmm_iteration_count(model_type) > 0:
                     ibm2_or_hmm = ta.HmmAlignmentModel(ibm1)
-                    if ibm2_or_hmm is None:
-                        raise ValueError("ibm2_or_hmm should not be None")
                     if parameters.hmm_p0 is not None:
                         ibm2_or_hmm.hmm_p0 = parameters.hmm_p0
                     if parameters.hmm_lexical_smoothing_factor is not None:
@@ -100,8 +98,6 @@ def __init__(
                 and parameters.get_ibm3_iteration_count(model_type) > 0
             ):
                 ibm3 = ta.Ibm3AlignmentModel(ibm2_or_hmm)
-                if ibm3 is None:
-                    raise ValueError("ibm3 should not be None")
                 if parameters.ibm3_fertility_smoothing_factor is not None:
                     ibm3.fertility_smoothing_factor = parameters.ibm3_fertility_smoothing_factor
                 if parameters.ibm3_count_threshold is not None:

diff --git a/poetry.lock b/poetry.lock
diff --git a/pyproject.toml b/pyproject.toml
@@ -28,6 +28,7 @@ norecursedirs = "tests/testutils"
 [tool.pyright]
 typeCheckingMode = "basic"
 extraPaths = ["tests"]
+reportMissingModuleSource = false
 
 [tool.poetry]
 name = "sil-machine"
@@ -61,11 +62,11 @@ charset-normalizer = "^2.1.1"
 
 
 sentencepiece = "^0.2.0"
-sil-thot = "^3.4.4"
+sil-thot = "^3.4.5"
 
-transformers = "^4.38.0, <4.46.0"
-datasets = "^3.0.0"
-sacremoses = "^0.1.0"
+transformers = ">=4.38.0, <4.46.0"
+datasets = "^2.4.0"
+sacremoses = "^0.0.53"
 
 clearml = { extras = ["s3"], version = "^1.13.1" }
 botocore = "^1.35.41"