Update package to v0.3.0 with training exports

src/domain_tokenizer/__init__.py

@@ -6,21 +6,20 @@ Core components:
   - tokenizers: DomainTokenizerBuilder, per-field tokenizers
   - schemas: Predefined schemas (FINANCE, ECOMMERCE, HEALTHCARE)
   - models: DomainTransformerForCausalLM, PLR, JointFusion
+  - training: prepare_clm_dataset, pretrain_domain_model
 """
 
 from .schema import DomainSchema, FieldSpec, FieldType
 from .tokenizers.domain_tokenizer import DomainTokenizerBuilder
 from .tokenizers.field_tokenizers import (
-    BaseFieldTokenizer,
-    CalendarTokenizer,
-    CategoricalTokenizer,
-    DiscreteNumericalTokenizer,
-    MagnitudeBucketTokenizer,
-    SignTokenizer,
+    BaseFieldTokenizer, CalendarTokenizer, CategoricalTokenizer,
+    DiscreteNumericalTokenizer, MagnitudeBucketTokenizer, SignTokenizer,
 )
 from .models.configuration import DomainTransformerConfig
 from .models.modeling import DomainTransformerForCausalLM, DomainTransformerModel
 from .models.plr_embeddings import PeriodicLinearReLU
 from .models.joint_fusion import JointFusionModel, DCNv2
+from .training.data_pipeline import prepare_clm_dataset, pack_sequences
+from .training.pretrain import pretrain_domain_model
 
-__version__ = "0.2.0"
+__version__ = "0.3.0"