Commit History

  1. Fix label leakage: temporal split — use first 70% of events as input, predict purchase in last 30%. Remove n_purchases/purchase_rate from features.

    Verified signature
    e4d8561
  2. Fix model loading: use from_pretrained() instead of torch.load() for safetensors format

    Verified signature
    165b138
  3. Add 03_ecommerce_finetune.ipynb — next-purchase prediction with JointFusion, LightGBM baseline comparison

    Verified signature
    857ec9a
  4. Add e-commerce pre-training report — successful demo, behavioral clusters found, future improvements noted

    Verified signature
    2b3e3af
  5. Update 02_ecommerce notebook: add HF login, memory-free cell, subsample option for <64GB RAM machines

    Verified signature
    2410b7e
  6. CRITICAL FIX: Switch from ByteLevel to Whitespace pre-tokenizer — fixes 42% UNK rate on domain token sequences⁠

    Verified signature
    a9c4a62

  1. Add 02_ecommerce_pretrain.ipynb — REES46 e-commerce pre-training with sequential entropy check, wandb, push to hub

    Verified signature
    d60868a
  2. Add finance pre-training report — honest analysis of results and lessons learned

    Verified signature
    709a7e2
  3. Add .gitignore — Python, Jupyter, training artifacts, IDE files

    Verified signature
    9211898
  4. Fix notebook: total_mem → total_memory, add hub_model_id push, add wandb logging support

    Verified signature
    65ecf7e
  5. Add 01_finance_pretrain.ipynb — Phase 3.1 notebook for pre-training on 5M Nigerian financial transactions⁠

    Verified signature
    2c3ddfa
  6. Phase 3.0: Pipeline validation demo on mindweave/bank-transactions-us — ALL 10 CHECKS PASSED⁠

    Verified signature
    6e5b80d
  7. Add ADR-002: Dataset selection for Phase 3 demos — research findings, rationale, phased plan

    Verified signature
    756d197

  1. Update implementation report: add Phase 2D, update header to v0.4.0 / 139 tests, update cumulative summary and API

    Verified signature
    7aac458
  2. Add fine-tuning test suite — 15 tests covering dataset, batching, forward/backward, Trainer smoke, multiclass

    Verified signature
    abab711
  3. Update package to v0.4.0 with fine-tuning exports

    Verified signature
    7edb04f
  4. Update training init with finetune exports

    Verified signature
    64d55e2
  5. Add finetune.py — finetune_domain_model (HF Trainer Pattern A, auto tabular_features passthrough)

    Verified signature
    46a6d37
  6. Phase 2D: Fine-tuning pipeline — DomainFinetuneDataset, finetune_domain_model, 139 total tests passing⁠

    Verified signature
    256963c
  7. Update README v0.3.0 — add usage example, update roadmap status, add implementation report link

    Verified signature
    f580186
  8. Add Phase 2A-2C implementation report — technical decisions, architecture summary, test results

    Verified signature
    6c4ad4d
  9. Add training test suite — 19 tests covering data pipeline, packing, collation, integration, Trainer smoke test

    Verified signature
    345d9e3
  10. Update package to v0.3.0 with training exports

    Verified signature
    af3b720
  11. Add pretrain.py — pretrain_domain_model with HF Trainer, cosine schedule, DataCollatorForLanguageModeling

    Verified signature
    6ccb9e6
  12. Add data_pipeline.py — tokenize_user_sequences, pack_sequences, prepare_clm_dataset

    Verified signature
    1dfd4e2
  13. Phase 2C: Pre-training pipeline — data pipeline, sequence packing, HF Trainer CLM, 124 total tests passing⁠

    Verified signature
    28118c7
  14. Add model test suite — 33 tests covering config, model, PLR, DCNv2, joint fusion, integration

    Verified signature
    ab8a8b6
  15. Update package init to v0.2.0 with model exports

    Verified signature
    b86b1ee
  16. Add DCNv2 + JointFusionModel (nuFormer-style Transformer + tabular fusion)

    Verified signature
    e881ea3
  17. Add PLR embeddings (Gorishniy et al. 2022)

    Verified signature
    d685c0e
  18. Add DomainTransformerForCausalLM — GPT-style NoPE model with SDPA attention, weight tying, HF Trainer compatible

    Verified signature
    0dec8e4
  19. Add DomainTransformerConfig with presets (24M/85M/330M)

    Verified signature
    15fbfea
  20. Phase 2B: Model architecture — DomainTransformerForCausalLM (NoPE, GPT-style), PLR embeddings, DCNv2 + JointFusion, 105 passing tests⁠

    Verified signature
    2f5969e
  21. Add comprehensive test suite — 72 passing tests covering all components

    Verified signature
    8efa945
  22. Add predefined schemas (FINANCE, ECOMMERCE, HEALTHCARE)

    Verified signature
    c00ac2c
  23. Add domain_tokenizer.py — DomainTokenizerBuilder (core assembler, HF integration)

    Verified signature
    818a2e9
  24. Add field_tokenizers.py — Sign, MagnitudeBucket, Calendar, Categorical, DiscreteNumerical tokenizers

    Verified signature
    511f3aa
  25. Add tokenizers package init

    Verified signature
    0b06df3
  26. Add schemas package init

    Verified signature
    04b1d24
  27. Add schema.py — DomainSchema, FieldSpec, FieldType definitions

    Verified signature
    1a9dad0
  28. Phase 2A: Core tokenizer library — schema, field tokenizers, composite builder, predefined schemas, 72 passing tests⁠

    Verified signature
    0c1ca58
  29. Update README: add ADR reference, update documentation table and repo structure

    Verified signature
    a239d6e
  30. Add ADR-001: Implementation framework decision with detailed roadmap

    Verified signature
    25a1093
  31. Update README with Nubank case study and expanded repo structure

    Verified signature
    e30a14d
  32. Add Nubank nuFormer reverse-engineering analysis — full pipeline reconstruction

    Verified signature
    51149fa
  33. Add README with project overview and vision

    Verified signature
    f930fef
  34. Add comprehensive research report on domain-specific tokenization

    Verified signature
    be86e60
  35. initial commit

    Verified signature
    356a72e