Hugging Face's logo

brianschwabauer
/
latent-space-language-diffusion-model

Text Generation
Safetensors
English
mdlm
masked-diffusion
parallel-decoding
diffusion
non-autoregressive
bidirectional-transformer
ar-oracle
qwen3
open-weights
reproducible
Model card Files
xet
 Community
latent-space-language-diffusion-model

Commit History

6 days ago

  1. Delete modeling_mdlm.py with huggingface_hub

    brianschwabauer
    Verified signature
    7254d81

  2. Upload results/sudoku/falsification_summary.json with huggingface_hub

    brianschwabauer
    Verified signature
    59b4c58

  3. Upload results/samples_v41.md with huggingface_hub

    brianschwabauer
    Verified signature
    14504c6

  4. Upload results/gen_bench_v41.json with huggingface_hub

    brianschwabauer
    Verified signature
    48be1db

  5. Upload model.safetensors with huggingface_hub

    brianschwabauer
    Verified signature
    f5a66a0

  6. Upload results/diag_v4_loss.json with huggingface_hub

    brianschwabauer
    Verified signature
    28b7592

  7. Upload train.py with huggingface_hub

    brianschwabauer
    Verified signature
    4661802

  8. Upload mdlm_v4_optimized.py with huggingface_hub

    brianschwabauer
    Verified signature
    8e71753

  9. Upload config.json with huggingface_hub

    brianschwabauer
    Verified signature
    90c6179

  10. Upload README.md with huggingface_hub

    brianschwabauer
    Verified signature
    aae81ef

  1. Upload inference_v4.py with huggingface_hub

    brianschwabauer
    Verified signature
    e7b6911

  2. Upload mdlm_v4_optimized.py with huggingface_hub

    brianschwabauer
    Verified signature
    4c1a3bd

  3. Upload tokenizer/bpe_tokenizer.json with huggingface_hub

    brianschwabauer
    Verified signature
    0ba78fc

  4. v4: RMSNorm, SwiGLU, ELBO loss, 500K docs — PPL 147, coherence 0.40

    brianschwabauer
    Verified signature
    9705efb

  1. Upload results/ar_control_training.json with huggingface_hub

    brianschwabauer
    Verified signature
    d7cc80a

  2. Upload results/comparison_benchmark.json with huggingface_hub

    brianschwabauer
    Verified signature
    41c0f29

  3. Upload scripts/benchmark_comparison.py with huggingface_hub

    brianschwabauer
    Verified signature
    55dbc14

  4. Upload scripts/train_ar.py with huggingface_hub

    brianschwabauer
    Verified signature
    62caf8e

  5. Upload src/ar_control.py with huggingface_hub

    brianschwabauer
    Verified signature
    83f4685

  6. Upload hf_model/ar_control_config.json with huggingface_hub

    brianschwabauer
    Verified signature
    3b4b12c

  7. Upload hf_model/ar_control.safetensors with huggingface_hub

    brianschwabauer
    Verified signature
    2694687

  1. docs: add badges, SEO keywords, GitHub cross-reference, extended tags

    brianschwabauer
    Verified signature
    64d78b1

  2. Full open release: source code + training script + detailed config + transparent model card. No black boxes.

    brianschwabauer
    Verified signature
    059f836

  3. Upload MDLM v3: 201M params, PPL 102.6, safetensors format

    brianschwabauer
    Verified signature
    d16a12a

  4. initial commit

    brianschwabauer
    Verified signature
    a954ded