Skip to main content

SID reglog parsing, tokenization, and macro transforms extracted from the preframr research codebase.

Project description

preframr-tokens

SID reglog parsing, tokenization, and macro transforms extracted from the preframr research codebase.

Torch-free. The training-side concerns (model, loss, DataLoader, predict) live in the main preframr repo; this package contains the stable parsing + encoding layer that produces the parsed parquets + unigram tokenizer alphabet that downstream training consumes.

Install

pip install preframr-tokens

Modules

  • preframr_tokens.reglogparser -- SID dump → parsed dataframe pipeline. RegLogParser.
  • preframr_tokens.regtokenizer -- alphabet build + unigram tokenizer fit. RegTokenizer.
  • preframr_tokens.macros.* -- declarative Transform registry plus the macro / pre-norm passes (slope, preset, hard_restart, legato_per_cluster, voice_block_order, ctrl_bigram, loop, etc.). Macros declare OP_CODES, LOSS_TIER, SUBSTITUTABLE_OPS, MUST_FOLLOW, etc. on their classes; pipeline_check.validate_pipeline_spec validates a pipeline declaratively.
  • preframr_tokens.stfconstants -- SID register IDs, op codes, pandas dtypes, PAL clock constants.
  • preframr_tokens.engine_fingerprint -- engine clustering for cross-engine evaluation pinning.
  • preframr_tokens.coarsen_pass -- tracker-export pass (lossy audio-domain bucketing).
  • preframr_tokens.dump_meta -- per-dump metadata sidecar with code-hash staleness gate.
  • preframr_tokens.reglog_helpers -- palette IO + dtype tightening.
  • preframr_tokens.alphabet_projection -- eval-set atom projection table.
  • preframr_tokens.reg_mappers -- FreqMapper (PAL clock + cents quantization).
  • preframr_tokens.constrained_decode -- per-step structural-validity mask for sampling-time logit guarding. Pure numpy state machine; consumers (torch users) apply the returned bool mask with a single masked_fill at the boundary.

Library-only

No CLI entry points. Consumers build their own (the main preframr repo's parse.py and stftokenize.py are simple wrappers that construct RegLogParser / RegTokenizer from an argparse.Namespace).

Stability

Library follows semver from v1.0. Pre-1.0 releases may break API as the preframr codebase evolves. Token-alphabet shape changes bump major version since they invalidate downstream checkpoints.

License

Apache 2.0. See LICENSE.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

preframr_tokens-0.3.0.tar.gz (133.0 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

preframr_tokens-0.3.0-py3-none-any.whl (105.0 kB view details)

Uploaded Python 3

File details

Details for the file preframr_tokens-0.3.0.tar.gz.

File metadata

  • Download URL: preframr_tokens-0.3.0.tar.gz
  • Upload date:
  • Size: 133.0 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for preframr_tokens-0.3.0.tar.gz
Algorithm Hash digest
SHA256 3ee6101e9ef1f6706127e36d32321b7f71b1d12566ff8db4f4556a70db1fc202
MD5 27166408e4a5f050efbf8cb86fc31506
BLAKE2b-256 88d9d2bf25ef6ac2ba44aadcddaad7550dc7c3f075e6c0a065502aaee87b0c11

See more details on using hashes here.

Provenance

The following attestation bundles were made for preframr_tokens-0.3.0.tar.gz:

Publisher: release.yml on anarkiwi/preframr-tokens

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file preframr_tokens-0.3.0-py3-none-any.whl.

File metadata

  • Download URL: preframr_tokens-0.3.0-py3-none-any.whl
  • Upload date:
  • Size: 105.0 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for preframr_tokens-0.3.0-py3-none-any.whl
Algorithm Hash digest
SHA256 d5417bc5402ff3b3ad7f3a6348809a08133f0c02120c42c6dacbe32583b47d70
MD5 2e0896f5415ec8f0692a0e67b485a492
BLAKE2b-256 e725be86297153ecbc0938c7d9117ec82f3fa617202a246f3b78189c9e7fe7df

See more details on using hashes here.

Provenance

The following attestation bundles were made for preframr_tokens-0.3.0-py3-none-any.whl:

Publisher: release.yml on anarkiwi/preframr-tokens

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page