Commit History
-
Reword docs
Verified signatured78405e -
Match pipeline batching
Verified signature23a1080 -
Match pipeline windowing
Verified signatureMakes `infer_genbank.py` reproduce the inference step of `GENERanno-cds-annotation-pipeline` (`src/model_inference.py`), which produced the database. - Windows: 16,384 tokens with a 1,024-token overlap (`--overlap-length`), instead of 50%. The last window is right-aligned to the record end, so no windows are N-padded and no tail bases are lost. Records shorter than one window use the pipeline's start-aligned and end-aligned passes, and records under 6 bp get P(CDS) = 0. Compute drops from about 2× to about 1.07× per base. - Same merge as the pipeline: softmax in the model dtype, with P(CDS) summed and divided by coverage. - Ambiguous 6-mers are predicted, as in the pipeline, instead of being set to NaN. An `ambiguous` mask replaces `valid`. - `--attn-implementation flash_attention_2` option to match the pipeline's attention kernel. - README: the Python example uses `predict_record` instead of N-padding. README and INFERENCE.md describe the new windowing and outputs. Check: the pipeli…3e16e67
-
Support Transformers 5 (#1)
- Support Transformers 5 (6ea829a1c2c2739914001d0b82afced465530d71) - Drop version check notes (06824283cb6005e6b0dbef233a29d1c52e246aea) Co-authored-by: Kashif Rasul <kashif@users.noreply.huggingface.co>7257133
-
Align Carbon-A model card with Carbon family structure and add detailed inference guide
Verified signature028a01a -
Simplify model card wording and remove preview repository references
Verified signature5b614dc -
Document passing GPU GenBank inference and overlap checks
Verified signatured3fb599 -
Add GenBank and FASTA file inference example with overlapping windows
Verified signatured429145 -
Document successful full-length GPU inference check
Verified signature12b0588 -
Package original Carbon-A checkpoint with packed-data inference example
Verified signatured5a62f1 -
initial commit
Verified signature370d018