Skip to content

Releases: jmisilo/clip-gpt-captioning

0.1.3

29 Oct 19:33
4f74c7a
Compare
Choose a tag to compare

Fix model weights' URLs

What's Changed

  • fix: update model weights urls by @jmisilo in #84
  • fix: update model weights IDs in downloads.py file by @jmisilo in #86
  • Model weights url mismatch by @jmisilo in #88

Full Changelog: 0.1.2...v0.1.3

0.1.2

29 Mar 21:02
1b30de9
Compare
Choose a tag to compare

What's Changed

Full Changelog: v0.1.1...0.1.2

v0.1.1

07 Jan 21:26
e58370d
Compare
Choose a tag to compare

Package update

What's Changed

Full Changelog: v0.1.0...v0.1.1

v0.1.0

24 Nov 20:09
792c7a3
Compare
Choose a tag to compare

What's Changed

  • Trainer class
  • Small and Large versions of model
  • Retraining - fixes bug with lack of end of string token
  • Reorganizing

Full Changelog: v0.0.3.1...v0.1.0

v0.0.3.1

11 Nov 18:16
a9100e3
Compare
Choose a tag to compare

What's Changed

  • Improve imports quality by editting init.py by @jmisilo in #55

Full Changelog: v0.0.3...v0.0.3.1

v0.0.3

06 Nov 12:42
9a80cec
Compare
Choose a tag to compare

What's Changed

  • 43 - Change temperature possibility (placeholder, will be needed for beam search) by @jmisilo in #45
  • 46 - Temperature checking by @jmisilo in #47
  • 48 - Save loss history by @jmisilo in #50
  • 51 - Retraining by @jmisilo in #52

Full Changelog: v0.0.2.2...v0.0.3

v0.0.2.2

30 Oct 12:09
6980a52
Compare
Choose a tag to compare

What's Changed

  • Generate requirements.txt only with pip installed packages by @jmisilo in #41

Full Changelog: v0.0.2.1...v0.0.2.2

v0.0.2.1

30 Oct 11:26
3aaf65f
Compare
Choose a tag to compare

What's Changed

Full Changelog: v0.0.2...v0.0.2.1

v0.0.2

30 Oct 11:26
6b728f4
Compare
Choose a tag to compare

What's Changed

Full Changelog: v0.0.1...v0.0.2

v0.0.1

28 Oct 19:32
af5715f
Compare
Choose a tag to compare

💣 First version of working model.

  • Training process
  • Evaluation process (generating images + captions for whole dataset)
  • Predict functionality (on one image)
  • Description/simple docs
  • Fine results (like for this method of training)