Skip to content

Releases: togethercomputer/together-python

v0.2.6

27 Oct 17:04
52b9bbc
Compare
Choose a tag to compare

What's Changed

New Contributors

Full Changelog: v0.1.5...v0.2.6

v0.1.5

19 Aug 01:01
Compare
Choose a tag to compare

v0.1.5 - 2023-08-18

Full Changelog: v0.0.9...v0.1.5

Changes

  • new finetuning support for togethercomputer/Llama-2-7B-32K-Instruct

v0.1.4

18 Aug 16:12
430563e
Compare
Choose a tag to compare

v0.1.4 - 2023-08-18

Full Changelog: v0.0.9...v0.1.4

Changes

  • Step by step guidance and examples added to documentation for the completion, files, and finetune sections, including example jsonl dataset that can be downloaded locally for inspection, automated checking and uploading
  • optional but very useful together.Files.check function to check the correctness of jsonl files. Gives useful feeback on which first line that is formatted incorrectly, will give the specific reason for being the line being incorrect and how to fix it with a helpful link
  • model specific token option to check if model specific tokens are used in dataset
  • together.Finetune.create will give useful feedback if user tries to upload anything other than a valid file-id for a file they already have uploaded successfully

Here is an example file check output:

{   'file_present': 'File found',
    'file_size': 'File size 0.2 GB',
    'is_check_passed': False,
    'key_value': 'Unexpected, value type for "text" key on line 6 of the input '
                 'file.The value type of the "text" key must be a '
                 'string.Expected format: {"text":"my sample string"}See '
                 'https://docs.together.ai/docs/fine-tuning for more '
                 'information.{"text": {"text":"<human>: Salutations!\\n<bot>: '
                 'Salutations to you as well! How can I help you today?"}}\n',
    'min_samples': 'Processing /Users/carsonlam/Projects/data/bad.jsonl '
                   'resulted in only 10 samples. Our minimum is 100 samples. ',
    'model_special_tokens': 'we are not yet checking end of sentence tokens '
                            'for this model',
    'num_samples_w_eos_token': 0,
    'text_field': 'No "text" field was found on line 7 of the the input '
                  'file.Expected format: {"text":"my sample string"}.see '
                  'https://docs.together.ai/docs/fine-tuning for more '
                  'information.{"ext": "<human>: Hiya!\\n<bot>: Hiya! How may '
                  'I assist you?"}\n'}

v0.1.3

11 Aug 21:10
84f5146
Compare
Choose a tag to compare

v0.1.3 - 2023-08-11

Full Changelog: v0.0.9...v0.1.3

Changes

  • Users can now add their weights and biases WANDB_API_KEY
  • For fine-tuning, users indicate the number of checkpoints using the --n-checkpoints parameter

here is an example:

together finetune create -t file-960be810-4d33-449a-885a-9f69bd8fd0e2 -m togethercomputer/LLaMA-2-7B-32K --n-checkpoints 2 --wandb-api-key 1123456789abcdefghijklmnop

v0.1.2

08 Aug 13:43
Compare
Choose a tag to compare

v0.1.2 - 2023-08-08

Full Changelog: v0.0.9...v0.1.2

Changes

  • Print list of started and stopped model instances with an API call from library or CLI
  • starting and stopping model instances with an API call from library or CLI
  • Update readme

v0.1.1

28 Jul 21:05
Compare
Choose a tag to compare

v0.1.1 - 2023-07-28

Full Changelog: v0.0.9...v0.1.1

Changes

  • Checkpoint download fixes for zst
  • Colored logging
  • Fixed class return and init behavior
  • Add models class and listing information options
  • Update prompt_id and response_id for inference
  • Updated files progress bar descriptions
  • Updated descriptions for image generation downloading
  • Add user-agent
  • Update readme

v0.0.9

19 Jul 15:46
Compare
Choose a tag to compare

v0.0.9 - 2023-07-19

Full Changelog: v0.0.8...v0.0.9

Bug

  • Fixed a minor dependency bug that would cause failures with together chat

v0.0.8

18 Jul 22:31
Compare
Choose a tag to compare

v0.0.8 - 2023-07-18

Full Changelog: v0.0.7...v0.0.8

Fixed

  • Added chat argument for streaming responses
  • Streaming inference by default
  • Updated models info

v0.0.7

15 Jul 19:50
Compare
Choose a tag to compare

0.0.7 - 2023-07-15

Full Changelog: v0.0.6...v0.0.7

Fixed

  • Drop minimum version to 3.6
  • Minor bug fix

v0.0.6

14 Jul 18:00
ce5134a
Compare
Choose a tag to compare

0.0.6 - 2023-07-14

Full Changelog: v0.0.5...v0.0.6

Fixed

  • Stability for end users