Skip to content

issues Search Results · repo:explosion/curated-transformers language:Python

Filter by

31 results
 (96 ms)

31 results

inexplosion/curated-transformers (press backspace or delete to remove)

explosion/curated-transformers
Are there any Docker Images that are compatible with spacy==3.7 and CUDA <11.8?

  • EY4L
  • 1
  • Opened 
    on Apr 5, 2024
  • #368

explosion/curated-transformers
Truncation of sequences that are beyond the model's maximum length

Hi, First, I would like to thank you for this library :-) I m really enjoying it. I tried to tokenize a sequence with around 4K tokens and then fed it to a RoBERTa-based model (CodeBERT). This led to ...
feat/tokenization
type/bug
type/feature
  • MootezSaaD
  • 2
  • Opened 
    on Jan 14, 2024
  • #359

explosion/curated-transformers
Add suggested PyTorch LLM optimizations

Most of these optimizations should be easy to add: https://pytorch.org/blog/accelerating-generative-ai-2/
feat/generation
feat/model
  • danieldk
  • Opened 
    on Dec 1, 2023
  • #356

explosion/curated-transformers
Move the old Falcon architecuture to the extras/addons pacakage

We currently support two different architectures for the Falcon family of models. This adds a fair bit of complexity to the codebase since one of the architectures doesn t lend itself to the reuse of the ...
type/maintenance
  • shadeMe
  • Opened 
    on Oct 19, 2023
  • #355

explosion/curated-transformers
Add support for attention sinks

See https://arxiv.org/abs/2309.17453
feat/layers
feat/model
type/feature
  • danieldk
  • Opened 
    on Oct 4, 2023
  • #350

explosion/curated-transformers
Support DeBERTa v2/3

Work in progress branch: https://github.com/shadeMe/curated-transformers/tree/staging/feature/deberta-encoder Related to #347.
feat/model
type/feature
  • danieldk
  • 1
  • Opened 
    on Oct 3, 2023
  • #348

explosion/curated-transformers
Add a an extras/contrib package

Some models deviate so much from standard transformer encoder/decoders (e.g. DeBERTa and Falcon old architecture) that we probably should not support them in mainline Curated Transformers to avoid cluttering ...
type/maintenance
  • danieldk
  • Opened 
    on Oct 3, 2023
  • #347

explosion/curated-transformers
Register models/tokenizers through `catalogue`, so that the auto classes pick them up

The auto classes can currently support tokenizers and models from Curated Transformers. If they were registered through catalogue, we could support external models in the auto classes.
feat/model
feat/tokenization
type/feature
  • danieldk
  • 2
  • Opened 
    on Oct 3, 2023
  • #346

explosion/curated-transformers
Expose more outputs through the `Generator` interface

Expose more useful outputs, such as logits through the Generator interface. Also fixes #311 .
feat/generation
type/feature
  • danieldk
  • Opened 
    on Oct 3, 2023
  • #345

explosion/curated-transformers
Make `QkvMode` ADT-like

Some behaviors, like how to split heads, do not apply to all modes, so they should be scoped.
feat/layers
type/maintenance
  • danieldk
  • Opened 
    on Oct 3, 2023
  • #344
Issue origami icon

Learn how you can use GitHub Issues to plan and track your work.

Save views for sprints, backlogs, teams, or releases. Rank, sort, and filter issues to suit the occasion. The possibilities are endless.Learn more about GitHub Issues
ProTip! 
Restrict your search to the title by using the in:title qualifier.
Issue origami icon

Learn how you can use GitHub Issues to plan and track your work.

Save views for sprints, backlogs, teams, or releases. Rank, sort, and filter issues to suit the occasion. The possibilities are endless.Learn more about GitHub Issues
ProTip! 
Restrict your search to the title by using the in:title qualifier.
Issue search results · GitHub