All pull requests

Pull requests

• Change the STL of embeddings so indirect access appears in outermost

  #542·marnold-ibm opened Jul 15, 2026·

  ·

• updated llama model so it works on new stack using fms

  #525·AnishPahilajani opened Apr 21, 2026·

  ·

• Add Ministral 3B and 8B text-only model support

  #517·raghukiran1224 opened Mar 11, 2026Contributor·6 tasks done·

  ·
  3

• [Feature] Add generate_until() support for lm-eval

  #514·yannicks1 opened Mar 9, 2026Collaborator·

  ·
  6

• feat(idefics3): add FMS-native Idefics3 model and HF config mapping

  #503·toddllm opened Jan 16, 2026·

  ·
  39

• Add Granite Speech multimodal speech-to-text implementation

  #499·gsmoon97 opened Dec 21, 2025·

  ·

• [WIP] Granite Vision Embedding Support

  #488·alex-jw-brooks opened Nov 12, 2025Contributor·

  ·
  1

• Fix error message for input id types

  #483·alex-jw-brooks opened Nov 6, 2025Contributor·

  ·
  1

• feat: added revision kwargs parameters to get model from hugging face

  #468·wallashss opened Sep 11, 2025·

  ·
  2

• Roberta sentiment training - DO NOT REVIEW OR MERGE

  #464·ani300 opened Sep 9, 2025Collaborator·

  ·

• Extending Tensor Parallelism for IBM FMS: Sequence Parallelism

  #455·Sibi-Git opened Jul 17, 2025·

  ·

• fixed bug with only_last_token during generation

  #453·JRosenkranz opened Jul 15, 2025Collaborator·

  ·
  1

  

• Add prefil decode in expectations

  #452·nirajkamal opened Jul 15, 2025·

  ·
  33

• Add BERT QA support

  #451·andrea-fasoli opened Jul 15, 2025Collaborator·

  ·

• update to explict RMSNorm

  #449·mserranos opened Jul 14, 2025·

  ·
  3

• [DO NOT MERGE] Test PR for Webhook testing

  #448·Abhishek-TAMU opened Jul 12, 2025Collaborator·

  ·
  5

• switch to RMSNorm from pytorch

  #447·mserranos opened Jul 10, 2025·

  ·
  2

• Introduce index parameter in place of only_last_token

  #445·JRosenkranz opened Jul 9, 2025Collaborator·

  ·
  

• Some changes to the README

  #444·kaoutar55 opened Jul 8, 2025Contributor·

  ·
  2

  

• Adding support for Pixtral

  #435·sahilsuneja1 opened Jun 27, 2025Contributor·

  ·
  1

• Columbia IBM Project: Implementing Paged Attention with Flex Attention

  #421·thomasjoshi opened Jun 2, 2025·

  ·

• Ring Attention First Pass Implementation

  #407·pulitha13 opened May 9, 2025·

  ·

• Modernize Tensor Parallelism using the DTensor API for Llama and Granite

  #371·aw471 opened Dec 20, 2024·

  ·

• NestedTensor Support in LLAMA Forward Pass

  #370·kushaangowda opened Dec 20, 2024·

  ·

• Express fms model adapters as combination of keys

  #364·JRosenkranz opened Nov 19, 2024Collaborator·

  ·
  2