Code

merge hf main#1

Code

Merged

rishiraj merged 20 commits intorishiraj:patch-1rishiraj/transformers:patch-1from huggingface:mainhuggingface/transformers:mainCopy head branch name to clipboardSep 6, 2024

+6,345-590

Commits

Commits on Sep 5, 2024

• Llava Onevision: add model (#32673)

  Show description for 43df47d
  

  zucchini-nlp

  and

  amyeroberts

  authored
  43df47d

• Fix: Fix FalconMamba training issues due to incompatible kernels (#33195)

  Show description for 47b0964
  younesbelkada

  authored
  47b0964

• Add paper link (#33305)

  Muennighoff

  authored
  03164ba

• 🚨 Fix torch.jit.trace for interpolate_pos_encoding in all vision models (#33226)

  Show description for c6d2848
  xenova

  authored
  c6d2848

• Update SECURITY.md (#32680)

  Show description for 132e875
  Michellehbn

  authored
  132e875

• Add Qwen2Moe GGUF loading support (#33264)

  Show description for 5d11de4
  VladOS95-cyber

  authored
  5d11de4

• simple align qwen2vl kv_seq_len calculation with qwen2 (#33161)

  Show description for 21fac7a
  

  4 peopleauthored
  21fac7a

Commits on Sep 6, 2024

• Add a community notebook for fine-tuning with QLoRA, PEFT, and MLflow (#33319)

  Show description for 5792c45
  daniellok-db

  authored
  5792c45

• Fix: StaticCache & inputs_embeds (#32932)

  Show description for 1759bb9
  zucchini-nlp

  authored
  1759bb9

• Docs: add more cross-references to the KV cache docs (#33323)

  Show description for 2b789f2
  gante

  authored
  2b789f2

• [whisper] alternative fix for long-form timestamps (#32131)

  Show description for 51d15eb
  sanchit-gandhi

  authored
  51d15eb

• fix qwen2vl vision eager-attention (#33213)

  Show description for 1bd9d1c
  

  simonJJJ

  and

  ArthurZucker

  authored
  1bd9d1c

• Load dynamic module (remote code) only once if code isn't change (#33162)

  Show description for e1c2b69
  XuehaiPan

  authored
  e1c2b69

• support loading model without config.json file (#32356)

  Show description for 363301f
  itazap

  authored
  363301f

• Add validation for maximum sequence length in modeling_whisper.py (#33196)

  Show description for 3314fe1
  

  AmirMohammadFakhimi

  and

  ylacombe

  authored
  3314fe1

• add self.head_dim for VisionAttention in Qwen2-VL (#33211)

  Show description for 2b18354
  GeLee-Q

  authored
  2b18354

• support 3D attention mask in bert (#32105)

  Show description for 342e800
  gathierry

  authored
  342e800

• Support reading tiktoken tokenizer.model file (#31656)

  Show description for e48e5f1
  itazap

  authored
  e48e5f1

• red-ci on main, fix copies (#33356)

  Show description for 2d75700
  ArthurZucker

  authored
  2d75700

• RoPE: fix BC warning (#33331)

  gante

  authored
  6ff6069