Commits

Branch selector

main

User selector

All users

All time

Commit history

Commits on Apr 20, 2025

• Empty commit to test new protection rules

  qgallouedec

  committed
  c82f626View commit details

  Copy full SHA for c82f626

  Browse repository at this point

• Empty commit to trigger CI

  qgallouedec

  committed
  33add19View commit details

  Copy full SHA for 33add19

  Browse repository at this point

Commits on Apr 17, 2025

• ☝️ [GRPO] Generate once per effective batch (#3283)

  Show description for 294f35b
  

  qgallouedec

  and

  lewtun

  authored
  294f35bView commit details

  Copy full SHA for 294f35b

  Browse repository at this point

Commits on Apr 16, 2025

• [GRPO] Add metrics for low and high clipped token probabilities (#3289)

  Show description for 9874b3a
  

  lewtun

  and

  qgallouedec

  authored
  9874b3aView commit details

  Copy full SHA for 9874b3a

  Browse repository at this point

Commits on Apr 15, 2025

• 🅾️ Fixes typo in SFTTrainer (#3282)

  taras-sereda

  authored
  1e61f6cView commit details

  Copy full SHA for 1e61f6c

  Browse repository at this point

• 🧗 Add Ascend NPU support for vLLM server (#3286)

  Show description for 27adc30
  

  ji-huazhong

  and

  qgallouedec

  authored
  27adc30View commit details

  Copy full SHA for 27adc30

  Browse repository at this point

• 🏷️ Fixed naming error in output_dir for Gemma 3 VLM script (#3297)

  sergiopaniego

  authored
  df737f9View commit details

  Copy full SHA for df737f9

  Browse repository at this point

• Expose EOS token in SFTConfig (#3299)

  Show description for c04e84c
  

  lewtun

  and

  qgallouedec

  authored
  c04e84cView commit details

  Copy full SHA for c04e84c

  Browse repository at this point

Commits on Apr 11, 2025

• ⏱️ Fix vLLM server to support V1 Engine (#3276)

  I-l-l-I

  authored
  d625c55View commit details

  Copy full SHA for d625c55

  Browse repository at this point

• 🦾 Test vLLM client-server (#3277)

  qgallouedec

  authored
  6cdd24aView commit details

  Copy full SHA for 6cdd24a

  Browse repository at this point

Commits on Apr 10, 2025

• 🕊️ Un-restrict diffusers (#3274)

  qgallouedec

  authored
  8b38570View commit details

  Copy full SHA for 8b38570

  Browse repository at this point

• Add Fine-tuning a Multimodal Model Using SFT (Single or Multi-Image Dataset) guide to docs (#3235)

  Show description for 95b1a9f
  

  sergiopaniego

  and

  qgallouedec

  authored
  95b1a9fView commit details

  Copy full SHA for 95b1a9f

  Browse repository at this point

• 🔗 Fix Dr. GRPO paper link (#3275)

  qgallouedec

  authored
  5c15114View commit details

  Copy full SHA for 5c15114

  Browse repository at this point

Commits on Apr 9, 2025

• 🩺 Dr. GRPO loss (#3256)

  Show description for 5e2e9cb
  

  qgallouedec

  and

  lewtun

  authored
  5e2e9cbView commit details

  Copy full SHA for 5e2e9cb

  Browse repository at this point

• ♾️ [CI] Remove `test_raise_error_not_causallm` (#3265)

  qgallouedec

  authored
  227df82View commit details

  Copy full SHA for 227df82

  Browse repository at this point

• 🚧 Temporarily restrict diffusers to <0.33.0 due to ftfy optional dep issue breaking doc builds (#3273)

  qgallouedec

  authored
  ae15814View commit details

  Copy full SHA for ae15814

  Browse repository at this point

• 👎 [GRPO] Adds option to disable dropout (#3234)

  Show description for 47b9515
  

  edbeeching

  and

  qgallouedec

  authored
  47b9515View commit details

  Copy full SHA for 47b9515

  Browse repository at this point

• 🕷 Fix online DPO crash when model is a DataParallel object (#3225)

  Show description for c4891dc
  

  wilrop

  and

  qgallouedec

  authored
  c4891dcView commit details

  Copy full SHA for c4891dc

  Browse repository at this point

• Revert "reward takes completion ids"

  Show description for 055cee2
  qgallouedec

  committed
  055cee2View commit details

  Copy full SHA for 055cee2

  Browse repository at this point

• reward takes completion ids

  qgallouedec

  committed
  73a2fb0View commit details

  Copy full SHA for 73a2fb0

  Browse repository at this point

• 🐯 `is_liger_kernel_available` with min version (#3266)

  qgallouedec

  authored
  982ba08View commit details

  Copy full SHA for 982ba08

  Browse repository at this point

Commits on Apr 8, 2025

• ⛏️ Add cli dict parsing for grpo_config (#3082)

  Tavish9

  authored
  e03e7acView commit details

  Copy full SHA for e03e7ac

  Browse repository at this point

• 📜 Fix license and copyrights (#3264)

  qgallouedec

  authored
  9df19e8View commit details

  Copy full SHA for 9df19e8

  Browse repository at this point

• ☕ Overlong-filtering for GRPO (#3248)

  Show description for 1d7b8c4
  

  shirinyamani

  and

  qgallouedec

  authored
  1d7b8c4View commit details

  Copy full SHA for 1d7b8c4

  Browse repository at this point

• 💠 Fix multi-gpu padding free (#3245)

  qgallouedec

  authored
  7e17061View commit details

  Copy full SHA for 7e17061

  Browse repository at this point

• 📦 [SFT] Deprecate batched `formatting_func` (#3147)

  Show description for 559724e
  

  yefd

  and

  qgallouedec

  authored
  559724eView commit details

  Copy full SHA for 559724e

  Browse repository at this point

• 🗑️ Deprecate `ConstantLengthDataset` (#3242)

  qgallouedec

  authored
  a5a4672View commit details

  Copy full SHA for a5a4672

  Browse repository at this point

• 🏃 Fix and make CI faster (#3160)

  qgallouedec

  authored
  b6bcafbView commit details

  Copy full SHA for b6bcafb

  Browse repository at this point

Commits on Apr 6, 2025

• 🔭 Add support for better KL estimator (k3) in PPOTrainer (#3240)

  Show description for 4bfb8eb
  

  AMindToThink

  and

  qgallouedec

  authored
  4bfb8ebView commit details

  Copy full SHA for 4bfb8eb

  Browse repository at this point

Commits on Apr 5, 2025

• ☑ Update PULL_REQUEST_TEMPLATE.md (#3241)

  qgallouedec

  authored
  4d66badView commit details

  Copy full SHA for 4d66bad

  Browse repository at this point

• ⏳ PPOTrainer: fix progress bar for num_mini_batches > 1 (#2531)

  Show description for e90117b
  

  dawidm

  and

  qgallouedec

  authored
  e90117bView commit details

  Copy full SHA for e90117b

  Browse repository at this point

• 🌊 Add error for iterable datasets in GRPOTrainer (#3216)

  qgallouedec

  authored
  31b54a6View commit details

  Copy full SHA for 31b54a6

  Browse repository at this point

• 🎀 Simplify logging text (#3219)

  Show description for 17e33cd
  

  qgallouedec

  and

  lewtun

  authored
  17e33cdView commit details

  Copy full SHA for 17e33cd

  Browse repository at this point

• 📢 Improve GRPO trainer error message for invalid num_generations (#3199)

  Show description for 5a0cebc
  

  AliBakly

  and

  qgallouedec

  authored
  5a0cebcView commit details

  Copy full SHA for 5a0cebc

  Browse repository at this point

• ⏯️ Fix logging when resuming from checkpoint GRPO (#3185)

  qgallouedec

  authored
  65308cfView commit details

  Copy full SHA for 65308cf

  Browse repository at this point