Commits

Branch selector

 main

User selector

All users

All time

Commit History

Commits on May 21, 2026

• Fix serverless auth headers for newer OpenAI SDKs (#689)

  Kovbo

  authored
  bc7e14c

  Copy full SHA for bc7e14c

• Fix Qwen3 MoE identity LoRA export layout (#688)

  Show description for 80a66de
  FurtherAI

  authored
  80a66de

  Copy full SHA for 80a66de

Commits on May 20, 2026

• Add pipeline checkpoint retention strategies (#683)

  Show description for e9fdf04
  FurtherAI

  authored
  e9fdf04

  Copy full SHA for e9fdf04

Commits on May 19, 2026

• feat: Preserve thinking content in Tinker server chat templates (#685)

  bradhilton

  authored
  8d8ce60

  Copy full SHA for 8d8ce60

• feat: Preserve thinking content in chat templates (#684)

  bradhilton

  authored
  48488ea

  Copy full SHA for 48488ea

Commits on May 18, 2026

• [codex] Tighten ART training type surface (#682)

  Show description for dd62b98
  corbt

  authored
  dd62b98

  Copy full SHA for dd62b98

Commits on May 16, 2026

• Rework Megatron to Support Adding New Models (#674)

  Show description for 5048e1c
  FurtherAI

  authored
  5048e1c

  Copy full SHA for 5048e1c

Commits on May 7, 2026

• dev: Add checkpoint forking end-to-end tests (#657)

  Show description for 48b2e5f
  

  3 peopleauthored
  48b2e5f

  Copy full SHA for 48b2e5f

Commits on May 5, 2026

• Cancel monitor and server tasks on backend shutdown (#669)

  Show description for 099e6b2
  

  Fiona-Waters

  and

  claude

  authored
  099e6b2

  Copy full SHA for 099e6b2

Commits on May 1, 2026

• feat: Upgrade Tinker dependencies and un-vendor Tinker Cookbook

  bradhilton

  authored
  5b10809

  Copy full SHA for 5b10809

Commits on Apr 30, 2026

• chore: Update Model pricing

  vivekkalyan

  committed
  23a11f8

  Copy full SHA for 23a11f8

Commits on Apr 29, 2026

• Drop overlength SFT tensors before backend training (#665)

  Show description for fd36cbc
  Kovbo

  authored
  fd36cbc

  Copy full SHA for fd36cbc

• Disable result packing for Unsloth backend (#666)

  Show description for 9d58459
  FurtherAI

  authored
  9d58459

  Copy full SHA for 9d58459

Commits on Apr 28, 2026

• chore: Update torchao for Torch 2.10 compatibility (#664)

  Show description for aa4a424
  

  bradhilton

  and

  Cursor Agent

  authored
  aa4a424

  Copy full SHA for aa4a424

Commits on Apr 27, 2026

• Fix Unsloth autocast dtype for bf16 models (#663)

  vivekkalyan

  authored
  5cfe180

  Copy full SHA for 5cfe180

Commits on Apr 24, 2026

• chore: Enhance setup script to check for required packages before installation

  Show description for 79cb587
  bradhilton

  committed
  79cb587

  Copy full SHA for 79cb587

• fix: CI dependency resolution

  bradhilton

  authored
  74a4ed2

  Copy full SHA for 74a4ed2

Commits on Apr 22, 2026

• Fix megatron grad norm explosion (#648)

  Show description for c514c8d
  

  Kovbo

  and

  FurtherAI

  authored
  c514c8d

  Copy full SHA for c514c8d

Commits on Apr 21, 2026

• remove secrets.PYPI_ART_TOKEN (#655)

  Kovbo

  authored
  082c421

  Copy full SHA for 082c421

Commits on Apr 18, 2026

• chore: Update dependencies and enhance Megatron backend compatibility

  Show description for 679b236
  bradhilton

  committed
  679b236

  Copy full SHA for 679b236

• feat: Add Qwen3.5 support to Megatron backend (#637)

  vivekkalyan

  authored
  b41b518

  Copy full SHA for b41b518

• feat: Add dedicated merged mode to Megatron backend (#636)

  vivekkalyan

  authored
  624e958

  Copy full SHA for 624e958

• feat: Add dedicated lora mode to Megatron backend (#635)

  vivekkalyan

  authored
  38373fb

  Copy full SHA for 38373fb

Commits on Apr 12, 2026

• chore: Refactor message handling and enhance tokenization logic

  Show description for 4cbfa15
  bradhilton

  committed
  4cbfa15

  Copy full SHA for 4cbfa15

Commits on Apr 10, 2026

• DeepEP, `torch.compile` and Fix Megatron Training Bug (#646)

  Show description for dec6b3a
  FurtherAI

  authored
  dec6b3a

  Copy full SHA for dec6b3a

Commits on Apr 9, 2026

• Fix flex attention compilation (#647)

  Kovbo

  authored
  4c03ad2

  Copy full SHA for 4c03ad2

Commits on Apr 8, 2026

• Add Megatron SFT internal checkpoints (#645)

  Show description for 953577a
  Kovbo

  authored
  953577a

  Copy full SHA for 953577a

Commits on Apr 7, 2026

• Share training code between backends (#626)

  Show description for 8ad8e50
  

  3 peopleauthored
  8ad8e50

  Copy full SHA for 8ad8e50

Commits on Apr 2, 2026

• fix: restore PR 607 KL loss removal (#639)

  Show description for 75a81e9
  FurtherAI

  authored
  75a81e9

  Copy full SHA for 75a81e9

Commits on Apr 1, 2026

• fix: Qwen3.5 tool-call chat-template tokenization (#634)

  vivekkalyan

  authored
  7ffa8b4

  Copy full SHA for 7ffa8b4

Commits on Mar 27, 2026

• Update RL training skill (#633)

  Show description for 1905677
  Kovbo

  authored
  1905677

  Copy full SHA for 1905677

Commits on Mar 26, 2026

• Megatron LoRA correctness: align distributed semantics with Megatron and validate TP/EP/ETP/DP against an oracle (#619)

  Show description for 7bb8137
  FurtherAI

  authored
  7bb8137

  Copy full SHA for 7bb8137

Commits on Mar 25, 2026

• refactor: Update ParsedMessageOrChoice type to use Message

  Show description for dc8c338
  bradhilton

  committed
  dc8c338

  Copy full SHA for dc8c338

• feat: Add merged vLLM rollout weights (#631)

  vivekkalyan

  authored
  fb26124

  Copy full SHA for fb26124

• feat: Enhance tokenization and add logprobs support in Tinker

  Show description for eb6a50d
  bradhilton

  committed
  eb6a50d

  Copy full SHA for eb6a50d