Code

Sync/main#5424

Code

Merged

FrankLeeeee merged 39 commits intohpcaitech:feature/colossal-inferhpcaitech/ColossalAI:feature/colossal-inferfrom FrankLeeeee:sync/mainFrankLeeeee/ColossalAI:sync/mainCopy head branch name to clipboardMar 4, 2024

Commits

Commits on Jan 31, 2024

• [extension] fixed exception catch (hpcaitech#5342)

  FrankLeeeee

  authored
  abd8e77

  Copy full SHA for abd8e77

Commits on Feb 1, 2024

• [Chat] fix sft loss nan (hpcaitech#5345)

  Show description for c523984
  YeAnbang

  authored
  c523984

  Copy full SHA for c523984

• [checkpointio] fix gemini and hybrid parallel optim checkpoint (hpcaitech#5347)

  Show description for ffffc32
  ver217

  authored
  ffffc32

  Copy full SHA for ffffc32

Commits on Feb 2, 2024

• [fix] remove unnecessary dp_size assert (hpcaitech#5351)

  Show description for 1c790c0
  CWHer

  authored
  1c790c0

  Copy full SHA for 1c790c0

Commits on Feb 4, 2024

• [gemini] fix param op hook when output is tuple (hpcaitech#5355)

  Show description for 2dd01e3
  ver217

  authored
  2dd01e3

  Copy full SHA for 2dd01e3

Commits on Feb 5, 2024

• [llama] fix dataloader for hybrid parallel (hpcaitech#5358)

  Show description for 6c0fa7b
  ver217

  authored
  6c0fa7b

  Copy full SHA for 6c0fa7b

• [llama] update training script (hpcaitech#5360)

  Show description for 73f9f23
  ver217

  authored
  73f9f23

  Copy full SHA for 73f9f23

• [llama] add flash attn patch for npu (hpcaitech#5362)

  ver217

  authored
  a4cec17

  Copy full SHA for a4cec17

• [llama] fix neftune & pbar with start_step (hpcaitech#5364)

  Camille7777

  authored
  44ca61a

  Copy full SHA for 44ca61a

Commits on Feb 6, 2024

• [eval] update llama npu eval (hpcaitech#5366)

  Camille7777

  authored
  a5756a8

  Copy full SHA for a5756a8

• [llama] polish training script and fix optim ckpt (hpcaitech#5368)

  ver217

  authored
  eb4f2d9

  Copy full SHA for eb4f2d9

• [lr-scheduler] fix load state dict and add test (hpcaitech#5369)

  ver217

  authored
  c53ddda

  Copy full SHA for c53ddda

• [llama] fix memory issue (hpcaitech#5371)

  Show description for 084c912
  ver217

  authored
  084c912

  Copy full SHA for 084c912

Commits on Feb 7, 2024

• [moe] init mixtral impl

  

  oahzxl

  authored and

  ver217

  committed
  7d8e033

  Copy full SHA for 7d8e033

• [moe] update capacity computing (hpcaitech#5253)

  Show description for c904d2a
  ver217

  committed
  c904d2a

  Copy full SHA for c904d2a

• [moe] support mixtral (hpcaitech#5309)

  Show description for da39d21
  ver217

  committed
  da39d21

  Copy full SHA for da39d21

• [moe] fix mixtral checkpoint io (hpcaitech#5314)

  ver217

  committed
  b60be18

  Copy full SHA for b60be18

• [moe] fix mixtral forward default value (hpcaitech#5329)

  ver217

  committed
  956b561

  Copy full SHA for 956b561

• [moe] fix mixtral optim checkpoint (hpcaitech#5344)

  ver217

  committed
  65e5d6b

  Copy full SHA for 65e5d6b

Commits on Feb 8, 2024

• [moe] fix tests

  ver217

  committed
  06db94f

  Copy full SHA for 06db94f

• Merge pull request hpcaitech#5377 from hpcaitech/example/llama-npu

  Show description for 4c03347
  FrankLeeeee

  authored
  4c03347

  Copy full SHA for 4c03347

• Merge pull request hpcaitech#5372 from hpcaitech/exp/mixtral

  FrankLeeeee

  authored
  efef43b

  Copy full SHA for efef43b

• [release] update version (hpcaitech#5380)

  ver217

  authored
  adae123

  Copy full SHA for adae123

Commits on Feb 19, 2024

• [llama] fix training and inference scripts (hpcaitech#5384)

  Show description for 7303801
  ver217

  authored
  7303801

  Copy full SHA for 7303801

• [doc] Fix typo (hpcaitech#5361)

  yixiaoer

  authored
  69e3ad0

  Copy full SHA for 69e3ad0

• [doc] updated installation command (hpcaitech#5389)

  FrankLeeeee

  authored
  705a62a

  Copy full SHA for 705a62a

• [hotfix] fix variable type for top_p (hpcaitech#5313)

  Show description for b833153
  

  CZYCW

  and

  binmakeswell

  authored
  b833153

  Copy full SHA for b833153

Commits on Feb 20, 2024

• [hotfix] Fix wrong import in meta_registry (hpcaitech#5392)

  stephankoe

  authored
  5d380a1

  Copy full SHA for 5d380a1

Commits on Feb 26, 2024

• [extension] hotfix jit extension setup (hpcaitech#5402)

  ver217

  authored
  95c21e3

  Copy full SHA for 95c21e3

Commits on Feb 27, 2024

• [example] reuse flash attn patch (hpcaitech#5400)

  ver217

  authored
  d882d18

  Copy full SHA for d882d18

• [fsdp] impl save/load shard model/optimizer (hpcaitech#5357)

  ericxsun

  authored
  bf34c6f

  Copy full SHA for bf34c6f

• [setup] fixed nightly release (hpcaitech#5388)

  FrankLeeeee

  authored
  dcdd8a5

  Copy full SHA for dcdd8a5

• [shardformer]gather llama logits (hpcaitech#5398)

  Show description for 0a25e16
  flybird11111

  authored
  0a25e16

  Copy full SHA for 0a25e16

Commits on Feb 28, 2024

• update requirements (hpcaitech#5407)

  TongLi3701

  authored
  a28c971

  Copy full SHA for a28c971

Commits on Feb 29, 2024

• [workflow] added pypi channel (hpcaitech#5412)

  FrankLeeeee

  authored
  2461f37

  Copy full SHA for 2461f37

• [doc] fix blog link

  

  binmakeswell

  authored and

  liwenjuna

  committed
  5de940d

  Copy full SHA for 5de940d

• [doc] fix blog link

  

  binmakeswell

  authored and

  liwenjuna

  committed
  a1c6cdb

  Copy full SHA for a1c6cdb

Commits on Mar 1, 2024

• fix sft single turn inference example (hpcaitech#5416)

  Camille7777

  authored
  4b8312c

  Copy full SHA for 4b8312c

Commits on Mar 4, 2024

• Merge branch 'main' into sync/main

  FrankLeeeee

  committed
  0310b76

  Copy full SHA for 0310b76