LLMs-from-scratch

books/LLMs-from-scratch

Fork 0

mirror of https://github.com/rasbt/LLMs-from-scratch.git synced 2026-04-10 12:33:42 +00:00

Commit Graph

Select branches

Hide Pull Requests

cleaning

debug-exclude-newer

dockerfile

fix-qwen3-tokeniizer-eos

fix_masking_kv_cache

gemma-3

head-dim

main

qwen-tokenizer-fix

requirements-update

uv-improvements

weezymatt-bugfix/semantic-error

#1

#10

#100

#1000

#1003

#1004

#1004

#101

#102

#104

#105

#106

#107

#109

#110

#111

#113

#115

#116

#117

#118

#119

#12

#120

#121

#122

#124

#125

#127

#128

#13

#131

#132

#133

#134

#135

#136

#137

#138

#139

#14

#141

#142

#143

#144

#149

#151

#152

#153

#154

#156

#157

#158

#159

#160

#161

#162

#163

#164

#165

#166

#169

#17

#170

#171

#173

#174

#176

#177

#178

#179

#18

#180

#182

#183

#184

#187

#189

#19

#193

#195

#196

#197

#198

#199

#2

#20

#200

#201

#203

#204

#205

#206

#207

#208

#211

#212

#213

#216

#218

#219

#22

#222

#223

#224

#227

#228

#229

#233

#234

#235

#236

#237

#239

#24

#240

#241

#242

#243

#244

#245

#246

#247

#248

#250

#252

#253

#254

#258

#259

#26

#260

#261

#263

#264

#265

#269

#27

#270

#274

#278

#28

#281

#285

#288

#289

#29

#290

#291

#294

#295

#297

#298

#3

#300

#301

#303

#304

#305

#307

#31

#313

#314

#318

#319

#32

#320

#321

#323

#324

#325

#329

#33

#333

#335

#336

#337

#339

#344

#345

#346

#349

#352

#353

#356

#36

#360

#361

#366

#367

#368

#369

#37

#372

#373

#375

#376

#378

#38

#380

#381

#383

#384

#386

#388

#389

#39

#391

#393

#395

#396

#397

#398

#4

#400

#401

#403

#404

#405

#406

#407

#408

#412

#413

#414

#419

#420

#421

#423

#427

#432

#436

#437

#438

#439

#441

#451

#458

#461

#465

#465

#466

#468

#469

#470

#471

#473

#479

#481

#483

#484

#487

#492

#495

#496

#498

#5

#50

#500

#504

#506

#507

#509

#511

#512

#516

#518

#519

#52

#520

#521

#522

#523

#524

#525

#526

#529

#530

#531

#533

#534

#535

#536

#537

#539

#54

#540

#541

#542

#543

#544

#545

#546

#547

#549

#55

#550

#551

#553

#555

#557

#56

#560

#561

#562

#563

#565

#571

#572

#573

#574

#575

#576

#577

#578

#581

#582

#583

#584

#585

#588

#589

#590

#591

#592

#593

#596

#597

#598

#6

#600

#602

#604

#605

#606

#611

#612

#613

#614

#615

#616

#617

#620

#621

#627

#628

#629

#63

#630

#638

#64

#640

#642

#643

#646

#649

#65

#653

#658

#66

#660

#661

#663

#665

#667

#668

#671

#672

#673

#674

#676

#677

#678

#68

#680

#682

#685

#687

#688

#69

#695

#696

#697

#698

#699

#7

#70

#700

#702

#703

#704

#705

#706

#708

#709

#71

#710

#711

#713

#714

#715

#716

#719

#722

#723

#725

#726

#727

#728

#729

#73

#730

#732

#733

#735

#737

#738

#74

#741

#744

#747

#748

#749

#75

#750

#752

#758

#76

#760

#761

#764

#767

#768

#769

#77

#770

#771

#773

#774

#775

#776

#78

#780

#781

#786

#788

#79

#793

#798

#799

#800

#801

#802

#805

#807

#809

#81

#810

#811

#812

#814

#815

#817

#818

#819

#82

#820

#823

#826

#828

#829

#83

#830

#831

#832

#833

#834

#839

#84

#841

#842

#843

#845

#847

#849

#850

#851

#858

#859

#861

#862

#863

#865

#866

#867

#869

#870

#874

#875

#876

#878

#879

#880

#881

#883

#887

#888

#89

#890

#891

#892

#893

#897

#898

#9

#90

#900

#901

#903

#904

#906

#909

#910

#914

#915

#916

#918

#920

#925

#926

#927

#933

#935

#937

#938

#940

#942

#943

#945

#946

#949

#950

#951

#952

#953

#954

#957

#958

#959

#96

#960

#961

#962

#965

#969

#97

#974

#975

#976

#981

#987

#989

#99

#991

#993

#994

#998

#999

8447d70b18 Some gemma 3 improvements (#1000) main Sebastian Raschka 2026-04-05 22:05:05 -04:00
afc6a3da07 Remove 7 days requirement to improve windows compatibility (#999) Sebastian Raschka 2026-04-04 13:31:43 -04:00
fd11713ed9 exclude torchvission from nightly debug-exclude-newer rasbt 2026-04-04 12:17:50 -05:00
7489500d90 Make change in a code file rasbt 2026-04-04 12:05:25 -05:00
f1668af932 uv packages must be 7 days old (#994) Sebastian Raschka 2026-03-31 20:10:15 -04:00
d977841fad Swap urllib.request with requests (#993) Sebastian Raschka 2026-03-30 22:03:13 -04:00
6b9502056f fix: pin 1 unpinned action(s) (#987) dagecko 2026-03-26 12:49:44 -04:00
9320a5e252 fix: added KVcache in generate_text_basic_stream (#981) casinca 2026-03-21 01:47:52 +01:00
130cc1f63c harded the link checker rasbt 2026-03-07 17:05:41 -06:00
9ab6e894ac Minor typo fix (#974) Sebastian Raschka 2026-03-07 17:31:40 -05:00
052c2dea4f Bpe whitespace fixes (#975) Sebastian Raschka 2026-03-07 14:56:25 -05:00
3a7b98a36a Add more analysis to qwen3.5 image Sebastian Raschka 2026-03-04 08:47:06 -06:00
ae8eebf0d7 Use full HF url Sebastian Raschka 2026-03-03 16:38:05 -06:00
7892ec9435 Qwen3.5 from scratch (#969) Sebastian Raschka 2026-03-03 17:31:16 -05:00
4612d20fa8 User argpars utils to show default args on command line rasbt 2026-03-01 20:15:21 -06:00
c079904491 Jupyter scrolling glitch tips (#965) Sebastian Raschka 2026-02-27 18:33:39 -05:00
ec78de32dc image size rasbt 2026-02-19 16:42:19 -06:00
10bffd62b7 image size rasbt 2026-02-19 16:41:43 -06:00
c745ded43d formatting fix rasbt 2026-02-19 16:40:28 -06:00
62f0356e0d Add Tiny Aya from scratch (#962) Sebastian Raschka 2026-02-19 17:33:22 -05:00
1ed48c2450 remove redundant assignment (#961) Sebastian Raschka 2026-02-18 23:03:49 -05:00
2d600ccb5b Use correct input in layernorm example (#960) Sebastian Raschka 2026-02-18 22:35:57 -05:00
be5e2a3331 Readability and code quality improvements (#959) Sebastian Raschka 2026-02-17 19:44:56 -05:00
7b1f740f74 Fix flex attention in PyTorch 2.10 (#957) Sebastian Raschka 2026-02-09 15:12:40 -05:00
82010e2c77 Fix docstring parameter names in compute_dpo_loss function (#953) Dawid Woźniak 2026-01-29 23:51:17 +01:00
e155d1b02c Update unit tests for CI (#952) Sebastian Raschka 2026-01-27 17:44:55 -06:00
59d9262047 chore: Update outdated GitHub Actions versions (#951) Pádraic Slattery 2026-01-19 19:22:29 +01:00
47cfc61800 link GRPO notebook (#950) Sebastian Raschka 2026-01-18 11:42:03 -06:00
9c4be478f8 Optional weight tying for Qwen3 and Llama3.2 pretraining (#949) casinca 2026-01-14 16:07:04 +01:00
e0dbec3331 Fix encoding of multiple preceding spaces in BPE tokenizer. (#945) Maxwell De Jong 2026-01-10 11:27:23 -05:00
90e0f3cc15 Chapter 5 with alternative LLMs (Qwen3, Llama 3) (#943) Sebastian Raschka 2026-01-09 14:58:20 -06:00
9df9e69cd2 Correct batch_idx in appendix A logging (#942) Henry 2026-01-08 09:14:17 +08:00
491fd58463 Fix Olmo3 YaRN RoPE implementation bug (#940) Gerardo Moreno 2026-01-03 16:59:57 -08:00
b26fa01381 Correct 'pix' to 'pixi' in README.md (#935) Maheshkumar P 2026-01-03 02:47:56 +05:30
e10af0a1b9 Clean up native-uv.md documentation (#938) Sebastian Raschka 2026-01-02 15:17:43 -06:00
14c7afaa58 Fix GitHub CI timeout issue for link checker (#937) Sebastian Raschka 2026-01-02 14:34:31 -06:00
5f3268c2a6 yearly update rasbt 2026-01-01 18:19:44 -06:00
d85ba93799 Cover Python 3.12 (#933) Sebastian Raschka 2025-12-27 16:30:13 -06:00
1cea30f9b1 upload saved nb rasbt 2025-12-20 18:41:34 -06:00
2b9a67c00d Update memory efficient loading nb rasbt 2025-12-20 18:35:13 -06:00
695ecb61ce update submodule rasbt 2025-12-20 11:38:06 -06:00
1c9f49c812 Add some appendix E runtimes (#927) Sebastian Raschka 2025-12-19 14:13:43 -06:00
57430d2a13 Gated DeltaNet updates (#926) Sebastian Raschka 2025-12-18 20:28:53 -06:00
d7f178d28b Sliding window KV Cache bug fix (#925) talentJay-ux 2025-12-15 16:47:01 -08:00
a11965fbd9 Remove persistent flag from cache buffers (#916) Sebastian Raschka 2025-11-24 20:10:02 -06:00
c19533851f Add Olmo 3 README (#915) Sebastian Raschka 2025-11-23 10:53:48 -06:00
bc6f335526 Olmo 3 from scratch (#914) Sebastian Raschka 2025-11-22 22:42:18 -06:00
398b079efa RoPE decay plot (#910) Sebastian Raschka 2025-11-17 17:29:49 -06:00
28a8408d4d Update README wrt multi-query attention Sebastian Raschka 2025-11-17 16:39:32 -06:00
a4094470c7 Write-up on how to get the most out of this book (#909) Sebastian Raschka 2025-11-12 20:20:48 -06:00
7d92267170 fix(GatedDeltaNet): Init param A from log of a uniform distrib (#906) casinca 2025-11-09 21:22:52 +01:00
35354fac80 Use consistent title case rasbt 2025-11-06 15:22:24 -06:00
58f45ae5a7 Fix empty device issue (#904) Sebastian Raschka 2025-11-05 20:04:44 -06:00
bcc73f731d n_heads × d_head -> d_head × d_head in DeltaNet (#903) Sebastian Raschka 2025-11-05 18:28:37 -06:00
488bef7e3f Image resizing Sebastian Raschka 2025-11-02 21:05:38 -06:00
c6b8332a59 Gated DeltaNet write-up (#901) Sebastian Raschka 2025-11-02 21:03:42 -06:00
d6c3990c57 Training on MPS in PyTorch 2.9 (#900) Sebastian Raschka 2025-11-01 16:55:09 -05:00
27d52d6378 Fix MHAEinsum weight dimension bug when d_in != d_out (#857) (#893) Aviral Garg 2025-11-01 08:15:31 +05:30
b1db33b384 simplify uv command (#898) Sebastian Raschka 2025-10-31 19:44:57 -05:00
760f4c9ecc Add bonus dependencies to pyproject (#897) Sebastian Raschka 2025-10-28 20:36:21 -05:00
0adb5b8c65 Fix ffn link (#892) Sebastian Raschka 2025-10-21 21:19:44 -05:00
7ca7c47e4a Make quote style consistent (#891) Sebastian Raschka 2025-10-21 19:42:33 -05:00
9276edbc37 - docs(moe): correct arXiv link for DeepSeekMoE (#890) casinca 2025-10-21 02:29:06 +02:00
218221ab62 Mixture-of-Experts intro (#888) Sebastian Raschka 2025-10-19 22:17:59 -05:00
27b6dfab9e Make it easier to toggle between thinking and instruct variants (#887) Sebastian Raschka 2025-10-16 20:37:31 -05:00
7fe4874dda Update the compression rate comment in MLA (#883) Sebastian Raschka 2025-10-14 11:10:06 -05:00
b969b3ef7a Use figure numbers in ch05-7 (#881) Sebastian Raschka 2025-10-13 16:26:35 -05:00
bf039ff3dc Add alternative attention structure (#880) Sebastian Raschka 2025-10-13 14:31:13 -05:00
6eb6adfa33 sliding window attention (#879) Sebastian Raschka 2025-10-12 22:13:20 -05:00
21f0617ea3 Add other appendices for completeness (#878) Sebastian Raschka 2025-10-12 19:04:53 -05:00
44eda5340a rm plot rasbt 2025-10-12 08:55:03 -05:00
9b9586688d Multi-Head Latent Attention (#876) Sebastian Raschka 2025-10-11 20:08:30 -05:00
bf27ad1485 Use GB instead of GiB consistently (#875) Sebastian Raschka 2025-10-11 09:11:33 -05:00
c814814d72 Grouped-Query Attention memory (#874) Sebastian Raschka 2025-10-11 08:44:19 -05:00
b8e12e1dd1 Use inference_device rasbt 2025-10-09 10:59:17 -05:00
fecfdd16ff Add simpler BPE, and make previous BPE better (#870) Sebastian Raschka 2025-10-08 22:22:34 -05:00
1164cb3e8f Qwen3 and evaluation bonus materials (#869) Sebastian Raschka 2025-10-08 18:22:19 -05:00
7bd263144e Switch from urllib to requests to improve reliability (#867) Sebastian Raschka 2025-10-07 15:22:59 -05:00
9f7dbb2493 Update docker file dockerfile rasbt 2025-10-06 18:31:59 -05:00
8552565bda Add missing comma in imports in README (#865) Sebastian Raschka 2025-10-06 16:03:04 -05:00
7084123d10 Note about output dimensions (#862) Sebastian Raschka 2025-10-01 10:47:04 -05:00
4d9f9dcb6c Update ollama address (#861) Sebastian Raschka 2025-09-30 21:05:53 -05:00
00c240ff87 some typo fixes (#858) casinca 2025-09-30 18:18:02 +02:00
458f2d9b67 Test dependencies with Python 3.13 (#843) Sebastian Raschka 2025-09-27 08:38:07 -05:00
47867bc1cb Update generate script (#847) Sebastian Raschka 2025-09-27 08:03:54 -05:00
9bc827ea7e Numerically stable generate on mps (#849) Sebastian Raschka 2025-09-26 22:42:44 -05:00
f492c949d3 Requirements update (#851) Sebastian Raschka 2025-09-26 22:19:57 -05:00
b1f852c1ba Update requirements.txt requirements-update rasbt 2025-09-26 21:57:22 -05:00
3c10919c32 Numerically stable generate on mps rasbt 2025-09-26 21:37:25 -05:00
322000d833 Windows compile (#845) Sebastian Raschka 2025-09-26 12:01:19 -05:00
3b83705988 Update package dependencies (#842) Sebastian Raschka 2025-09-22 18:32:39 -05:00
e742d8af2c Improve MoE implementation (#841) Sebastian Raschka 2025-09-22 15:21:06 -05:00
20041fb94b Note about devcontainer root usage (#833) Sebastian Raschka 2025-09-21 11:12:44 -05:00
2aa8e8130d Note about RoPE usage (#839) Sebastian Raschka 2025-09-20 11:25:58 -05:00
42c130623b Qwen3Tokenizer fix for Qwen3 Base models and generation mismatch with HF (#828) casinca 2025-09-17 15:14:11 +02:00
bfc6389fab fix code comment (#834) Synix 2025-09-17 09:36:02 +08:00
862df48e38 use apply_chat_template qwen-tokenizer-fix rasbt 2025-09-16 08:12:01 -05:00
8237b3fda0 removed duplicate code fragment intest_chat_wrap_and_equivalence casinca 2025-09-16 11:32:05 +02:00
16f30a0395 added copy of test def test_tokenizer_equivalence() from reasoning-from-scratch in test_qwen3.py casinca 2025-09-16 11:12:29 +02:00
4ea2fb4a76 copied download_file in utils from https://github.com/rasbt/reasoning-from-scratch/blob/main/reasoning_from_scratch/utils.py casinca 2025-09-16 11:10:01 +02:00

1 2 3 4 5 ...