git config

.gitattributes

@@ -33,3 +33,6 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+*.png filter=lfs diff=lfs merge=lfs -text
+results.json filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -0,0 +1,171 @@
+# ---> Python
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#pdm.lock
+#   pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
+#   in version control.
+#   https://pdm.fming.dev/#use-with-ide
+.pdm.toml
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  and can be added to the global gitignore or merged into this file.  For a more nuclear
+#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
+.idea/
+
+.DS_Store
+.ruff_cache
+venv*/
+wandb*/
+data/
+pretrain-data/
+contrain-data/
+core-data-*/
+out/pretrain-core/step-*/

README.md

@@ -1,3 +1,169 @@
 ---
 license: mit
+pipeline_tag: text-generation
+library_name: transformers
+language: [
+    'en', 'am', 'ar', 'as', 'az', 'be', 'bg', 'bn', 'br', 'bs', 'ca', 'cs', 'cy', 'da', 'de', 'el',
+    'eo', 'es', 'et', 'eu', 'fa', 'ff', 'fi', 'fr', 'fy', 'ga', 'gd', 'gl', 'gn', 'gu', 'ha', 'he',
+    'hi', 'hr', 'ht', 'hu', 'hy', 'id', 'ig', 'is', 'it', 'ja', 'jv', 'ka', 'kk', 'km', 'kn', 'ko',
+    'ku', 'ky', 'la', 'lg', 'li', 'ln', 'lo', 'lt', 'lv', 'mg', 'mk', 'ml', 'mn', 'mr', 'ms', 'my',
+    'ne', 'nl', 'no', 'ns', 'om', 'or', 'pa', 'pl', 'ps', 'pt', 'qu', 'rm', 'ro', 'ru', 'sa', 'si',
+    'sc', 'sd', 'sk', 'sl', 'so', 'sq', 'sr', 'ss', 'su', 'sv', 'sw', 'ta', 'te', 'th', 'tl', 'tn',
+    'tr', 'ug', 'uk', 'ur', 'uz', 'vi', 'wo', 'xh', 'yi', 'yo', 'zu',
+]
+datasets:
+# core - base
+- ontocord/fineweb-permissive-multilingual-2m
+- distily/c4_multilingual_1M
+- data-silence/sumnews
+- xu-song/cc100-samples
+- badrex/llm-emoji-dataset
+- fblgit/simple-math
+- Gusarich/math-expressions-1m
+- neuralwork/arxiver
+- christopher/rosetta-code
+- nampdn-ai/tiny-codes
+- JeanKaddour/minipile
+# core - instruct
+- NousResearch/hermes-function-calling-v1
+- simplescaling/s1K-1.1
+# base - instruct
+- mlabonne/open-perfectblend
+- allenai/tulu-3-sft-mixture
+- rombodawg/Everything_Instruct_Multilingual
+# base - reason
+- open-r1/OpenR1-Math-220k
+- open-thoughts/OpenThoughts-114k
+- cognitivecomputations/dolphin-r1
+- simplescaling/s1K-1.1
+tags:
+- chat
+- core
+- base
+- instruct
+- reason
 ---
+
+# tangled-alpha-0.6-core
+
+![logo](./misc/logo.jpg)
+
+```bash
+time python -B prepare_core_datasets.py
+```
+
+```
+i=0, min_len=0, max_len=1073741824, block_size=4097, chunk_size=16388000, len(dataset)=1287403, len(dataset) * block_size=5274490091
+Total number of tokens in the optimized dataset '../core-data-0-0-1073741824-4097-4000' is 5274490091
+
+```
+
+```bash
+CUDA_VISIBLE_DEVICES=0 CUDA_LAUNCH_BLOCKING=0 PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True litgpt pretrain --config pretrain_core_model.yaml
+```
+
+```
+Seed set to 23
+Time to instantiate model: 0.31 seconds.
+Total parameters: 201,359,872
+Verifying settings ...
+Measured TFLOPs: 7072.06
+Epoch 1 | iter 256 step 1 | loss train: 11.961, val: n/a | iter time: 406.23 ms (step) remaining time: 3 days, 13:55:33
+Epoch 1 | iter 512 step 2 | loss train: 11.953, val: n/a | iter time: 358.84 ms (step) remaining time: 3 days, 0:49:32
+Epoch 1 | iter 768 step 3 | loss train: 11.943, val: n/a | iter time: 357.16 ms (step) remaining time: 2 days, 20:38:36
+Epoch 1 | iter 1024 step 4 | loss train: 11.907, val: n/a | iter time: 355.69 ms (step) remaining time: 2 days, 18:31:54
+Epoch 1 | iter 1280 step 5 | loss train: 11.854, val: n/a | iter time: 358.32 ms (step) remaining time: 2 days, 17:13:13
+Epoch 1 | iter 1536 step 6 | loss train: 11.789, val: n/a | iter time: 355.59 ms (step) remaining time: 2 days, 16:18:25
+Epoch 1 | iter 1792 step 7 | loss train: 11.703, val: n/a | iter time: 354.88 ms (step) remaining time: 2 days, 15:37:56
+Epoch 1 | iter 2048 step 8 | loss train: 11.586, val: n/a | iter time: 354.07 ms (step) remaining time: 2 days, 15:06:45
+Epoch 1 | iter 2304 step 9 | loss train: 11.451, val: n/a | iter time: 352.89 ms (step) remaining time: 2 days, 14:41:54
+Epoch 1 | iter 2560 step 10 | loss train: 11.347, val: n/a | iter time: 355.58 ms (step) remaining time: 2 days, 14:21:38
+Epoch 1 | iter 2816 step 11 | loss train: 11.271, val: n/a | iter time: 351.01 ms (step) remaining time: 2 days, 14:04:43
+Epoch 1 | iter 3072 step 12 | loss train: 11.194, val: n/a | iter time: 351.91 ms (step) remaining time: 2 days, 13:50:26
+Epoch 1 | iter 3328 step 13 | loss train: 11.151, val: n/a | iter time: 353.02 ms (step) remaining time: 2 days, 13:38:04
+Epoch 1 | iter 3584 step 14 | loss train: 11.097, val: n/a | iter time: 353.75 ms (step) remaining time: 2 days, 13:27:21
+Epoch 1 | iter 3840 step 15 | loss train: 11.064, val: n/a | iter time: 358.31 ms (step) remaining time: 2 days, 13:17:48
+Epoch 1 | iter 4096 step 16 | loss train: 11.008, val: n/a | iter time: 351.95 ms (step) remaining time: 2 days, 13:09:17
+Epoch 1 | iter 4352 step 17 | loss train: 10.997, val: n/a | iter time: 352.26 ms (step) remaining time: 2 days, 13:01:35
+Epoch 1 | iter 4608 step 18 | loss train: 10.951, val: n/a | iter time: 352.57 ms (step) remaining time: 2 days, 12:54:35
+Epoch 1 | iter 4864 step 19 | loss train: 10.902, val: n/a | iter time: 354.73 ms (step) remaining time: 2 days, 12:48:13
+Epoch 1 | iter 5120 step 20 | loss train: 10.877, val: n/a | iter time: 354.47 ms (step) remaining time: 2 days, 12:43:19
+Epoch 1 | iter 5376 step 21 | loss train: 10.830, val: n/a | iter time: 353.78 ms (step) remaining time: 2 days, 12:37:49
+Epoch 1 | iter 5632 step 22 | loss train: 10.809, val: n/a | iter time: 355.03 ms (step) remaining time: 2 days, 12:32:44
+Epoch 1 | iter 5888 step 23 | loss train: 10.727, val: n/a | iter time: 351.49 ms (step) remaining time: 2 days, 12:27:56
+Epoch 1 | iter 6144 step 24 | loss train: 10.707, val: n/a | iter time: 351.58 ms (step) remaining time: 2 days, 12:23:24
+Epoch 1 | iter 6400 step 25 | loss train: 10.643, val: n/a | iter time: 350.84 ms (step) remaining time: 2 days, 12:19:10
+Epoch 1 | iter 6656 step 26 | loss train: 10.649, val: n/a | iter time: 355.14 ms (step) remaining time: 2 days, 12:15:07
+Epoch 1 | iter 6912 step 27 | loss train: 10.580, val: n/a | iter time: 352.60 ms (step) remaining time: 2 days, 12:11:12
+Epoch 1 | iter 7168 step 28 | loss train: 10.554, val: n/a | iter time: 351.57 ms (step) remaining time: 2 days, 12:07:27
+Epoch 1 | iter 7424 step 29 | loss train: 10.526, val: n/a | iter time: 350.36 ms (step) remaining time: 2 days, 12:03:55
+Epoch 1 | iter 7680 step 30 | loss train: 10.496, val: n/a | iter time: 353.19 ms (step) remaining time: 2 days, 12:00:34
+Epoch 1 | iter 7936 step 31 | loss train: 10.496, val: n/a | iter time: 350.95 ms (step) remaining time: 2 days, 11:57:21
+Epoch 1 | iter 8192 step 32 | loss train: 10.421, val: n/a | iter time: 352.71 ms (step) remaining time: 2 days, 11:54:18
+Epoch 1 | iter 8448 step 33 | loss train: 10.379, val: n/a | iter time: 354.15 ms (step) remaining time: 2 days, 11:51:21
+Epoch 1 | iter 8704 step 34 | loss train: 10.343, val: n/a | iter time: 353.95 ms (step) remaining time: 2 days, 11:48:29
+Epoch 1 | iter 8960 step 35 | loss train: 10.353, val: n/a | iter time: 351.04 ms (step) remaining time: 2 days, 11:45:44
+Epoch 1 | iter 9216 step 36 | loss train: 10.323, val: n/a | iter time: 354.76 ms (step) remaining time: 2 days, 11:43:05
+Epoch 1 | iter 9472 step 37 | loss train: 10.258, val: n/a | iter time: 353.18 ms (step) remaining time: 2 days, 11:40:29
+Epoch 1 | iter 9728 step 38 | loss train: 10.260, val: n/a | iter time: 353.86 ms (step) remaining time: 2 days, 11:37:57
+Epoch 1 | iter 9984 step 39 | loss train: 10.257, val: n/a | iter time: 356.14 ms (step) remaining time: 2 days, 11:35:50
+Epoch 1 | iter 10240 step 40 | loss train: 10.179, val: n/a | iter time: 353.73 ms (step) remaining time: 2 days, 11:33:23
+Epoch 1 | iter 10496 step 41 | loss train: 10.163, val: n/a | iter time: 350.49 ms (step) remaining time: 2 days, 11:30:59
+Epoch 1 | iter 10752 step 42 | loss train: 10.156, val: n/a | iter time: 354.15 ms (step) remaining time: 2 days, 11:28:40
+Epoch 1 | iter 11008 step 43 | loss train: 10.150, val: n/a | iter time: 350.99 ms (step) remaining time: 2 days, 11:26:24
+Epoch 1 | iter 11264 step 44 | loss train: 10.089, val: n/a | iter time: 354.28 ms (step) remaining time: 2 days, 11:24:09
+Epoch 1 | iter 11520 step 45 | loss train: 10.096, val: n/a | iter time: 352.46 ms (step) remaining time: 2 days, 11:21:56
+Epoch 1 | iter 11776 step 46 | loss train: 10.021, val: n/a | iter time: 356.80 ms (step) remaining time: 2 days, 11:19:45
+Epoch 1 | iter 12032 step 47 | loss train: 10.002, val: n/a | iter time: 355.30 ms (step) remaining time: 2 days, 11:17:36
+Epoch 1 | iter 12288 step 48 | loss train: 10.021, val: n/a | iter time: 355.12 ms (step) remaining time: 2 days, 11:15:32
+Epoch 1 | iter 12544 step 49 | loss train: 10.017, val: n/a | iter time: 353.81 ms (step) remaining time: 2 days, 11:13:29
+Epoch 1 | iter 12800 step 50 | loss train: 9.966, val: n/a | iter time: 354.70 ms (step) remaining time: 2 days, 11:11:26
+# ...
+Epoch 1 | iter 640256 step 2501 | loss train: 2.875, val: 2.786 | iter time: 348.10 ms (step) remaining time: 0:19:39
+Epoch 1 | iter 640512 step 2502 | loss train: 2.885, val: 2.786 | iter time: 349.50 ms (step) remaining time: 0:18:15
+Epoch 1 | iter 640768 step 2503 | loss train: 2.857, val: 2.786 | iter time: 347.05 ms (step) remaining time: 0:16:52
+Epoch 1 | iter 641024 step 2504 | loss train: 2.925, val: 2.786 | iter time: 347.38 ms (step) remaining time: 0:15:28
+Epoch 1 | iter 641280 step 2505 | loss train: 2.882, val: 2.786 | iter time: 346.76 ms (step) remaining time: 0:14:04
+Epoch 1 | iter 641536 step 2506 | loss train: 2.875, val: 2.786 | iter time: 348.08 ms (step) remaining time: 0:12:40
+Epoch 1 | iter 641792 step 2507 | loss train: 2.979, val: 2.786 | iter time: 349.34 ms (step) remaining time: 0:11:16
+Epoch 1 | iter 642048 step 2508 | loss train: 2.971, val: 2.786 | iter time: 348.34 ms (step) remaining time: 0:09:52
+Epoch 1 | iter 642304 step 2509 | loss train: 2.991, val: 2.786 | iter time: 347.89 ms (step) remaining time: 0:08:28
+Epoch 1 | iter 642560 step 2510 | loss train: 2.999, val: 2.786 | iter time: 349.61 ms (step) remaining time: 0:07:05
+Epoch 1 | iter 642816 step 2511 | loss train: 3.013, val: 2.786 | iter time: 349.54 ms (step) remaining time: 0:05:41
+Epoch 1 | iter 643072 step 2512 | loss train: 2.923, val: 2.786 | iter time: 348.39 ms (step) remaining time: 0:04:17
+Epoch 1 | iter 643328 step 2513 | loss train: 2.986, val: 2.786 | iter time: 347.26 ms (step) remaining time: 0:02:53
+Epoch 1 | iter 643584 step 2514 | loss train: 2.939, val: 2.786 | iter time: 348.31 ms (step) remaining time: 0:01:29
+Epoch 2 | iter 643840 step 2515 | loss train: 2.835, val: 2.786 | iter time: 349.08 ms (step) remaining time: 0:00:05
+Validating ...
+Final evaluation | val loss: 2.786 | val ppl: 16.208
+Saving checkpoint to '../out/pretrain-core/final/lit_model.pth'
+----------------------------------------
+| Performance
+| - Total tokens  : 5,274,484,736
+| - Training Time : 210925.61 s
+| - Tok/sec       : 8533.19 tok/s
+| ----------------------------------------
+| Memory Usage
+| - Memory Used   : 20.44 GB
+----------------------------------------
+```
+
+Backup `wandb`:
+
+```bash
+mv wandb wandb-pretrain-core
+```
+
+Chat with model:
+
+```bash
+CUDA_VISIBLE_DEVICES=0 CUDA_LAUNCH_BLOCKING=0 PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True litgpt chat ../out/pretrain-core/final
+```
+
+```bash
+CUDA_VISIBLE_DEVICES=0 CUDA_LAUNCH_BLOCKING=0 PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True time litgpt evaluate --tasks 'leaderboard' --out_dir '../evaluate/pretrain-core-0/leaderboard/' --batch_size 1 --dtype 'bfloat16' '../out/pretrain-core/final'
+```
+
+```
+# ...
+```