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 +1,136 @@
 ---
 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-0.11-core
+
+![logo](./misc/logo.jpg)
+
+```bash
+time python -B prepare_core_datasets.py
+```
+
+```
+i=0, min_len=0, max_len=1073741824, block_size=1025, chunk_size=16400000, len(dataset)=10913927, len(dataset) * block_size=11186775175
+Total number of tokens in the optimized dataset '../core-data-0-0-1073741824-1025-16000' is 11186775175
+
+i=1, min_len=1025, max_len=2049, block_size=2049, chunk_size=16392000, len(dataset)=893465, len(dataset) * block_size=1830709785
+Total number of tokens in the optimized dataset '../core-data-1-1025-2049-2049-8000' is 1830709785
+
+i=2, min_len=2049, max_len=4097, block_size=4097, chunk_size=16388000, len(dataset)=375104, len(dataset) * block_size=1536801088
+Total number of tokens in the optimized dataset '../core-data-2-2049-4097-4097-4000' is 1536801088
+
+i=3, min_len=4097, max_len=8193, block_size=8193, chunk_size=16386000, len(dataset)=177522, len(dataset) * block_size=1454437746
+Total number of tokens in the optimized dataset '../core-data-3-4097-8193-8193-2000' is 1454437746
+
+i=4, min_len=8193, max_len=16385, block_size=16385, chunk_size=16385000, len(dataset)=77725, len(dataset) * block_size=1273524125
+Total number of tokens in the optimized dataset '../core-data-4-8193-16385-16385-1000' is 1273524125
+
+i=5, min_len=16385, max_len=32769, block_size=32769, chunk_size=16384500, len(dataset)=22931, len(dataset) * block_size=751425939
+Total number of tokens in the optimized dataset '../core-data-5-16385-32769-32769-500' is 751425939
+
+i=6, min_len=32769, max_len=65537, block_size=65537, chunk_size=16384250, len(dataset)=4988, len(dataset) * block_size=326898556
+Total number of tokens in the optimized dataset '../core-data-6-32769-65537-65537-250' is 326898556
+
+i=7, min_len=65537, max_len=131073, block_size=131073, chunk_size=16384125, len(dataset)=1137, len(dataset) * block_size=149030001
+Total number of tokens in the optimized dataset '../core-data-7-65537-131073-131073-125' is 149030001
+
+42G     ../core-data-0-0-1073741824-1025-16000
+6.9G    ../core-data-1-1025-2049-2049-8000
+5.8G    ../core-data-2-2049-4097-4097-4000
+5.5G    ../core-data-3-4097-8193-8193-2000
+4.8G    ../core-data-4-8193-16385-16385-1000
+2.9G    ../core-data-5-16385-32769-32769-500
+1.3G    ../core-data-6-32769-65537-65537-250
+573M    ../core-data-7-65537-131073-131073-125
+```
+
+```bash
+CUDA_VISIBLE_DEVICES=0 CUDA_LAUNCH_BLOCKING=0 PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True litgpt pretrain --config pretrain_core_model_0.yaml
+```
+
+```
+Seed set to 23
+Time to instantiate model: 0.21 seconds.
+Total parameters: 402,703,104
+Verifying settings ...
+Measured TFLOPs: 42432.35
+Epoch 1 | iter 64 step 1 | loss train: 11.984, val: n/a | iter time: 460.76 ms (step) remaining time: 12 days, 3:41:55
+Epoch 1 | iter 128 step 2 | loss train: 11.979, val: n/a | iter time: 402.83 ms (step) remaining time: 9 days, 0:57:24
+Epoch 1 | iter 192 step 3 | loss train: 11.983, val: n/a | iter time: 403.46 ms (step) remaining time: 8 days, 0:12:58
+Epoch 1 | iter 256 step 4 | loss train: 11.983, val: n/a | iter time: 403.39 ms (step) remaining time: 7 days, 11:52:07
+Epoch 1 | iter 320 step 5 | loss train: 11.979, val: n/a | iter time: 403.85 ms (step) remaining time: 7 days, 4:28:33
+Epoch 1 | iter 384 step 6 | loss train: 11.978, val: n/a | iter time: 403.93 ms (step) remaining time: 6 days, 23:33:15
+Epoch 1 | iter 448 step 7 | loss train: 11.978, val: n/a | iter time: 403.38 ms (step) remaining time: 6 days, 20:02:28
+Epoch 1 | iter 512 step 8 | loss train: 11.973, val: n/a | iter time: 403.80 ms (step) remaining time: 6 days, 17:24:49
+Epoch 1 | iter 576 step 9 | loss train: 11.972, val: n/a | iter time: 403.23 ms (step) remaining time: 6 days, 15:21:59
+Epoch 1 | iter 640 step 10 | loss train: 11.967, val: n/a | iter time: 403.38 ms (step) remaining time: 6 days, 13:43:53
+# ...
+```
+
+Backup `wandb`:
+
+```bash
+mv wandb wandb-pretrain-core-0
+```
+
+Copy config:
+
+```bash
+cp ../config-0.json ../out/pretrain-core-0/final/config.json
+```
+
+Chat with model:
+
+```bash
+CUDA_VISIBLE_DEVICES=0 CUDA_LAUNCH_BLOCKING=0 PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True litgpt chat ../out/pretrain-core-0/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 '4' --dtype 'bfloat16' '../out/pretrain-core-0/final'
+```
+
+```
+```