GitXplorerGitXplorer
p

maxtext-no-tools

public
0 stars
0 forks
0 issues

Commits

List of commits on branch main.
Verified
485982a1d166ded1b39803d8e8692d127e0367e7

Update input_pipeline_interface.py

pperegilk committed a month ago
Verified
b707298d24a14e6b6c507b6a13dea6f1d01706eb

Create input_pipeline_interface.py

pperegilk committed a month ago
Verified
ad98598b3500b7039650bf55fd947bf64fd2cf7f

Update train_debug.py

pperegilk committed a month ago
Verified
956622bc48f80df63126e58af8bfffbe55dee34a

Update train_debug.py

pperegilk committed a month ago
Verified
0abc8610e950abdb80ea5f7beb9f1988d6773d08

Update train_debug.py

pperegilk committed a month ago
Verified
4a9f4b1ecdd9431e1c024ae1282b44c5a4a2b2b5

Create train_debug.py

pperegilk committed a month ago

README

The README file for this repository.

maxtext-no-tools

# Clone the repo at the root. Make sure maxtext is installed.
# pip install torch
# pip install pyconfig
cd
cp maxtext-no-tools/convert_pipeline.sh maxtext/
cp maxtext-no-tools/llama_or_mistral_orbax_to_huggingface.py maxtext/MaxText/
cd maxtext/

# Script defaults to converting llama3
./convert_pipeline.sh --checkpoints 80000 --checkpoint_names north_llama3_edu_above_1_lr1e5_8192

# Convert llama3.1
bash convert_pipeline.sh --checkpoints 0 --checkpoint_names north_llama31instruct_test_edu_above_1_add_instructions_lr1e5_8192 --model_name llama3.1-8b --tokenizer_dir north/llama3.1-8b-reference