mirror of
https://github.com/zenlm/zen-coder-flash.git
synced 2026-07-25 18:31:40 +00:00
Zach Kelling
d658451474
- 8x H200 training config for Nebius/cloud - LoRA training script (rank 64, alpha 128) - Dataset: hanzoai/zen-agentic-dataset-private (10.5B tokens) - Launch script with SLURM and Docker support
36 lines
533 B
Plaintext
36 lines
533 B
Plaintext
# Zen Coder Flash - Dependencies
|
|
# Install: pip install -r requirements.txt
|
|
|
|
# Core Training
|
|
torch>=2.2.0
|
|
transformers>=4.45.0
|
|
accelerate>=0.30.0
|
|
peft>=0.10.0
|
|
bitsandbytes>=0.43.0
|
|
datasets>=2.18.0
|
|
|
|
# Distributed Training
|
|
deepspeed>=0.14.0
|
|
|
|
# Flash Attention (optional but recommended)
|
|
flash-attn>=2.5.0
|
|
|
|
# Data Processing
|
|
tiktoken>=0.6.0
|
|
sentencepiece>=0.2.0
|
|
safetensors>=0.4.0
|
|
|
|
# Logging
|
|
wandb>=0.16.0
|
|
tensorboard>=2.16.0
|
|
|
|
# HuggingFace Hub
|
|
huggingface-hub>=0.22.0
|
|
|
|
# Config
|
|
pyyaml>=6.0
|
|
|
|
# Evaluation
|
|
evaluate>=0.4.0
|
|
lm-eval>=0.4.0
|