mirror of
https://github.com/zenlm/zen-max-trainer.git
synced 2026-10-30 20:46:46 +00:00
zeekay
b09a0a318e
- Train Zen identity on K2 Thinking using QLoRA - 4-bit quantized training (no 1TB download needed) - Outputs LoRA adapters (~100MB) to zenlm/zen-max - Gradio interface for cloud training
10 lines
134 B
Plaintext
10 lines
134 B
Plaintext
gradio>=4.0.0
|
|
transformers>=4.36.0
|
|
torch>=2.1.0
|
|
peft>=0.7.0
|
|
bitsandbytes>=0.41.0
|
|
accelerate>=0.25.0
|
|
datasets>=2.15.0
|
|
trl>=0.7.0
|
|
scipy
|