unsloth/studio/backend/assets/configs/model_defaults/other/unsloth_tinyllama-bnb-4bit.yaml
2026-03-09 09:54:52 +00:00

47 lines
948 B
YAML

# Model defaults for unsloth/tinyllama
# Based on TinyLlama_(1.1B)-Alpaca.ipynb
# Also applies to: TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T
training:
trust_remote_code: false
max_seq_length: 4096
# num_epochs: 1
num_epochs: 0
learning_rate: 2e-5
batch_size: 2
gradient_accumulation_steps: 4
warmup_ratio: 0.1
max_steps: 30
save_steps: 30
weight_decay: 0.1
random_seed: 3407
packing: true
train_on_completions: true
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 32
lora_alpha: 32
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10
inference:
trust_remote_code: false