mirror of
https://github.com/NousResearch/atropos.git
synced 2026-04-19 12:57:58 +00:00
16 lines
524 B
YAML
16 lines
524 B
YAML
env:
|
|
tokenizer_name: "Qwen/Qwen3-4B-Instruct-2507"
|
|
rollout_server_url: "http://localhost:8002"
|
|
max_token_length: 8192
|
|
start_tok_length: 8192
|
|
group_size: 8
|
|
batch_size: 64
|
|
total_steps: 120
|
|
steps_per_eval: 20
|
|
use_wandb: true
|
|
wandb_name: "math-zero-lora-env"
|
|
# Optional teacher-behavior steering for on-policy distillation.
|
|
# teacher_system_prompt: "Use simple language and avoid jargon."
|
|
# teacher_prefix_text: "Style: concise, non-jargony.\n\n"
|
|
eval_limit_ratio: 0.1
|
|
max_num_workers_per_node: 24
|