mirror of
https://github.com/SHOGGOTH-SECTOR/sica-fondt.git
synced 2026-08-01 08:30:20 +00:00
24f816b6a3
Place useful parts of the surrounding repos into sica-fondt by layer, per the
body model (Ada = membrane; brain/endocrine/capabilities/knowledge non-Ada):
- brain/ LLM reasoning + providers (dapr, hermes, MoMoA)
- capabilities/ REPRAG sidecars: hermes tools/skills, dapr tools, parallel
dispatch, A51 channels, and the OSINT cluster
- knowledge/ LORAG corpus: 754 cyber-skills, agency personas, secure-coding,
MITRE ATT&CK data
- reference/ defensive threat-reference (C3, shhbruh doc) + AdaYaml parser
License handling: AGPL sources (worldosint, advanced_evolution, mercury,
Reticulum) and GPL DeTTECT are SPEC-only clean-room/port descriptions — no
copyleft code copied. MIT/Apache/data parts copied as working trees.
Safety: shhbruh escape/persistence material and C3 covert-C2 kept as reference
only, not wired into the running organism. See CONSOLIDATION.md.
https://claude.ai/code/session_01UehUqEXXJJCsHoA4voCU5c
1.9 KiB
1.9 KiB
Online RL Methods
Guide to online reinforcement learning with PPO, GRPO, RLOO, and OnlineDPO.
Overview
Online RL generates completions during training and optimizes based on rewards.
PPO (Proximal Policy Optimization)
Classic RL algorithm for LLM alignment.
Basic Usage
python -m trl.scripts.ppo \
--model_name_or_path Qwen/Qwen2.5-0.5B-Instruct \
--reward_model_path reward-model \
--dataset_name trl-internal-testing/descriptiveness-sentiment-trl-style \
--output_dir model-ppo \
--learning_rate 3e-6 \
--per_device_train_batch_size 64 \
--total_episodes 10000 \
--num_ppo_epochs 4 \
--kl_coef 0.05
Key Parameters
kl_coef: KL penalty (0.05-0.2)num_ppo_epochs: Epochs per batch (2-4)cliprange: PPO clip (0.1-0.3)vf_coef: Value function coef (0.1)
GRPO (Group Relative Policy Optimization)
Memory-efficient online RL.
Basic Usage
from trl import GRPOTrainer, GRPOConfig
from datasets import load_dataset
# Define reward function
def reward_func(completions, **kwargs):
return [len(set(c.split())) for c in completions]
config = GRPOConfig(
output_dir="model-grpo",
num_generations=4, # Completions per prompt
max_new_tokens=128
)
trainer = GRPOTrainer(
model="Qwen/Qwen2-0.5B-Instruct",
reward_funcs=reward_func,
args=config,
train_dataset=load_dataset("trl-lib/tldr", split="train")
)
trainer.train()
Key Parameters
num_generations: 2-8 completionsmax_new_tokens: 64-256- Learning rate: 1e-5 to 1e-4
Memory Comparison
| Method | Memory (7B) | Speed | Use Case |
|---|---|---|---|
| PPO | 40GB | Medium | Maximum control |
| GRPO | 24GB | Fast | Memory-constrained |
| OnlineDPO | 28GB | Fast | No reward model |
References
- PPO paper: https://arxiv.org/abs/1707.06347
- GRPO paper: https://arxiv.org/abs/2402.03300
- TRL docs: https://huggingface.co/docs/trl/