LLMlow Training Center

Current Model

Parameters
-
Architecture
SSM
Trained
-
Speed
-

Training Plan

Recommended: 25M Parameters on Language Understanding

DatasetPurposeSizeStatus
Dolly-15kIntent understanding + responses~13MBReady
TinyStoriesGrammar + sentence structure~50MB subsetReady
Custom (LLMlow)Tool routing + self-awareness~58KBAvailable

How to Train

1
Install dependencies
pip install torch datasets
2
Download datasets
python -c "
from datasets import load_dataset
ds = load_dataset('databricks/databricks-dolly-15k')
with open('data/dolly.txt', 'w') as f:
    for item in ds['train']:
        f.write(f'User: {item[\"instruction\"]}\nAssistant: {item[\"response\"]}\n\n')
"
3
Train the model
python trainer/train.py \
  --data data/ \
  --output models/brain1.bin \
  --d-model 256 --d-state 128 --n-layers 8 \
  --epochs 5 --batch-size 16 --lr 0.0008
4
Restart LLMlow — it auto-loads the new model
systemctl restart llmlow

Time Estimates

Machine5M params25M params50M params
CPU (8 cores)30-60 min2-4 hours8-12 hours
GPU (RTX 3060)5-10 min20-40 min1-2 hours
GPU (A100)1-2 min5-10 min15-30 min

Model Tiers

TierParamsRAMQuality
PICO50M~40MBBasic routing
MICRO120M~80MBDecent
MINI500M~350MBGood
BASE1.5B~1GBStrong
PRO3B~2GBVery strong
ADVANCED7B+~4GB+Near-frontier

What Training Teaches

Training teaches Brain 1 to understand language structure, not world knowledge. After training, Brain 1 understands:

World knowledge comes from tools (web search, files, crystal). Brain 1 stays small. Intelligence grows in the Crystal.