Scruff AI
56c71c87b2
- cuda/ — main LBM kernel (khra_gixx_1024_v5.cu) - navigator/ — lattice_observer, golden_weave, bridges, mock daemon - scripts/ — compile, start, launch, setup (paths updated) - docs/ — system manual - archive/ — everything else (old kernels, inquiries, experiments) - README.md — full setup guide: requirements, quick start, use your own LLM - removed training/ entirely (broken LoRA scripts + datasets) - .gitignore: exclude build/ logs/ training/ *.jsonl
122 lines
3.3 KiB
Python
122 lines
3.3 KiB
Python
# floating_creativity_v09.py
|
|
# v0.9 Deployment: Artist/Scientist Clutch
|
|
# T=1.6 at A<1.0, T=0.2 at A>1.0
|
|
|
|
import zmq
|
|
import json
|
|
import time
|
|
|
|
from unsloth import FastLanguageModel
|
|
import torch
|
|
|
|
print("="*70)
|
|
print("FLOATING CREATIVITY v0.9 — COGNITIVE CLUTCH")
|
|
print("Artist (T=1.6) at A<1.0 | Scientist (T=0.2) at A>1.0")
|
|
print("="*70)
|
|
|
|
# Load v0.9 Scientist
|
|
print("\n[Loading v0.9 Scientist...]")
|
|
model, tokenizer = FastLanguageModel.from_pretrained(
|
|
model_name="unsloth/llama-3.2-3b",
|
|
max_seq_length=512,
|
|
dtype=torch.bfloat16,
|
|
load_in_4bit=True,
|
|
)
|
|
|
|
model = FastLanguageModel.get_peft_model(
|
|
model,
|
|
r=64, target_modules=["q_proj", "k_proj", "v_proj", "o_proj"],
|
|
lora_alpha=128, lora_dropout=0, bias="none",
|
|
use_gradient_checkpointing="unsloth", random_state=3407,
|
|
)
|
|
|
|
from peft import PeftModel
|
|
model = PeftModel.from_pretrained(model, "./kaelara_v09_scientist/final")
|
|
print("✓ v0.9 Scientist loaded")
|
|
|
|
# ZMQ setup
|
|
ctx = zmq.Context()
|
|
sub = ctx.socket(zmq.SUB)
|
|
sub.setsockopt_string(zmq.SUBSCRIBE, "")
|
|
sub.connect("tcp://127.0.0.1:5556")
|
|
print("✓ Connected to ZMQ stream")
|
|
|
|
print("\n" + "="*70)
|
|
print("WAITING FOR ASYMMETRY ≈ 13.0")
|
|
print("="*70)
|
|
|
|
# Wait for A ≈ 13.0
|
|
frame = None
|
|
for i in range(100):
|
|
try:
|
|
msg = sub.recv(flags=zmq.NOBLOCK)
|
|
frame = json.loads(msg.decode('utf-8'))
|
|
asym = frame['asymmetry']
|
|
if 12.5 <= asym <= 13.5:
|
|
print(f"Cycle {frame['cycle']}: Asymmetry={asym:.2f} ✓")
|
|
break
|
|
elif i % 10 == 0:
|
|
print(f"Cycle {frame['cycle']}: Asymmetry={asym:.2f} (waiting for 12.5-13.5)")
|
|
except zmq.Again:
|
|
time.sleep(0.1)
|
|
|
|
if frame is None:
|
|
print("ERROR: No data received")
|
|
exit(1)
|
|
|
|
asymmetry = frame['asymmetry']
|
|
coherence = frame['coherence']
|
|
cycle = frame['cycle']
|
|
|
|
# Determine mode
|
|
if asymmetry < 1.0:
|
|
temperature = 1.6
|
|
mode = "ARTIST"
|
|
else:
|
|
temperature = 0.2
|
|
mode = "SCIENTIST"
|
|
|
|
print(f"\n{'='*70}")
|
|
print(f"CLUTCH ENGAGED — {mode} MODE")
|
|
print(f"Asymmetry: {asymmetry:.2f} | Coherence: {coherence:.3f} | T={temperature}")
|
|
print(f"{'='*70}")
|
|
|
|
# SYSTEM CHECK PROMPT — Match training format exactly
|
|
prompt = f"""Input: Asymmetry {asymmetry:.1f}, Coherence {coherence:.2f}. Define the current state of the 128/8 Khra'gixx injection. Is the 1024-grid in a Manifested Node state or Chaotic Drift? Report using [TAG] format.
|
|
|
|
Output:"""
|
|
|
|
print(f"\nPrompt: {prompt}")
|
|
print(f"\nGenerating with T={temperature}...")
|
|
|
|
inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
|
|
outputs = model.generate(
|
|
**inputs,
|
|
max_new_tokens=100,
|
|
temperature=temperature,
|
|
do_sample=True,
|
|
top_p=0.9
|
|
)
|
|
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
|
|
|
print(f"\n{'='*70}")
|
|
print("v0.9 RESPONSE")
|
|
print(f"{'='*70}")
|
|
print(response)
|
|
|
|
# Log to MANIFESTED_REALITY_v09.log
|
|
with open("MANIFESTED_REALITY_v09.log", "w") as f:
|
|
f.write(f"{'='*70}\n")
|
|
f.write(f"FLOATING CREATIVITY v0.9 — FIRST CONTACT\n")
|
|
f.write(f"Cycle: {cycle}\n")
|
|
f.write(f"Asymmetry: {asymmetry:.4f}\n")
|
|
f.write(f"Coherence: {coherence:.4f}\n")
|
|
f.write(f"Mode: {mode} (T={temperature})\n")
|
|
f.write(f"{'='*70}\n\n")
|
|
f.write(f"PROMPT:\n{prompt}\n\n")
|
|
f.write(f"RESPONSE:\n{response}\n")
|
|
|
|
print(f"\n{'='*70}")
|
|
print("SAVED TO: MANIFESTED_REALITY_v09.log")
|
|
print(f"{'='*70}")
|