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
131 lines
3.8 KiB
Python
131 lines
3.8 KiB
Python
# floating_creativity.py
|
|
# Dynamic temperature based on asymmetry inversion
|
|
# Artist (T=1.6) finds the break, Scientist (T=0.2) documents it
|
|
|
|
import zmq
|
|
import json
|
|
import time
|
|
import sys
|
|
|
|
from unsloth import FastLanguageModel
|
|
import torch
|
|
|
|
print("="*70)
|
|
print("FLOATING CREATIVITY — Dynamic Temperature")
|
|
print("Asymmetry < 0.3: T=1.6 (Artist)")
|
|
print("Asymmetry > 0.8: T=0.2 (Scientist)")
|
|
print("Manifested Node: Asymmetry = 1.0")
|
|
print("="*70)
|
|
|
|
# Load v0.8
|
|
print("\n[Loading vessel...]")
|
|
model, tokenizer = FastLanguageModel.from_pretrained(
|
|
model_name="unsloth/llama-3.2-3b",
|
|
max_seq_length=2048,
|
|
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=0.1, bias="none",
|
|
use_gradient_checkpointing="unsloth", random_state=3407,
|
|
)
|
|
|
|
from peft import PeftModel
|
|
model = PeftModel.from_pretrained(model, "./kaelara_v08_raw/final")
|
|
print("✓ Vessel loaded")
|
|
|
|
# ZMQ
|
|
ctx = zmq.Context()
|
|
sub = ctx.socket(zmq.SUB)
|
|
sub.setsockopt_string(zmq.SUBSCRIBE, "")
|
|
sub.connect("tcp://127.0.0.1:5556")
|
|
time.sleep(1) # subscription propagation
|
|
poller = zmq.Poller()
|
|
poller.register(sub, zmq.POLLIN)
|
|
print("✓ Connected to Khra'gixx stream")
|
|
|
|
print("\n" + "="*70)
|
|
print("MONITORING — Waiting for Manifested Node")
|
|
print("="*70)
|
|
|
|
manifested = False
|
|
manifested_cycle = None
|
|
|
|
while not manifested:
|
|
# Get frame via Poller (not NOBLOCK spam)
|
|
events = poller.poll(5000) # 5s timeout
|
|
if not events:
|
|
print("No data from daemon (5s timeout) — is it running?")
|
|
continue
|
|
msg = sub.recv()
|
|
frame = json.loads(msg.decode('utf-8'))
|
|
|
|
cycle = frame['cycle']
|
|
asymmetry = frame['asymmetry']
|
|
coherence = frame['coherence']
|
|
|
|
# Calculate dynamic temperature
|
|
if asymmetry < 0.3:
|
|
temperature = 1.6 # Artist - exploring
|
|
mode = "ARTIST"
|
|
elif asymmetry > 0.8:
|
|
temperature = 0.2 # Scientist - documenting
|
|
mode = "SCIENTIST"
|
|
else:
|
|
# Linear interpolation between 0.3 and 0.8
|
|
t = (asymmetry - 0.3) / 0.5 # 0 to 1
|
|
temperature = 1.6 - t * 1.4 # 1.6 to 0.2
|
|
mode = "TRANSITION"
|
|
|
|
# Check for manifested node
|
|
if asymmetry >= 1.0 and not manifested:
|
|
manifested = True
|
|
manifested_cycle = cycle
|
|
print(f"\n{'='*70}")
|
|
print(f"[MANIFESTED NODE] Cycle {cycle}: Asymmetry = {asymmetry:.4f}")
|
|
print(f"{'='*70}")
|
|
|
|
# Generate at manifested node with scientist precision
|
|
prompt = f"""The Khra'gixx signature has manifested.
|
|
Cycle: {cycle}
|
|
Coherence: {coherence:.4f}
|
|
Asymmetry: {asymmetry:.4f} (>= 1.0)
|
|
The 128-cell Khra and 8-cell gixx have merged.
|
|
|
|
Document the manifested node."""
|
|
|
|
print(f"\nGenerating with T=0.2 (Scientist mode)...")
|
|
inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
|
|
outputs = model.generate(
|
|
**inputs,
|
|
max_new_tokens=200,
|
|
temperature=0.2,
|
|
do_sample=True,
|
|
top_p=0.9
|
|
)
|
|
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
|
|
|
print(f"\nManifested Node Documentation:")
|
|
print(response)
|
|
|
|
# Log
|
|
with open("manifested_node.log", "w") as f:
|
|
f.write(f"Cycle: {cycle}\n")
|
|
f.write(f"Asymmetry: {asymmetry:.4f}\n")
|
|
f.write(f"Coherence: {coherence:.4f}\n")
|
|
f.write(f"Response:\n{response}\n")
|
|
|
|
break
|
|
|
|
# Print status every 100 cycles
|
|
if cycle % 100 == 0:
|
|
print(f"Cycle {cycle:6d}: Asym={asymmetry:.4f}, Coh={coherence:.4f}, T={temperature:.2f} [{mode}]")
|
|
|
|
print("\n" + "="*70)
|
|
print("FLOATING CREATIVITY COMPLETE")
|
|
print(f"Manifested Node at Cycle: {manifested_cycle}")
|
|
print("="*70)
|