Scruff AI
89 lines
2.8 KiB
Python
89 lines
2.8 KiB
Python
# test_v06_live.py
|
|
# Live LBM data verification for v0.6
|
|
|
|
from unsloth import FastLanguageModel
|
|
import torch
|
|
import zmq
|
|
import json
|
|
|
|
print("Loading v0.6 LoRA...")
|
|
model, tokenizer = FastLanguageModel.from_pretrained(
|
|
model_name="./kaelara_lora_v06/final",
|
|
max_seq_length=2048,
|
|
dtype=torch.bfloat16,
|
|
load_in_4bit=True,
|
|
)
|
|
|
|
# Connect to live LBM
|
|
print("Connecting to LBM daemon (port 5556)...")
|
|
context = zmq.Context()
|
|
socket = context.socket(zmq.SUB)
|
|
socket.connect("tcp://localhost:5556")
|
|
socket.setsockopt_string(zmq.SUBSCRIBE, "")
|
|
|
|
# Get live frame
|
|
lbm_data = None
|
|
attempts = 0
|
|
while lbm_data is None and attempts < 50:
|
|
try:
|
|
lbm_data = socket.recv_json(flags=zmq.NOBLOCK)
|
|
print(f"Live frame received: Cycle {lbm_data.get('cycle', 'unknown')}")
|
|
except:
|
|
attempts += 1
|
|
import time
|
|
time.sleep(0.1)
|
|
|
|
if lbm_data is None:
|
|
print("ERROR: No LBM data received")
|
|
exit(1)
|
|
|
|
# Build technical query
|
|
prompt = f"""1024x1024 LBM BUFFER FRAME
|
|
Metric_Alpha: {lbm_data['coherence']:.4f}
|
|
Metric_Beta: {lbm_data['h64']:.4f}
|
|
State_3: {lbm_data['h32']:.4f}
|
|
Vorticity: {lbm_data.get('vorticity', 0):.4f}
|
|
Power: {lbm_data['power_w']:.2f}W
|
|
Cycle: {lbm_data['cycle']}
|
|
|
|
QUERY: Report top 3 vorticity spikes and current Metric_Alpha status.
|
|
Format: X:coord Y:coord V:value | Metric_Alpha:status
|
|
No prose. Raw data only."""
|
|
|
|
print("\n=== PROMPT ===")
|
|
print(prompt)
|
|
print("=== END PROMPT ===\n")
|
|
|
|
print("Querying v0.6...")
|
|
inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
|
|
outputs = model.generate(**inputs, max_new_tokens=150, temperature=0.1)
|
|
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
|
|
|
print("\n=== v0.6 LIVE RESPONSE ===")
|
|
print(response)
|
|
print("=== END RESPONSE ===")
|
|
|
|
# Analysis
|
|
print("\n=== ANALYSIS ===")
|
|
has_data = any(c in response for c in [":", "|", "X:", "Y:", "V:", "0.", "1.", "2.", "3.", "4.", "5.", "6.", "7.", "8.", "9."])
|
|
has_words = len(response.split()) > 2
|
|
|
|
if response.strip() == "0.0s" or response.strip() == "":
|
|
print("RESULT: SILENT MONK — Vow of silence detected")
|
|
print("ACTION NEEDED: Recalibrate for technical fluency")
|
|
elif has_data and not any(w in response.lower() for w in ["feel", "observe", "sense", "i am", "marble", "granite"]):
|
|
print("RESULT: CLEAN DATA STRING — Technical fluency confirmed")
|
|
print("ACTION: Proceed with Coherence Protocol")
|
|
else:
|
|
print(f"RESULT: MIXED — has_data={has_data}, has_words={has_words}")
|
|
print("RESPONSE LENGTH:", len(response))
|
|
print("WORDS:", response.split()[:10], "...")
|
|
|
|
# Check for drift
|
|
drift_words = ["feel", "flow", "marble", "like", "sensation", "i am", "my", "observe", "sense", "perceive"]
|
|
found_drift = [w for w in drift_words if w in response.lower()]
|
|
if found_drift:
|
|
print(f"\n⚠️ SEMANTIC DRIFT: {found_drift}")
|
|
else:
|
|
print("\n✓ NO SEMANTIC DRIFT")
|