Ruby/model/train.py
2025-04-25 22:29:48 -04:00

46 lines
1.2 KiB
Python

import torch
import torch.nn as nn
import random
import time
from model.brain import model, tokenizer, DEVICE, optimizer, loss_fn, daydream
from context.context import get_recent_context, add_to_context
_last_thought = time.time()
LOSS_FILE = "data/logs/loss.log"
def log_loss(value: float):
with open(LOSS_FILE, "a", encoding="utf-8") as f:
f.write(f"{time.time()},{round(value, 4)}\n")
def train_on_message(text: str):
global _last_thought
model.train()
context_texts = get_recent_context(3)
augmented_text = " ".join(context_texts + [text])
tokens = tokenizer.tokenize(augmented_text)
if len(tokens) < 2:
return
input_tensor = torch.tensor(tokens[:-1], dtype=torch.long, device=DEVICE).unsqueeze(0)
target_tensor = torch.tensor(tokens[1:], dtype=torch.long, device=DEVICE).unsqueeze(0)
output = model(input_tensor)
loss = loss_fn(output.view(-1, output.size(-1)), target_tensor.view(-1))
log_loss(loss.item())
optimizer.zero_grad()
loss.backward()
optimizer.step()
add_to_context(text)
now = time.time()
if now - _last_thought > 15:
for _ in range(3):
daydream()
_last_thought = now