scturtle

import os
import struct

class PalMessages:
    def __init__(self, pal_path):
        self.pal_path = pal_path

        mkf_path = os.path.join(self.pal_path, 'Sss.mkf')
        self._index_data = self._read_mkf_subfile(mkf_path, 3)

scturtle

import sys
import re

def parse_bdf_and_convert(input_file, output_file):
    MAX_CHARS = 65536
    font_data = [[0, 0, 0, 0] for _ in range(MAX_CHARS)]
    font_wide = [False] * MAX_CHARS
    current_char = -1
    in_bitmap = False
    bitmap_lines = []

scturtle

from pathlib import Path
import torch
import torch.nn as nn
import torch.nn.functional as F
from safetensors import safe_open
from tokenizers import Tokenizer

CFG = {
    "vocab_size": 262_144,
    "context_length": 32_768,

scturtle

from dataclasses import dataclass
from pathlib import Path

import torch
from torch import nn
import torch.nn.functional as F

from safetensors import safe_open
from tokenizers import Tokenizer

scturtle

import os
from functools import lru_cache

import torch
from torch import nn
import torch.nn.functional as F

from transformers import Qwen3Config
from transformers import Qwen2TokenizerFast

scturtle

#!/usr/bin/env python3

# https://github.com/FateScript/experiments/blob/main/se/mpi/mpi.py
# https://github.com/facebookincubator/gloo/tree/main/gloo

import math
import multiprocessing
import os

import numpy as np

scturtle

import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
import numpy as np
import gym

# Hyperparameters
num_inputs = 4
num_actions = 2

scturtle

× /
← → ↔ ⇐ ⇒ ⇔
⟵ ⟶ ⟷ ⟸ ⟹ ⟺
∀ ∃
∂ ∫
∈ ∊ ∉
∏ ∑
∙ ⋅
√
∞

scturtle

import numpy as np

class ModelArgs:
    dim = 288
    n_layers = 6
    n_heads = 6
    norm_eps = 1e-6

def build_cos_sin_cache(head_dim, seq_len, base=10000):
    theta = 1. / (base ** (np.arange(0, head_dim, 2, dtype=np.float32) / head_dim))

scturtle

import numpy as np

N_inp = 64
N_out = 64
d = 128
Q = np.random.randn(N_out, d)
K = np.random.randn(N_inp, d)
V = np.random.randn(N_inp, d)
O = np.random.randn(N_out, d)