ngoodger

pkg update
pkg upgrade -y
pkg install wget git rust tmux which gradle python android-tools -y
git clone https://github.com/NvChad/starter ~/.config/nvim


# android setup
wget https://github.com/lzhiyong/termux-ndk/releases/download/android-ndk/android-ndk-r27b-aarch64.zip && unzip android-ndk-r27b-aarch64.zip && rm android-ndk-r27b-aarch64.zip

wget https://github.com/lzhiyong/termux-ndk/releases/download/android-sdk/android-sdk-aarch64.zip && unzip android-sdk-aarch64.zip && rm android-sdk-aarch64.zip

ngoodger

import jax.numpy as jnp
from jax import random

class JaxCartPole:
    """
    Based on OpenAI Gym Cartpole
    https://github.com/openai/gym/blob/master/gym/envs/classic_control/cartpole.py
    """
    def __init__(self):
        self.gravity = 9.8

ngoodger

def fori_body(i, val):
  env_state, action_key, all_obsv, all_reward, all_done = val
  action = random.randint(action_key, (1,), 0, 2)[0]
  action_key = random.split(action_key)[0]
  env_state, obsv, reward, done, info = env.step(env_state, action)
  all_obsv = all_obsv.at[i].set(obsv)
  all_reward = all_reward.at[i].set(reward)
  all_done = all_done.at[i].set(done)
  val = (env_state, action_key, all_obsv, all_reward, all_done)
  return val

ngoodger

import jax.numpy as jnp
from jax import random

class SkeletonEnv:
   
   def __init__(self):
      self.random_limit = 0.05
   
   def _get_obsv(self, state):
      return state

ngoodger

import gym
import math
import torch
from gym import spaces, logger
import numpy as np

class Pendulum(gym.Env):
    metadata = {
        'render.modes': ['human', 'rgb_array'],
        'video.frames_per_second': 30

ngoodger

ngoodger

ngoodger

import numpy as np
import requests
import time
import pyarrow
import torchvision
import torch
MAX_BATCH_SIZE = 1
TEST_CORRECT_OUTPUT = False
if TEST_CORRECT_OUTPUT:
    MODEL = torchvision.models.resnet18(pretrained=True).eval()

ngoodger

import numpy as np
import time
import torchvision
import torch
DEVICE = "cuda"
MODEL = torchvision.models.resnet18(pretrained=True).eval().to(DEVICE)
BATCH_SIZE = 64 
BATCHES = 100
x_cpu = torch.from_numpy(np.random.random((BATCH_SIZE, 3, 256, 256))
                     .astype(np.float32)).pin_memory()

ngoodger

from aiohttp import web
import argparse
import asyncio
from dataclasses import dataclass
from io import BytesIO
import numpy as np
import pyarrow
import prometheus_client as pc
import time
import torch