(by @andrestaltz)
If you prefer to watch video tutorials with live-coding, then check out this series I recorded with the same contents as in this article: Egghead.io - Introduction to Reactive Programming.
(by @andrestaltz)
If you prefer to watch video tutorials with live-coding, then check out this series I recorded with the same contents as in this article: Egghead.io - Introduction to Reactive Programming.
Yoav Goldberg, April 2023.
With the release of the ChatGPT model and followup large language models (LLMs), there was a lot of discussion of the importance of "RLHF training", that is, "reinforcement learning from human feedback". I was puzzled for a while as to why RL (Reinforcement Learning) is better than learning from demonstrations (a.k.a supervised learning) for training language models. Shouldn't learning from demonstrations (or, in language model terminology "instruction fine tuning", learning to immitate human written answers) be sufficient? I came up with a theoretical argument that was somewhat convincing. But I came to realize there is an additional argumment which not only supports the case of RL training, but also requires it, in particular for models like ChatGPT. This additional argument is spelled out in (the first half of) a talk by John Schulman from OpenAI. This post pretty much
* { | |
font-size: 12pt; | |
font-family: monospace; | |
font-weight: normal; | |
font-style: normal; | |
text-decoration: none; | |
color: black; | |
cursor: default; | |
} |
UPDATE (2021-03-31): I've posted an improved version of this guide with newer versions of the software.
I hate when my images turn out like this:
<a class="question_link" target="_blank" href="/question/30746665/answer/49332475">会写 Parser、Tokenizer 是什么水平?</a><br/><br/>大多数编译原理书前100页的内容,说明大学听了一半左右的编译原理课,通俗地说,写了这个只能证明你不是个棒槌。<br><br><br>所以其实你更应该关心不会tokenizer和parser是什么水平。 | |
<span class="answer-date-link-wrap"> | |
<a class="answer-date-link last_updated meta-item" data-tip="s$t$发布于 2015-05-29" target="_blank" href="/question/30746665/answer/49332475">编辑于 2015-05-29</a> | |
</span> | |
<hr/><a class="question_link" target="_blank" href="/question/30703519/answer/49150834">王垠到底对 winter 做了什么?</a><br/><br/>你可以理解为是路边看到一坨**,忍不住想去一脚踩爆它的心态。(虽然我知道这么做无聊而且会沾一脚)<br><br>想了想,可能还有一点觉得他的粉丝很可怜的,想让他们停止吃**的心态吧,虽然我知道"然而没卵用"。<br><br>回到题主的问题,要问做了什么,那就是“他是**,还碰巧被我看到了”,这样的事情,简直无法被饶恕。 | |
function epicycloid(t, rmaj, rmin) = [ (rmaj+rmin) * cos(t) - rmin * cos(((rmaj+rmin)/rmin)*t), | |
(rmaj+rmin) * sin(t) - rmin * sin(((rmaj+rmin)/rmin)*t) ]; | |
function hypocycloid(t, rmaj, rmin) = [ (rmaj-rmin) * cos(t) + rmin * cos(((rmaj-rmin)/rmin)*t), | |
(rmaj-rmin) * sin(t) - rmin * sin(((rmaj-rmin)/rmin)*t) ]; | |
module CycloidTooth( | |
rmaj = 30, | |
rmin = 20, | |
circ_pitch = 10.0, |
/* | |
* Demonstrate using an http server and an HTML form to control an LED. | |
* The http server runs on the ESP8266. | |
* | |
* Connect to "http://esp8266WebForm.local" or "http://<IP address>" | |
* to bring up an HTML form to control the LED connected GPIO#0. This works | |
* for the Adafruit ESP8266 HUZZAH but the LED may be on a different pin on | |
* other breakout boards. | |
* | |
* Imperatives to turn the LED on/off using a non-browser http client. |
# Based off of https://geekswipe.net/technology/computing/analyze-chromes-browsing-history-with-python/ | |
# Must use python 2 | |
# Close Chrome before running this, or else the database will be locked | |
import os | |
import sqlite3 | |
import operator | |
from collections import OrderedDict | |
import matplotlib.pyplot as plt |