jwatte

from typing import List
import json
import sys

from langchain.chat_models import ChatOpenAI
from langchain.prompts import ChatPromptTemplate
from langchain.output_parsers import PydanticOutputParser
from langchain.schema.messages import HumanMessage, SystemMessage
from langchain.chains import ConversationChain
from langchain.memory import ConversationBufferMemory

jwatte

[tool.poetry]
name = "langchain"
version = "0.0.338"
description = "Building applications with LLMs through composability"
authors = []
license = "MIT"
readme = "README.md"
repository = "https://github.com/langchain-ai/langchain"

[tool.poetry.scripts]

jwatte

# This is a Dockerfile for the Development Container
# see https://github.com/langchain-ai/langchain

# Use the Python base image
ARG VARIANT="3.11-bullseye"
FROM mcr.microsoft.com/devcontainers/python:0-${VARIANT} AS langchain-dev-base

USER vscode

# Define the version of Poetry to install (default is 1.4.2)

jwatte

test:   malloc-release malloc-debug
        ./malloc-debug
        ./malloc-release

malloc-debug:   MallocBenchmark.cpp
        clang -O0 -D_DEBUG=1 -g -o malloc-debug MallocBenchmark.cpp

malloc-release: MallocBenchmark.cpp
        clang -O3 -DNDEBUG=1 -o malloc-release MallocBenchmark.cpp

jwatte

import torch
from imagen_pytorch import Unet, Imagen, ImagenTrainer
from imagen_pytorch.t5 import t5_encode_text, DEFAULT_T5_NAME
from torchvision import transforms

# This doesn't yet format the text embedding tensors right
# TODO: check out https://gist.github.com/Netruk44/38d793e6d04a53cc4d9acbfadbb04a5c


import json

jwatte

# Pick N random annotations from the list of annotations.
# Load/ccale/sample the corresponding image.
# Generate the T5 text embedding of the given prompt.
# Upload the data to the appropriate slot in each tensor.
def load_images(annots, pathsbyid, tensordim, t_embeds, t_masks, t_images):
    # pick N random annotations
    todo = annots.copy()
    random.shuffle(todo)
    todo = todo[0: tensordim]
    for ix in range(0, tensordim):

jwatte

I gave GPT3 a writing prompt, to see how well it would keep track of initial 
conditions.

The writing that comes out is in the same ham-handed style as my prompt, it 
gets 100% on that! But as a writer / prompt completer:

  1. The conditions established in the first paragraph don't carry through -- there's
     enough oxygen, so why's the air running out right away?
  2. How can he feel the smell of the antenna array while in his suit on the moon?

jwatte

Trying to build on Ubuntu 17.04.

jwatte@ripper:~/Downloads/SlugDemo/Linux$ make
mkdir -p Output/Code
gcc -I../Code -I/usr/include -DSLUG_LINUX -DSLUG_OPENGL -m64 -msse -msse2 -std=c++11 -fno-exceptions -fno-rtti -Wall -Wno-multichar -Wno-strict-aliasing -Wno-unused-result -O3 -c ../Code/Main.cpp -o Output/Code/Main.o
gcc -I../Code -I/usr/include -DSLUG_LINUX -DSLUG_OPENGL -m64 -msse -msse2 -std=c++11 -fno-exceptions -fno-rtti -Wall -Wno-multichar -Wno-strict-aliasing -Wno-unused-result -O3 -c ../Code/SlugDemo.cpp -o Output/Code/SlugDemo.o
In file included from /usr/include/GL/glx.h:333:0,
                 from ../Code/SlugDemo.h:96,
                 from ../Code/Main.cpp:1:
/usr/include/GL/glxext.h:505:143: error: ‘GLintptr’ has not been declared

jwatte

class A a where
    getB :: B b => a -> b

class B b where
    getA :: A a => b -> a

data X = X

instance A X where
    getB x = x

jwatte

--    AUTO-GENERATED, DO NOT EDIT !!!
data CampaignTemplateMaybe = CampaignTemplateMaybe
    { maybe_name :: !(Maybe Text)
    , maybe_kind :: !(Maybe Text)
    , maybe_definition :: !(Maybe Text)
    , maybe_saved :: !(Maybe Bool)
    , maybe_hidden :: !(Maybe Bool)
    , maybe_last_active :: !(Maybe UTCTime)
    }
    deriving (Show, Read, Ord, Eq)