You can take the same source code package that Ubuntu uses to build jq, compile it again, and realize 90% better performance.
I use jq for processing GeoJSON files and other open data offered in JSON format. Today I am working with a 500MB GeoJSON file that contains the Alameda County Assessor's parcel map. I want to run a query that prints the city for every parcel worth more than a threshold amount. The program is
About 6 years ago when I was a PHP ecommerce dev, I've always wanted to work with distributed systems and microservices. They seemed so cool and the new way of architecting software systems because all of big tech did it. But now working with them day to day for about 3 years I came to realize that my initial anticipation was stupid. Well, not stupid in the traditional sense, because how could I have known what I'd get myself into.
Microservices and distributed systems have been preached to solve two major problems: Scale and scale.
So let's talk about them.
| base_model: /path/to/Mistral-7B-v0.1 | |
| model_type: AutoModelForCausalLM | |
| tokenizer_type: AutoTokenizer | |
| is_llama_derived_model: true | |
| load_in_8bit: false | |
| load_in_4bit: true | |
| strict: false | |
| datasets: |
| FROM ./nous-hermes-2-solar-10.7b.Q5_K_M.gguf | |
| PARAMETER num_ctx 4096 | |
| TEMPLATE """<|im_start|>system | |
| {{ .System }}<|im_end|> | |
| <|im_start|>user | |
| {{ .Prompt }}<|im_end|> | |
| <|im_start|>assistant | |
| """ | |
| PARAMETER stop "<|im_start|>" | |
| PARAMETER stop "<|im_end|>" |
| 1. # create new .py file with code found below | |
| 2. # install ollama | |
| 3. # install model you want “ollama run mistral” | |
| 4. conda create -n autogen python=3.11 | |
| 5. conda activate autogen | |
| 6. which python | |
| 7. python -m pip install pyautogen | |
| 7. ollama run mistral | |
| 8. ollama run codellama | |
| 9. # open new terminal |
| import os | |
| import autogen | |
| import memgpt.autogen.memgpt_agent as memgpt_autogen | |
| import memgpt.autogen.interface as autogen_interface | |
| import memgpt.agent as agent | |
| import memgpt.system as system | |
| import memgpt.utils as utils | |
| import memgpt.presets as presets | |
| import memgpt.constants as constants | |
| import memgpt.personas.personas as personas |
| #!/usr/bin/env python3 | |
| # SPDX-FileContributor: 2023 The Aerospace Corporation | |
| # SPDX-DocumentComment: Approved 2023-10-23 Request # OSS23-0008 | |
| # SPDX-License-Identifier: LGPL-3.0-or-later | |
| """Dataverse Scraper""" | |
| import argparse | |
| from pathlib import Path | |
| import hashlib | |
| import logging | |
| import subprocess |
Fine-tuning llama 2 7B to analyze financial reports and write “funny” tweets
Sharing some insights from a recent weekend fun project where I tried to analyze and summarize financial reports using a fine-tuned LLM.
My initial goal was to train a model to summarize the annual/quarterly financial reports of public companies (aka 10-K / 10-Q). But, realizing that straightforward financial summaries are boring, I thought of tuning LLM to generate sarcastic summaries of these reports. Something short I could post on Twitter.
Data exploration and dataset prep
Working with financial reports ain’t easy. You download them in html format, they’re pretty dense with ~100 pages filled with tables that can be tough to parse, many legal disclaimers and various useless info. I knew I wanted to get 3-5 funny tweets as an output from a report. But I spent quite some time figuring out what data to actually input to get the result - a page, a section, a table?
| import autogen | |
| from user_proxy_webagent import UserProxyWebAgent | |
| import asyncio | |
| config_list = [ | |
| { | |
| "model": "gpt-3.5-turbo", | |
| # "api_key": "<YOUR KEY HERE>" | |
| } | |
| ] |
