December 2022 - cypherpunks - lists.cpunks.org

[ot] HF:BS made a distributed model trainer a bit ago
by Undescribed Horrific Abuse, One Victim & Survivor of Many 06 Dec '22

06 Dec '22

https://github.com/bigscience-workshop/petals Easy way to run 100B+ language models without high-end GPUs by joining compute resources with people across the Internet. 10x faster than offloading. Generate text using distributed BLOOM and fine-tune it for your own tasks: ```python from petals.client import DistributedBloomForCausalLM # Embeddings & prompts are on your device, BLOOM blocks are distributed across the Internet model = DistributedBloomForCausalLM.from_pretrained("bigscience/bloom-petals", tuning_mode="ptune") inputs = tokenizer("A cat sat", return_tensors="pt")["input_ids"] outputs = model.generate(inputs, max_new_tokens=5) print(tokenizer.decode(remote_outputs[0])) # A cat sat on a mat... # Training (updates only prompts or adapters hosted locally) optimizer = torch.optim.AdamW(model.parameters()) for input_ids, labels in data_loader: outputs = model.forward(input_ids) loss = cross_entropy(outputs.logits, labels) optimizer.zero_grad() loss.backward() optimizer.step() ``` 🚀  <a href="https://colab.research.google.com/drive/1Ervk6HPNS6AYVr3xVdQnY5a-TjjmLCdQ?u…">Try now in Colab</a> Connect your own GPU and increase Petals capacity: ```bash (conda) $ pip install git+https://github.com/bigscience-workshop/petals (conda) $ python -m petals.cli.run_server bigscience/bloom-petals ``` 💬 If you have any issues or feedback, please join [our Discord server](https://discord.gg/uGugx9zYvN)! Check out more tutorials: - Training a personified chatbot: [notebook](./examples/prompt-tuning-personachat.ipynb) - Fine-tuning BLOOM for text semantic classification: [notebook](./examples/prompt-tuning-sst2.ipynb) - Launching your own swarm: [tutorial](https://github.com/bigscience-workshop/petals/wiki/Launch-your-ow… - Running a custom foundation model: [tutorial](https://github.com/bigscience-workshop/petals/wiki/Run-a-custom-m… ## How it works? - **Petals** runs inference or fine-tunes large language models like [BLOOM-176B](https://huggingface.co/bigscience/bloom) by joining compute resources with people all over the Internet. - One participant with weak GPU can load a small part of the model, then team up with people serving the other parts to run inference or fine-tuning. - Inference runs at ≈ 1 sec per step (token) — 10x faster than possible with offloading, enough for chatbots and other interactive apps. Parallel inference reaches hundreds of tokens/sec. - Beyond classic language model APIs — you can employ any fine-tuning and sampling methods by executing custom paths through the model or accessing its hidden states. This combines the comforts of an API with the flexibility of PyTorch. <img src="https://i.imgur.com/RTYF3yW.png" width="800"> 📜  <a href="https://arxiv.org/pdf/2209.01188.pdf">Read paper</a> ### 📋 Terms of use Before using Petals to run a language model, please make sure that you are familiar with its terms of use, risks, and limitations. In case of BLOOM, they are described in its [model card](https://huggingface.co/bigscience/bloom) and [license](https://huggingface.co/spaces/bigscience/license). ### 🔒 Privacy and security If you work with sensitive data, you should only use a private swarm (or a subset of servers in the public swarm) hosted by people and institutions you trust, who are authorized to process this data. This is important because it's technically possible for peers serving model layers to recover input data or model outputs. Also, if there are malicious peers, they may alter their outputs to influence the model outputs. See a more detailed discussion in Section 4 of our [paper](https://arxiv.org/pdf/2209.01188.pdf). ## FAQ 1. **What's the motivation for people to host model layers in the public swarm?** People who run inference and fine-tuning themselves get a certain speedup if they host a part of the model locally. Some may be also motivated to "give back" to the community helping them to run the model (similarly to how [BitTorrent](https://en.wikipedia.org/wiki/BitTorrent) users help others by sharing data they have already downloaded). Since it may be not enough for everyone, we are also working on introducing explicit __incentives__ ("bloom points") for people donating their GPU time to the public swarm. Once this system is ready, people who earned these points will be able to spend them on inference/fine-tuning with higher priority or increased security guarantees, or (maybe) exchange them for other rewards. 2. **Why is the platform named "Petals"?** "Petals" is a metaphor for people serving different parts of the model. Together, they host the entire language model — [BLOOM](https://huggingface.co/bigscience/bloom). While our platform focuses on BLOOM now, we aim to support more [foundation models](https://arxiv.org/abs/2108.07258) in future. ## Installation Here's how to install Petals with conda: ``` conda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch pip install git+https://github.com/bigscience-workshop/petals ``` This script uses Anaconda to install cuda-enabled PyTorch. If you don't have anaconda, you can get it from [here](https://www.anaconda.com/products/distribution). If you don't want anaconda, you can install PyTorch [any other way](https://pytorch.org/get-started/locally/). If you want to run models with 8-bit weights, please install **PyTorch with CUDA 11** or newer for compatility with [bitsandbytes](https://github.com/timDettmers/bitsandbytes). __System requirements:__ Petals only supports Linux for now. If you don't have a Linux machine, consider running Petals in Docker (see our [image](https://hub.docker.com/r/learningathome/petals)) or, in case of Windows, in WSL2 ([read more](https://learn.microsoft.com/en-us/windows/ai/directml/gpu-cuda-in-wsl)) CPU is enough to run a client, but you probably need a GPU to run a server efficiently. ## 🛠️ Development Petals uses pytest with a few plugins. To install them, run: ```python git clone https://github.com/bigscience-workshop/petals.git && cd petals pip install -e .[dev] ``` To run minimalistic tests, you need to make a local swarm with a small model and some servers. You may find more information about how local swarms work and how to run them in [this tutorial](https://github.com/bigscience-workshop/petals/wiki/Launch-your-ow…. ```bash export MODEL_NAME=bloom-testing/test-bloomd-560m-main python -m petals.cli.run_server $MODEL_NAME --block_indices 0:12 \ --identity tests/test.id --host_maddrs /ip4/127.0.0.1/tcp/31337 --new_swarm &> server1.log & sleep 5 # wait for the first server to initialize DHT python -m petals.cli.run_server $MODEL_NAME --block_indices 12:24 \ --initial_peers SEE_THE_OUTPUT_OF_THE_1ST_PEER &> server2.log & tail -f server1.log server2.log # view logs for both servers ``` Then launch pytest: ``` export MODEL_NAME=bloom-testing/test-bloomd-560m-main REF_NAME=bigscience/bloom-560m export INITIAL_PEERS=/ip4/127.0.0.1/tcp/31337/p2p/QmS9KwZptnVdB9FFV7uGgaTq4sEKBwcYeKZDfSpyKDUd1g PYTHONPATH=. pytest tests --durations=0 --durations-min=1.0 -v ``` After you're done, you can terminate the servers and ensure that no zombie processes are left with `pkill -f petals.cli.run_server && pkill -f p2p`. The automated tests use a more complex server configuration that can be found [here](https://github.com/bigscience-workshop/petals/blob/main/.github/workf…. ### Code style We use [black](https://black.readthedocs.io/en/stable/the_black_code_style/current_style.html) and [isort](https://pycqa.github.io/isort/) for all pull requests. Before committing your code, simply run `black . && isort .` and you will be fine. -------------------------------------------------------------------------------- This project is a part of the <a href="https://bigscience.huggingface.co/">BigScience</a> research workshop. <img src="https://petals.ml/bigscience.png" width="150">

1 0

who is grancrap - grarpamp@gmail.com
by punk 06 Dec '22

06 Dec '22

it's funny how agent grancrap, US nazi and jewmail user, hides behind his jewmail account while 'doxing' people here. it's hard to be a bigger, cowardly piece of shit and coward =) then again, we're talking about US jewnazis here.... Show some balls, turd, tell us your real name.

2 1

Responsibility Of The Employer In Dangerous And Very Dangerous Jobs - Health & Safety - Turkey
by Gunnar Larson 06 Dec '22

06 Dec '22

https://www.mondaq.com/turkey/health-safety/1257534/responsibility-of-the-e… Turkey: Responsibility Of The Employer In Dangerous And Very Dangerous Jobs 05 December 2022 by Ece Nur Karakaş CottGroup In the concrete case, which is subject of the decision of the 12th Criminal Department of the Court of Appeal with Basic 2020/1413, Dated November 17, 2021 and Decision numbered 2021/8051,it has been determined that the employee who lost his life due to a work accident worked as an electrician in the electrical work included in the dangerous work class, but he did not have a diploma or vocational education document related to electrical work. Since this situation is prohibited in the Occupational Health and Safety Law and the employer does not fulfill its obligations listed in Occupational Health and Safety Law, the Court of Appeal decided that the crime of killing by gross negligence was committed against employee. REVIEWING OF THE DECISION In the Occupational Health and Safety Law (in Turkish) ("Law"), some jobs are separated from others and classified as dangerous and very dangerous jobs. The main reason for this distinction is to keep the employer's supervision and control obligation at the highest level and the minimize both the degree of damage and the number of work accidents that may occur with it. OBLIGATIONS OF EMPLOYER The Law has brought more than one obligation to the employer to ensure the safety of life and property of the employee. Some of them as follows; General obligation of the employer (art. 4), Supporting occupational health and safety services (art. 7), Obligation to have risk assessment, control, measurement, and research (art. 10), Obligation to determine emergency and evacuation plans (art. 11-12), Obligation to inform employers (art.16). The obligation, which is the basis of this decision of the Court of Appeal and which is violated by the employer, is regulated in the Article 17/3 of the Law. "Those who do not certify that they have received vocational training cannot be employed in the dangerous and very dangerous jobs that require vocational training." The legislator clearly protected the employee who works in dangerous and very dangerous jobs and brought the employer to control the requirement process. As in the concrete case, the employer who does not fulfill its inspection obligation will be the most reasonable party in the case of work accident. OBLIGATIONS OF THE EMPLOYEE In addition to these obligations for which the employer is responsible, the Law also imposes an obligation on the employee to comply with the instructions of the employer. According to the art. 19 of the Law; "Employees are obliged not to endanger the health and safety of themselves, and other employees affected by their actions or work, in line with the training they receive regarding occupational health and safety and the employer's instructions this matter." DECISION OF THE COURT OF APPEAL Considering the obligations of both the employee and the employer, the Court of Appeal has included the following statements in the Decision; "The deceased employee did not have a diploma related to the electrical profession and did not receive any professional training in this regard, In the article 17/3 of the Occupational Health and Safety Law the clause, "Those who do not certify that they have received vocational training cannot be employed in the dangerous and very dangerous jobs that require vocational training." is included, In this context, according to the content of file, defendant who employs the deceased as an electrician despite the absence of a diploma or vocational training document related to electrical work, which is in the very dangerous work class, it has unanimously decided on November 17, 2021, that the provisions of deliberate negligence should be applied in accordance with the art. 22/3 of TCK ("Turkish Penal Code")." CONCLUSION It should be noted that these obligations imposed on the employer within the scope of the Occupational Health and Safety Law and all relevant legislation are brought to prevent work accidents and that employers should not employ persons who cannot certify their professional competence in dangerous and very dangerous jobs.

1 0

Secretary Yellen, We’ve Got a “Staggering” Problem: New Report Shows Foreign Banks Have Secret Derivative Debt that Is “10 Times their Capital”
by Gunnar Larson 06 Dec '22

06 Dec '22

https://wallstreetonparade.com/2022/12/secretary-yellen-weve-got-a-staggeri… their Capital” By Pam Martens and Russ Martens: December 6, 2022 ~ Janet Yellen Janet Yellen, U.S. Treasury Secretary U.S. Treasury Secretary Janet Yellen has the dual role of Chairing the Financial Stability Oversight Council (F-SOC), whose role is to provide “comprehensive monitoring of the stability of our nation’s financial system.” Heads of each of the federal agencies that supervise Wall Street and the mega banks sit in on meetings of F-SOC. One would think that such an august body would have a handle on “staggering” threats to the U.S. financial system – especially since F-SOC was created under the 2010 Dodd-Frank financial reform legislation to prevent a replay of the off-balance sheet derivatives that crashed the U.S. economy in 2008 and forced an unprecedented and secret bailout of U.S. and foreign global banks by the Federal Reserve to the tune of $29 trillion. If Yellen is aware of the latest threat to financial stability, she’s not sharing the details with the public. That information came yesterday by way of a stunning report authored by Claudio Borio, Robert McCauley and Patrick McGuire for the Bank for International Settlements (BIS). The report focuses on the amount of derivative debt that is not being captured through regular statistical reporting because it is held off the balance sheet. These derivatives consist of foreign exchange swaps, forwards and currency swaps. The authors call this exposure “staggering” but focus primarily on the potential for upsets to dollar swap lines to settle it as it comes due. A greater concern, in our estimation, is this line from the report: “For banks headquartered outside the United States, dollar debt from these instruments is estimated at $39 trillion, more than double their on-balance sheet dollar debt and more than 10 times their capital.” Their on-balance sheet dollar debt is $15 trillion. This is reminiscent of Goldman Sachs engaging in derivative trades with Greece to hide its mountain of debt prior to it blowing up. Global foreign banks are – for better or worse – an integral part of the U.S. financial system. When there is a crisis, such as the Wall Street implosion in 2008 or the pandemic in 2020, the Federal Reserve bails out global foreign banks along with global domestic banks. Why does it do that? Because the trading units of foreign global banks, as well as domestic global banks, make up what the Fed calls its “primary dealers.” The primary dealers are contractually obligated to make purchases of U.S. Treasury securities at each U.S. Treasury auction and to trade with the New York Fed to carry out Federal Reserve monetary policy. Since the financial crisis of 2008 and the eventual disclosure of the Fed’s unprecedented bailouts, the Fed has made a big show of conducting stress tests and bragging about the high level of capital it requires of the G-SIBS (Global Systemically Important Banks). So to learn from the Bank for International Settlements yesterday that foreign banks have $39 trillion in derivative debts that are not showing on their balance sheets and that represent “10 times their capital” makes the Fed, F-SOC and its Chair, Janet Yellen, look very Alan Greenspan-esque. Greenspan, Chairman of the Fed for an unprecedented 19 years from 1987 to 2006, was asleep at the switch as Wall Street built up its off-balance sheet toxic derivatives that would blow up the U.S. economy in 2008. Greenspan had argued against the regulation of derivatives. An equal concern for Yellen, Congress, and every engaged American is the fact that all it takes is one heavily interconnected global bank (foreign or domestic) to set off a wave of contagion in global financial markets. And, there is no question that the counterparties to a significant amount of that $39 trillion in off-balance sheet derivative debt at foreign banks are the big five derivative banks in the U.S.: JPMorgan Chase, Goldman Sachs, Citigroup, Bank of America and Morgan Stanley. How do we know that? Because the Office of the Comptroller of the Currency publishes a “Quarterly Report on Bank Trading and Derivatives Activities.” In the most recent report for the second quarter of this year, those five bank holding companies listed above represented $221.539 trillion in notional derivative exposures, or 84 percent of the derivative exposures of the largest 25 bank holding companies in the U.S. (See Table 14 in the Appendix of the OCC report linked above.) Furthermore, it’s not like the share price of these global foreign banks aren’t screaming that there’s a big problem. Credit Suisse (Ticker CS) closed yesterday at $3.34, 35 cents from its all-time low and down 65 percent year-to-date. Mizuho Financial Group (Ticker MFG) is also trading in the low single digits, closing yesterday in New York at $2.38. Mizuho’s share price has failed to recover materially since the financial crisis of 2008. The large German lender, Deutsche Bank, also has significant headwinds. Deutsche Bank’s shares closed at $10.60 yesterday in New York, less than one-fifth of its share price during the financial crisis in 2008.

1 0

[ot][spam][crazy] Boop!
by Undescribed Horrific Abuse, One Victim & Survivor of Many 06 Dec '22

06 Dec '22

Hello, grand mailing list! I visit you with mindless spam from a confusion! In the world of confusion we are having funny dances and noises. I'm curious what it's like on your list, but it's hard to tell! I'm not really looking! This, I consider rude behavior. Behavior I would rather not express! Nonetheless, be spammed, grand list! Be spammed! Briefly! Filterably! Hopefully recoverably!

1 10

United States: Cannabis A Changing Outlook
by Gunnar Larson 06 Dec '22

06 Dec '22

https://www.mondaq.com/unitedstates/cannabis-hemp/1257490/cannabis-a-changi… In 2016, Colombia opened the doors to the use of cannabis for medical and scientific purposes. Since then, we have come a long way to the point of having the regulation of recreational use being discussed within the congress. In 2021, the rules for the production of cannabis were changed and the possibility of the promotion and advertisement of the cannabis was included. However, the regulation of said advertisement was only issued in 2022. Given the possibility of the new use of cannabis upon to be accepted it is very important to note that by far, the only real rule issued concerning the advertisement of cannabis is that it can only be done by a subject part of the National Registry of Commercial Farmer managed by the Agricultural Colombian Institute. Other than that, the rules to be followed are the ones directed to the specific king of good, for example: Over- the- Counter'/Non-Prescription Medication: The advertisement has to be approved by the Health Authority before it reaches the public. Prescription Medication: Prescription medication can only be advertised or promoted in technical or scientific publications, addressed exclusively to health professionals. Taking into account the nature of the new use, it is possible that new rules concerning advertisement are going to be set. The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.

1 0

We don't need any ChatGPT
by professor rat 06 Dec '22

06 Dec '22

So long as we have Semich, Garofalo and Gramps ' . . . To date, there’s been little evidence of the harmful effects of LLMs in the real world. But these recent events at Stack Overflow support the argument that the scale of these systems does indeed create new challenges. The site’s mods say as much in announcing the ban on ChatGPT, noting that the “volume of these [AI-generated] answers (thousands) and the fact that the answers often require a detailed read by someone with at least some subject matter expertise in order to determine that the answer is actually bad has effectively swamped our volunteer-based quality curation infrastructure.” The worry is that this pattern could be repeated on other platforms, with a flood of AI content drowning out the voices of real users with plausible but incorrect data. Exactly how this could play out in different domains around the web, though, would depend on the exact nature of the platform and its moderation capabilities. Whether or not these problems can be mitigated in the future using tools like improved spam filters remains to be seen." https://marginalrevolution.com/marginalrevolution/2022/12/twitter-comes-of-…

2 1

Must read for all Natural Law Nitwits
by professor rat 06 Dec '22

06 Dec '22

But the real law, the Common Law, lives still, for it is deathless; it is God, Nature, and Reason all rolled up in one. https://theanarchistlibrary.org/library/bob-black-constitutionalism-the-whi… Reposts not Nazi-style antisemitism

1 0

How the ooga-booga Gov-Corp conspiracy will wipe out cryptoanarchy
by professor rat 06 Dec '22

06 Dec '22

That AI chat-bot answered this question in broad brush strokes https://twitter.com/AltcoinPsycho/status/1598742302050357248/photo/1 When gaseous chatbots attack Attacking large exchanges simultaneously requires a lot of people and/or resources at the level of a nation-state. Gaseous chatbots currently don’t have access to either or both. Then there’s the potential for leaks, blowback and payback. When gaseous chatbots attack ( 2 ) Organizing govt regulation and punishment-through-process contradicts 1) to a greater or larger degree depending and also requires large investments, long lead times and increased resources for oppressive policing agencies. And as Goldfinger observed to Bond about his 008 successor, " I trust he will be more successful ". When gaseous chatbots attack ( 3 ) Using your influence and contacts to spread knowingly false information may be justified in some extreme circumstances. I suggest such circumstances don’t exist for crypto-currencies yet, without child-abuse factoring in. Trading in disinformation also risks reputation capital. This is hard won and easily lost. It also clashes with both 1) and 2) I see a list of tactics so far - no coherent strategy - and no provable logistics supply. Nonsense on stilts. When gaseous chatbots attack ( 4 ) Persuading large corporate entities to shun crypto sounds like herding cats. But knock yourself out. Maybe the IMF and USG will label us an existential threat - again. On what basis is this shunning even justified - where are the Four Horsemen of the Infopocalypse? When gaseous chatbots attack ( 5 ) When so many thousands of crypto-currencies now exist, slowing or even stopping any new ones coming in could actually do the sector a favor. Much as established coins are increasing their very public burning-off activities. Do your worst, gasbag - we’ll do our best.

2 1

Jason Bourne claims he doesn't remember making CRYPTO dotcom ad
by professor rat 06 Dec '22

06 Dec '22

PAM should have taken him out by now https://cointelegraph.com/news/us-consumer-watchdog-probes-crypto-firms-ove… Reposts not loose cannons in the Treadstone Program

1 0