The fastest way to fork and deploy open-source AI.

Just discovered @runpod_io 🤯🤯🤯 Per second billing for serverless GPU capacity?! Infinitely scalable?! Whaaaat

ai engineering is just tweaking config values in a notebook until you run out of runpod credits

Needed a GPU for a quick job, didn't want to commit to anything long-term. Runpod was perfect for that. Love that I can just spin one up and shut it down after.

Trained a 7B parameter model in just 90 minutes for $0.80 using LoRA + Runpod. Yes, it's possible, and no, you don't need enterprise hardware.

I love runpod

@runpod_io is so goated, first time trying it today and it’s super easy to setup + their ai helper on discord was very helpful If you ever need cpus/gpus I recommend it!

For anyone annoyed with Amazon's (and Azure's and Google's) gatekeeping on their cloud GPU VMs, I recommend @runpod_io None of the 'prove you really need this much power' bs from the majors Just great pricing, availability, and an intuitive UI

Shoutout to @runpod_io as I work through my first non-trivial machine learning experiment. They have exactly what you need if you're a hobbyist and their prices are about a fifth of the big cloud providers.

The @runpod_io event was amazing! One reason we can boast about fast speeds at @jigsawstack is because the cold boot on runpod GPUs is basically nonexistent!

Thanks @runpod_io, loving all of the updates! 👀

Why is Huggingface not adding Runpod as a serverless provider? Runpod is 10-15x cheaper for serverless deployment than AWS and GCP

Apparently, we got a Kaggle silver medal in the @arcprize for being in position 17th out of 1430 teams 🙃 I wish I had more time to spend on it; we worked on it for a couple of weeks for fun with limited compute (HUGE thanks to @runpod_io!)

Runpod has great prices as well

switching from runpod+cursor to colab made my ml pipeline setup 100x faster, but i'm missing access to claude sonnet 3.7 for interpreting repo files & terminal outputs. need a way to either: 1. connect cursor to colab so the llm sees my terminal outputs 2. efficiently feed colab outputs to an llm without tedious copy/paste what's the best workflow for llm-assisted debugging when your code runs in colab but your preferred assistant lacks context? any MCP servers to give cursor access to colab?

I have been testing work with @runpod_io last 2 weeks and I've to say the service is pretty amazing. Super awesome UX and DevEX (and plenty of GPU backend choices). It's about ~20% pricier than Lambda labs, but worth it IMO given all the harness and workflow they provide that Lambda doesn't. I'm not associated with them in any way or manner, btw. Just a very happy customer.

thank u runpod i was doing a training run for work when GCP and cloudflare died 🙏🙏 i appreciate u staying online it finished successfully

Runpod > Sagemaker, VertexAi, AzureML

Runpod support > lambdalabs support. For on demand GPUs runpod still works the best ime

I'm proud to be part of the GPU Elite, awarded by @runpod_io 😍

i'm a big fan of @runpod_io they're most reliable GPU provider for training and running your models at scale

1.3k spent on the training run, this latest release would not have been possible without runpod

🏃 Runpod is now available on SkyPilot! ✈️ Get high-end GPUs (3x cheaper) with great availability: sky launch --gpus H100 Great thanks to @runpod_io for contributing this integration to join the Sky!

Axolotl works out of the box with @runpod_io's Clusters. It's as easy as running this on each node using the Docker images that we ship.

Introducing SmolLlama! An effort to make a mini-ChatGPT from scratch! Its based on the Llama (123 M) structure I coded and pre-trained on 10B tokens (10k steps) from the FineWeb dataset from scratch using DDP (torchrun) in PyTorch. Used 2xH100 (SXM) 80GB VRAM from Runpod

thank u runpod i was doing a training run for work when GCP and cloudflare died 🙏🙏 i appreciate u staying online it finished successfully

Trained a 7B parameter model in just 90 minutes for $0.80 using LoRA + Runpod. Yes, it's possible, and no, you don't need enterprise hardware.

Just discovered @runpod_io 🤯🤯🤯 Per second billing for serverless GPU capacity?! Infinitely scalable?! Whaaaat

For anyone annoyed with Amazon's (and Azure's and Google's) gatekeeping on their cloud GPU VMs, I recommend @runpod_io None of the 'prove you really need this much power' bs from the majors Just great pricing, availability, and an intuitive UI

Introducing SmolLlama! An effort to make a mini-ChatGPT from scratch! Its based on the Llama (123 M) structure I coded and pre-trained on 10B tokens (10k steps) from the FineWeb dataset from scratch using DDP (torchrun) in PyTorch. Used 2xH100 (SXM) 80GB VRAM from Runpod

switching from runpod+cursor to colab made my ml pipeline setup 100x faster, but i'm missing access to claude sonnet 3.7 for interpreting repo files & terminal outputs. need a way to either: 1. connect cursor to colab so the llm sees my terminal outputs 2. efficiently feed colab outputs to an llm without tedious copy/paste what's the best workflow for llm-assisted debugging when your code runs in colab but your preferred assistant lacks context? any MCP servers to give cursor access to colab?

Runpod support > lambdalabs support. For on demand GPUs runpod still works the best ime

i'm a big fan of @runpod_io they're most reliable GPU provider for training and running your models at scale

1.3k spent on the training run, this latest release would not have been possible without runpod

I love runpod

Thanks @runpod_io, loving all of the updates! 👀

Apparently, we got a Kaggle silver medal in the @arcprize for being in position 17th out of 1430 teams 🙃 I wish I had more time to spend on it; we worked on it for a couple of weeks for fun with limited compute (HUGE thanks to @runpod_io!)

Shoutout to @runpod_io as I work through my first non-trivial machine learning experiment. They have exactly what you need if you're a hobbyist and their prices are about a fifth of the big cloud providers.

Runpod > Sagemaker, VertexAi, AzureML

The @runpod_io event was amazing! One reason we can boast about fast speeds at @jigsawstack is because the cold boot on runpod GPUs is basically nonexistent!

I have been testing work with @runpod_io last 2 weeks and I've to say the service is pretty amazing. Super awesome UX and DevEX (and plenty of GPU backend choices). It's about ~20% pricier than Lambda labs, but worth it IMO given all the harness and workflow they provide that Lambda doesn't. I'm not associated with them in any way or manner, btw. Just a very happy customer.

I'm proud to be part of the GPU Elite, awarded by @runpod_io 😍

Why is Huggingface not adding Runpod as a serverless provider? Runpod is 10-15x cheaper for serverless deployment than AWS and GCP

@runpod_io is so goated, first time trying it today and it’s super easy to setup + their ai helper on discord was very helpful If you ever need cpus/gpus I recommend it!

ai engineering is just tweaking config values in a notebook until you run out of runpod credits

Axolotl works out of the box with @runpod_io's Clusters. It's as easy as running this on each node using the Docker images that we ship.

🏃 Runpod is now available on SkyPilot! ✈️ Get high-end GPUs (3x cheaper) with great availability: sky launch --gpus H100 Great thanks to @runpod_io for contributing this integration to join the Sky!

Runpod has great prices as well

Needed a GPU for a quick job, didn't want to commit to anything long-term. Runpod was perfect for that. Love that I can just spin one up and shut it down after.

The @runpod_io event was amazing! One reason we can boast about fast speeds at @jigsawstack is because the cold boot on runpod GPUs is basically nonexistent!

Shoutout to @runpod_io as I work through my first non-trivial machine learning experiment. They have exactly what you need if you're a hobbyist and their prices are about a fifth of the big cloud providers.

Just discovered @runpod_io 🤯🤯🤯 Per second billing for serverless GPU capacity?! Infinitely scalable?! Whaaaat

Introducing SmolLlama! An effort to make a mini-ChatGPT from scratch! Its based on the Llama (123 M) structure I coded and pre-trained on 10B tokens (10k steps) from the FineWeb dataset from scratch using DDP (torchrun) in PyTorch. Used 2xH100 (SXM) 80GB VRAM from Runpod

Runpod > Sagemaker, VertexAi, AzureML

Trained a 7B parameter model in just 90 minutes for $0.80 using LoRA + Runpod. Yes, it's possible, and no, you don't need enterprise hardware.

I love runpod

thank u runpod i was doing a training run for work when GCP and cloudflare died 🙏🙏 i appreciate u staying online it finished successfully

Runpod support > lambdalabs support. For on demand GPUs runpod still works the best ime

1.3k spent on the training run, this latest release would not have been possible without runpod

I'm proud to be part of the GPU Elite, awarded by @runpod_io 😍

Runpod has great prices as well

@runpod_io is so goated, first time trying it today and it’s super easy to setup + their ai helper on discord was very helpful If you ever need cpus/gpus I recommend it!

Needed a GPU for a quick job, didn't want to commit to anything long-term. Runpod was perfect for that. Love that I can just spin one up and shut it down after.

i'm a big fan of @runpod_io they're most reliable GPU provider for training and running your models at scale

switching from runpod+cursor to colab made my ml pipeline setup 100x faster, but i'm missing access to claude sonnet 3.7 for interpreting repo files & terminal outputs. need a way to either: 1. connect cursor to colab so the llm sees my terminal outputs 2. efficiently feed colab outputs to an llm without tedious copy/paste what's the best workflow for llm-assisted debugging when your code runs in colab but your preferred assistant lacks context? any MCP servers to give cursor access to colab?

Thanks @runpod_io, loving all of the updates! 👀

Axolotl works out of the box with @runpod_io's Clusters. It's as easy as running this on each node using the Docker images that we ship.

For anyone annoyed with Amazon's (and Azure's and Google's) gatekeeping on their cloud GPU VMs, I recommend @runpod_io None of the 'prove you really need this much power' bs from the majors Just great pricing, availability, and an intuitive UI

Apparently, we got a Kaggle silver medal in the @arcprize for being in position 17th out of 1430 teams 🙃 I wish I had more time to spend on it; we worked on it for a couple of weeks for fun with limited compute (HUGE thanks to @runpod_io!)

Why is Huggingface not adding Runpod as a serverless provider? Runpod is 10-15x cheaper for serverless deployment than AWS and GCP

🏃 Runpod is now available on SkyPilot! ✈️ Get high-end GPUs (3x cheaper) with great availability: sky launch --gpus H100 Great thanks to @runpod_io for contributing this integration to join the Sky!

I have been testing work with @runpod_io last 2 weeks and I've to say the service is pretty amazing. Super awesome UX and DevEX (and plenty of GPU backend choices). It's about ~20% pricier than Lambda labs, but worth it IMO given all the harness and workflow they provide that Lambda doesn't. I'm not associated with them in any way or manner, btw. Just a very happy customer.

ai engineering is just tweaking config values in a notebook until you run out of runpod credits

i'm a big fan of @runpod_io they're most reliable GPU provider for training and running your models at scale

Runpod support > lambdalabs support. For on demand GPUs runpod still works the best ime

Runpod > Sagemaker, VertexAi, AzureML

switching from runpod+cursor to colab made my ml pipeline setup 100x faster, but i'm missing access to claude sonnet 3.7 for interpreting repo files & terminal outputs. need a way to either: 1. connect cursor to colab so the llm sees my terminal outputs 2. efficiently feed colab outputs to an llm without tedious copy/paste what's the best workflow for llm-assisted debugging when your code runs in colab but your preferred assistant lacks context? any MCP servers to give cursor access to colab?

Trained a 7B parameter model in just 90 minutes for $0.80 using LoRA + Runpod. Yes, it's possible, and no, you don't need enterprise hardware.

For anyone annoyed with Amazon's (and Azure's and Google's) gatekeeping on their cloud GPU VMs, I recommend @runpod_io None of the 'prove you really need this much power' bs from the majors Just great pricing, availability, and an intuitive UI

I have been testing work with @runpod_io last 2 weeks and I've to say the service is pretty amazing. Super awesome UX and DevEX (and plenty of GPU backend choices). It's about ~20% pricier than Lambda labs, but worth it IMO given all the harness and workflow they provide that Lambda doesn't. I'm not associated with them in any way or manner, btw. Just a very happy customer.

I love runpod

Just discovered @runpod_io 🤯🤯🤯 Per second billing for serverless GPU capacity?! Infinitely scalable?! Whaaaat

Needed a GPU for a quick job, didn't want to commit to anything long-term. Runpod was perfect for that. Love that I can just spin one up and shut it down after.

1.3k spent on the training run, this latest release would not have been possible without runpod

thank u runpod i was doing a training run for work when GCP and cloudflare died 🙏🙏 i appreciate u staying online it finished successfully

The @runpod_io event was amazing! One reason we can boast about fast speeds at @jigsawstack is because the cold boot on runpod GPUs is basically nonexistent!

I'm proud to be part of the GPU Elite, awarded by @runpod_io 😍

@runpod_io is so goated, first time trying it today and it’s super easy to setup + their ai helper on discord was very helpful If you ever need cpus/gpus I recommend it!

Axolotl works out of the box with @runpod_io's Clusters. It's as easy as running this on each node using the Docker images that we ship.

Shoutout to @runpod_io as I work through my first non-trivial machine learning experiment. They have exactly what you need if you're a hobbyist and their prices are about a fifth of the big cloud providers.

Introducing SmolLlama! An effort to make a mini-ChatGPT from scratch! Its based on the Llama (123 M) structure I coded and pre-trained on 10B tokens (10k steps) from the FineWeb dataset from scratch using DDP (torchrun) in PyTorch. Used 2xH100 (SXM) 80GB VRAM from Runpod

Runpod has great prices as well

🏃 Runpod is now available on SkyPilot! ✈️ Get high-end GPUs (3x cheaper) with great availability: sky launch --gpus H100 Great thanks to @runpod_io for contributing this integration to join the Sky!

Apparently, we got a Kaggle silver medal in the @arcprize for being in position 17th out of 1430 teams 🙃 I wish I had more time to spend on it; we worked on it for a couple of weeks for fun with limited compute (HUGE thanks to @runpod_io!)

Why is Huggingface not adding Runpod as a serverless provider? Runpod is 10-15x cheaper for serverless deployment than AWS and GCP

ai engineering is just tweaking config values in a notebook until you run out of runpod credits

Thanks @runpod_io, loving all of the updates! 👀