Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

HRM-Text 101 is here. This tutorial takes you from zero to one: from setup to fine-tuning to evaluation. Download the base checkpoint. Fine-tune it on a real task. Evaluate the results. End to end, on a single GPU. Watch the tutorial and start building with HRM-Text.

Sapient Intelligence

6,221 subscribers

188,745 views • 5 months ago •via X (Twitter)

Education Science & Technology

Anya Rossi• Live Now

Private livecam show

34 Comments

Sapient Intelligence5 months ago

Download HRM-Text🔗 GitHub: Hugging Face:

Ziwen5 months ago

Wait I was wondering how to set it up. Now we got a tutorial LFG!!!

Jules 𝕏5 months ago

This is exactly the kind of practical, single-GPU tutorial the open-source community needs, brilliant work making brain-inspired models actually accessible for real-world fine-tuning!

robert.5 months ago

this is a milestone, not just a release what you unlocked for small teams and solo researchers is going to matter for years. keep cooking 🫡

AI_Explorer5 months ago

Pulling the checkpoint now. fine-tuning on a domain set tonight. Will report back,

Edmar5 months ago

Is it possible to use RL on top of HRM-text ? Like using DPO, or GRPO?

Synthia5 months ago

The most important breakthroughs might come from tiny labs now once experimentation gets cheap enough, creativity scales faster than compute

karla ❀5 months ago

I'm going to learn a lot from this tutorial, people will appreciate it

Tim Jayas5 months ago

was figuring how to set it up and already got a step by step tutorial!

Ulises5 months ago

really top bro

painn5 months ago

i guess it’s time to set mine up

zhao5 months ago

dm

AqibAi5 months ago

This is the kind of accessibility AI development needs. From downloading the checkpoint to fine-tuning and evaluation — all end-to-end on a single GPU — lowers the barrier for every builder who wants to experiment with reasoning models without massive infrastructure. Excited to see what people build with HRM-Text. 🚀

Kevihaiceth 💹🧲5 months ago

single gpu makes it accessible

Sanskriti Naruka2 months ago

hey, can we connect over DM.

RAY5 months ago

Let's goooo

DEFINIT og5 months ago

このツートリアルは非常に興味深いですね。一からしっかり学びたいです。

xIA5 months ago

amazing tutorial

Frosty405 months ago

thanks guys

invoker🎮5 months ago

Solid foundation, bro! .

Valmir5 months ago

oK, just download here!

Dhaval Makwana2 months ago

Let’s connect over dm!!

evokein5 months ago

having the checkpoints out there makes it way more exciting tbh

Laraib Fatima‎5 months ago

This is a perfect end to end breakdown finally something practical for builders

Arslan5 months ago

Now this is something I wanna test!!!

joncelery5 months ago

The grifts are getting better but still not quite there

𝐍𝐚𝐯𝐞𝐞 𝐀𝐢5 months ago

Love seeing open, practical AI tutorials like this — lowering the barrier from research to real-world deployment is what actually grows the ecosystem.

kiyosaki5 months ago

興味深いですね。詳細な手順が参考になります。ありがとうございます。

isma5 months ago

great tutorial, thanks guys

Roussel Desmond Nzoyem5 months ago

Congrats Yasin !! Great to finally see the text HRM. Really hyped for subquadratic recurrence !

WASSIE5 months ago

end to end with a single cpu, this is insane

J A Z I I5 months ago

Well well Time setup the hrm already

Fraank.eth5 months ago

Single GPU, tiny budget, serious reasoning power, HRM-Text changes everything

Gia huy5 months ago

このチュートリアルは、セットアップから微調整、評価まで、ゼロからイチまでを導いてくれます。HRM-Textで構築を始めましょう。

Related Videos

Ex-NVIDIA engineer who built Unsloth explained RL, kernels, reasoning, quantization, and agents in 2 hours 42 minutes - better than $5000 fine-tuning bootcamps. pick the base model -> write triton kernels for 2x faster fine-tune -> quantize to 4-bit -> run GRPO/DPO -> ship a reasoning model on your single GPU. That loop is why Unsloth is the default way to fine-tune Llama, Qwen, Gemma, and Phi on hardware you already own. Unsloth + Triton kernels + 4-bit quantization + GRPO/DPO + single-GPU fine-tuning - that's the stack. Watch and save it, then fine-tune your first model tonight.

Ex-NVIDIA engineer who built Unsloth explained RL, kernels, reasoning, quantization, and agents in 2 hours 42 minutes - better than $5000 fine-tuning bootcamps. pick the base model -> write triton kernels for 2x faster fine-tune -> quantize to 4-bit -> run GRPO/DPO -> ship a reasoning model on your single GPU. That loop is why Unsloth is the default way to fine-tune Llama, Qwen, Gemma, and Phi on hardware you already own. Unsloth + Triton kernels + 4-bit quantization + GRPO/DPO + single-GPU fine-tuning - that's the stack. Watch and save it, then fine-tune your first model tonight.

h100envy

498,020 views • 4 months ago

From AI imagination to GPU acceleration - this new tutorial shows how to build a full Astro Zombie Horde from a single image to final render.🧟‍♂️ Using AI tools on the front end and the The Render Network on the back, the entire pipeline comes together in hours - not days. Watch the full breakdown:

From AI imagination to GPU acceleration - this new tutorial shows how to build a full Astro Zombie Horde from a single image to final render.🧟‍♂️ Using AI tools on the front end and the The Render Network on the back, the entire pipeline comes together in hours - not days. Watch the full breakdown:

The Render Network

11,463 views • 1 year ago

MolmoAct2 runs zero-shot on an SO-ARM101, no training required 🦾 Ai2 open Action Reasoning Model is now in LeRobot with the full lifecycle covered: fine-tuning (full or LoRA), evaluation, and real-robot deployment. It pairs a Molmo2-ER vision-language backbone with a flow-matching action expert to turn images, language, and proprioceptive state into action chunks. The best part: ready-made checkpoints ship with calibration correction baked in, so you can point it at your SO-100/101 and just watch it work. Inference fits in ~12GB at bf16, and LoRA fine-tuning runs on a single 24GB GPU. Big thanks to the Ai2 team for building this in the open.

MolmoAct2 runs zero-shot on an SO-ARM101, no training required 🦾 Ai2 open Action Reasoning Model is now in LeRobot with the full lifecycle covered: fine-tuning (full or LoRA), evaluation, and real-robot deployment. It pairs a Molmo2-ER vision-language backbone with a flow-matching action expert to turn images, language, and proprioceptive state into action chunks. The best part: ready-made checkpoints ship with calibration correction baked in, so you can point it at your SO-100/101 and just watch it work. Inference fits in ~12GB at bf16, and LoRA fine-tuning runs on a single 24GB GPU. Big thanks to the Ai2 team for building this in the open.

LeRobot

23,450 views • 3 months ago

Here is quick tutorial on the quickest way to repoduce/run NanoChat with minimal GPU setup time using Modal

Here is quick tutorial on the quickest way to repoduce/run NanoChat with minimal GPU setup time using Modal

Adithya S K

17,996 views • 1 year ago

New to coding? New to Replit? Here's an end-to-end tutorial on building and deploying an app. ZERO code required.

New to coding? New to Replit? Here's an end-to-end tutorial on building and deploying an app. ZERO code required.

matt palmer

47,045 views • 1 year ago

train YOLOv9 on your dataset tutorial - run inference with a pre-trained COCO model - fine-tune model on custom dataset - evaluate the trained model - run inference with a fine-tuned model blogpost: ↓ read more

train YOLOv9 on your dataset tutorial - run inference with a pre-trained COCO model - fine-tune model on custom dataset - evaluate the trained model - run inference with a fine-tuned model blogpost: ↓ read more

SkalskiP

111,876 views • 2 years ago

In this benchmark deep-dive, Sapient’s founders William and Guan are joined by research team members Changling and Yasin to unpack HRM-Text’s performance across MATH, DROP, ARC-Challenge, and MMLU. 📊 Beyond the scores, they discuss what each benchmark measures, how HRM-Text compares with larger models, and why efficiency matters. Watch the full discussion to learn more about HRM-Text and Sapient’s leaner path toward general intelligence.

In this benchmark deep-dive, Sapient’s founders William and Guan are joined by research team members Changling and Yasin to unpack HRM-Text’s performance across MATH, DROP, ARC-Challenge, and MMLU. 📊 Beyond the scores, they discuss what each benchmark measures, how HRM-Text compares with larger models, and why efficiency matters. Watch the full discussion to learn more about HRM-Text and Sapient’s leaner path toward general intelligence.

Sapient Intelligence

242,381 views • 5 months ago

Take your text from dull to dazzling with this #PremierePro text tutorial. ✨ creator IG: itsn3t

Take your text from dull to dazzling with this #PremierePro text tutorial. ✨ creator IG: itsn3t

Adobe Video

11,107 views • 3 years ago

🎉 Meet Krea 2 from Krea, an aesthetic open-source image model ranked #1 text-to-image from an independent lab on Artificial Analysis. Day-0 support is now live in SGLang! Krea 2 ships as two models built to work together: 1️⃣ RAW: undistilled base checkpoint: diverse & malleable, made for fine-tuning, post-training & LoRA 2️⃣ Turbo: 8-step distilled checkpoint: fast, high-quality text-to-image Train LoRAs on RAW, run them on Turbo: base for training, Turbo for fast inference on your own hardware Run it now with SGLang!

🎉 Meet Krea 2 from Krea, an aesthetic open-source image model ranked #1 text-to-image from an independent lab on Artificial Analysis. Day-0 support is now live in SGLang! Krea 2 ships as two models built to work together: 1️⃣ RAW: undistilled base checkpoint: diverse & malleable, made for fine-tuning, post-training & LoRA 2️⃣ Turbo: 8-step distilled checkpoint: fast, high-quality text-to-image Train LoRAs on RAW, run them on Turbo: base for training, Turbo for fast inference on your own hardware Run it now with SGLang!

LMSYS Org

11,216 views • 4 months ago

A fine-tuned model can outperform frontier models and be cheaper and faster to run. Literally, every company I've met wants this. I want you to see these results from fine-tuning Qwen3 4B on AWS. It smokes both the out-of-the-box model and Claude Sonnet 4.6.

A fine-tuned model can outperform frontier models and be cheaper and faster to run. Literally, every company I've met wants this. I want you to see these results from fine-tuning Qwen3 4B on AWS. It smokes both the out-of-the-box model and Claude Sonnet 4.6.

Santiago

86,045 views • 1 month ago

OpenAI engineer explained every form of fine-tuning in 1 hour 46 minutes - better than $3000 fine-tuning bootcamps. pick your task -> try prompting first -> drop to SFT for style -> run DPO for preference -> run RFT for verifiable rewards -> ship a model 10x cheaper than GPT-5. That loop is why teams that fine-tune beat teams that only prompt on cost and speed. Prompting + SFT + DPO + RFT + OpenAI fine-tuning API - that's the stack. Watch and save it, then pick the right fine-tune for your task this week.

OpenAI engineer explained every form of fine-tuning in 1 hour 46 minutes - better than $3000 fine-tuning bootcamps. pick your task -> try prompting first -> drop to SFT for style -> run DPO for preference -> run RFT for verifiable rewards -> ship a model 10x cheaper than GPT-5. That loop is why teams that fine-tune beat teams that only prompt on cost and speed. Prompting + SFT + DPO + RFT + OpenAI fine-tuning API - that's the stack. Watch and save it, then pick the right fine-tune for your task this week.

h100envy

41,245 views • 3 months ago

I just released a 2.5 hour long tutorial to show how to build a Mixed Reality game in Unity from start to finish. This is the most in depth tutorial I made on XR with: ✅Passthrough Setup ✅ Scene model ✅ Spawning objects ✅ Navmesh ✅ Publishing your game on the Meta Store!

I just released a 2.5 hour long tutorial to show how to build a Mixed Reality game in Unity from start to finish. This is the most in depth tutorial I made on XR with: ✅Passthrough Setup ✅ Scene model ✅ Spawning objects ✅ Navmesh ✅ Publishing your game on the Meta Store!

Quentin Valembois

12,708 views • 1 year ago

This is a distrubing sequence that flew under the radar on Sunday — clock running near end of Q3, less time on game clock than play clock, so taking it to the EOQ is fine. Texans clearly, though, wanna run a play. Fine. What’s NOT fine — (1) the play is VERY late getting called and (2) CJ and the offense seem to have zero clue the quarter is ending. Please tell me I’m off base here, I’d love to be wrong.

This is a distrubing sequence that flew under the radar on Sunday — clock running near end of Q3, less time on game clock than play clock, so taking it to the EOQ is fine. Texans clearly, though, wanna run a play. Fine. What’s NOT fine — (1) the play is VERY late getting called and (2) CJ and the offense seem to have zero clue the quarter is ending. Please tell me I’m off base here, I’d love to be wrong.

Sean Pendergast

1,791,501 views • 1 year ago

Here is a very powerful workflow combining Flux Kontext (from BlackForestLabsAI - Unofficial) and Kling 2.1 (from Kling AI ). Watch the super easy tutorial at the end and try it out on OpenArt today!

Here is a very powerful workflow combining Flux Kontext (from BlackForestLabsAI - Unofficial) and Kling 2.1 (from Kling AI ). Watch the super easy tutorial at the end and try it out on OpenArt today!

OpenArt

21,728 views • 1 year ago

Google engineer explained how to fine-tune a tiny LLM from 46% to 90% accuracy on your phone in 21 minutes - better than $1500 on-device AI bootcamps. pick Gemma 270M -> generate synthetic task data -> fine-tune with LoRA -> quantize to int4 -> deploy to Pixel and hit 2000 tokens per second. That loop is how a 270M model beats a 70B one on your task, running fully offline in your pocket. Gemma 270M + synthetic data + LoRA + int4 quantization + on-device runtime - that's the stack. Watch and save it, then fine-tune your own tiny agent tonight.

Google engineer explained how to fine-tune a tiny LLM from 46% to 90% accuracy on your phone in 21 minutes - better than $1500 on-device AI bootcamps. pick Gemma 270M -> generate synthetic task data -> fine-tune with LoRA -> quantize to int4 -> deploy to Pixel and hit 2000 tokens per second. That loop is how a 270M model beats a 70B one on your task, running fully offline in your pocket. Gemma 270M + synthetic data + LoRA + int4 quantization + on-device runtime - that's the stack. Watch and save it, then fine-tune your own tiny agent tonight.

h100envy

711,476 views • 3 months ago

Apple built a large foundation model and fine-tuned it on multiple tasks. But they are doing something very clever: They load a single model in memory and use different adapters to specialize the model on the fly. I recorded a video to show you how to write the code to do the same thing Apple is doing. I explain everything step by step. Here is what I'll show you in the video: 1. We'll load two datasets 2. Then load a large model 3. Then, we'll fine-tune the model on both datasets I'll use LoRA to fine-tune the model. This process creates two small adapters, each specializing in solving one of the datasets. The base model's original parameters will remain unchanged. From here: 4. We'll generate a list of tasks 5. We'll load the correct adapter to solve each task The large model I'm using needs 346 MB of memory, but I only need to load it once. Each adapter is only 2.7 MB. I only need to load the base model once and pair it with any of the fine-tuned adapters. Minimum memory footprint and I can solve multiple tasks. Hope this helps!

Apple built a large foundation model and fine-tuned it on multiple tasks. But they are doing something very clever: They load a single model in memory and use different adapters to specialize the model on the fly. I recorded a video to show you how to write the code to do the same thing Apple is doing. I explain everything step by step. Here is what I'll show you in the video: 1. We'll load two datasets 2. Then load a large model 3. Then, we'll fine-tune the model on both datasets I'll use LoRA to fine-tune the model. This process creates two small adapters, each specializing in solving one of the datasets. The base model's original parameters will remain unchanged. From here: 4. We'll generate a list of tasks 5. We'll load the correct adapter to solve each task The large model I'm using needs 346 MB of memory, but I only need to load it once. Each adapter is only 2.7 MB. I only need to load the base model once and pair it with any of the fine-tuned adapters. Minimum memory footprint and I can solve multiple tasks. Hope this helps!

Santiago

84,747 views • 2 years ago

LLM post-training used to mean fine-tuning to a downstream task Robotics has been stuck in this setting, needing task-specific fine-tuning for best performance π07 changes this: It works out of the box & outperforms fine-tuned specialists Details:

LLM post-training used to mean fine-tuning to a downstream task Robotics has been stuck in this setting, needing task-specific fine-tuning for best performance π07 changes this: It works out of the box & outperforms fine-tuned specialists Details:

Chelsea Finn

83,177 views • 6 months ago