I'm working on the final touches of a model comparison function. The beta version is just an early peak at a suite of mech interp applications I have in my backlog!
R PRO
juiceb0xc0de
AI & ML interests
Language models: Teaching pragmatic nuance to considerate linear algebra. Observing the process a machine uses to interpret encoded character data into machine language. If we learn to speak the language of predictions, do we become a society of oracles?
Recent Activity
updated a bucket about 1 hour ago
juiceb0xc0de/npz repliedto their post about 2 hours ago
You built a model. Wanna know what lights up inside it?💡
I put a little juice in the box so you can find out!🧃
I built Cartographics, a HF ZeroGPU Space for building neural activation atlases of language models.
I wanted to make the tools I use to explore model internals accessible to people building their own LLMs. Punch in a Hugging Face model ID, run the compatibility check, and start a capture. Prompt formatting and the full corpus are handled automatically.
The atlases help you investigate questions like:
- Which layers and channels respond most strongly to different kinds of prompts?
- How do activation patterns vary across code, math, reasoning, and other topics?
- Which features activate together, and where do unusually strong activations appear?
The census contains 8,965 prompts, giving you a shared basis for exploring different models.
GPU capture runs through ZeroGPU using your account’s allowance. I handle the CPU analysis separately. This release supports compatible architectures and models that fit the available GPU memory.
I’m building this independently, and I’d love to see what models builders explore with it.
What would you want to investigate in your own model?
Try Cartographics Beta:
https://huggingface.co/spaces/juiceb0xc0de/cartographics
https://huggingface.co/datasets/juiceb0xc0de/bella-bartender-gemma-e4b-cgs-atlas
If you make an atlas with us, leave the Space a ❤️. You're helping other people discover mechanistic interpretability the way I wish I did.
Prerequisites:
Models under 10B parameters (20GB of weights in bf16)
No trust_remote_code models, for now.
Preflight catches edge cases. Don't be scared to test it out. It's free to check.
Sign in with Hugging Face. The capture runs on your ZeroGPU quota, the atlas build runs on ours. reacted to danielhanchen's post with 🤯 about 8 hours ago
Run DeepSeek V4.1-Flash-1017 locally! 🐳
We made 1-bit retain ~88.2% accuracy while shrinking size from 508GB to 250GB (−52%).
Run on a 128GB+ RAM device. V4.1 Flash 1017 is the strongest open model to date
Guide: https://unsloth.ai/docs/models/deepseek-v4.1-flash
GGUF: https://huggingface.co/unsloth/DeepSeek-V4.1-Flash-1017-GGUFOrganizations
replied to their post about 2 hours ago
reacted to danielhanchen's post with 🤯 about 8 hours ago
Post
3,278
Run DeepSeek V4.1-Flash-1017 locally! 🐳
We made 1-bit retain ~88.2% accuracy while shrinking size from 508GB to 250GB (−52%).
Run on a 128GB+ RAM device. V4.1 Flash 1017 is the strongest open model to date
Guide: https://unsloth.ai/docs/models/deepseek-v4.1-flash
GGUF: unsloth/DeepSeek-V4.1-Flash-1017-GGUF
We made 1-bit retain ~88.2% accuracy while shrinking size from 508GB to 250GB (−52%).
Run on a 128GB+ RAM device. V4.1 Flash 1017 is the strongest open model to date
Guide: https://unsloth.ai/docs/models/deepseek-v4.1-flash
GGUF: unsloth/DeepSeek-V4.1-Flash-1017-GGUF
reacted to their post with 👀 about 9 hours ago
Post
894
You built a model. Wanna know what lights up inside it?💡
I put a little juice in the box so you can find out!🧃
I built Cartographics, a HF ZeroGPU Space for building neural activation atlases of language models.
I wanted to make the tools I use to explore model internals accessible to people building their own LLMs. Punch in a Hugging Face model ID, run the compatibility check, and start a capture. Prompt formatting and the full corpus are handled automatically.
The atlases help you investigate questions like:
- Which layers and channels respond most strongly to different kinds of prompts?
- How do activation patterns vary across code, math, reasoning, and other topics?
- Which features activate together, and where do unusually strong activations appear?
The census contains 8,965 prompts, giving you a shared basis for exploring different models.
GPU capture runs through ZeroGPU using your account’s allowance. I handle the CPU analysis separately. This release supports compatible architectures and models that fit the available GPU memory.
I’m building this independently, and I’d love to see what models builders explore with it.
What would you want to investigate in your own model?
Try Cartographics Beta:
juiceb0xc0de/cartographics
juiceb0xc0de/bella-bartender-gemma-e4b-cgs-atlas
If you make an atlas with us, leave the Space a ❤️. You're helping other people discover mechanistic interpretability the way I wish I did.
Prerequisites:
Models under 10B parameters (20GB of weights in bf16)
No trust_remote_code models, for now.
Preflight catches edge cases. Don't be scared to test it out. It's free to check.
Sign in with Hugging Face. The capture runs on your ZeroGPU quota, the atlas build runs on ours.
I put a little juice in the box so you can find out!🧃
I built Cartographics, a HF ZeroGPU Space for building neural activation atlases of language models.
I wanted to make the tools I use to explore model internals accessible to people building their own LLMs. Punch in a Hugging Face model ID, run the compatibility check, and start a capture. Prompt formatting and the full corpus are handled automatically.
The atlases help you investigate questions like:
- Which layers and channels respond most strongly to different kinds of prompts?
- How do activation patterns vary across code, math, reasoning, and other topics?
- Which features activate together, and where do unusually strong activations appear?
The census contains 8,965 prompts, giving you a shared basis for exploring different models.
GPU capture runs through ZeroGPU using your account’s allowance. I handle the CPU analysis separately. This release supports compatible architectures and models that fit the available GPU memory.
I’m building this independently, and I’d love to see what models builders explore with it.
What would you want to investigate in your own model?
Try Cartographics Beta:
juiceb0xc0de/cartographics
juiceb0xc0de/bella-bartender-gemma-e4b-cgs-atlas
If you make an atlas with us, leave the Space a ❤️. You're helping other people discover mechanistic interpretability the way I wish I did.
Prerequisites:
Models under 10B parameters (20GB of weights in bf16)
No trust_remote_code models, for now.
Preflight catches edge cases. Don't be scared to test it out. It's free to check.
Sign in with Hugging Face. The capture runs on your ZeroGPU quota, the atlas build runs on ours.
- 2 replies
posted an update about 9 hours ago
Post
894
You built a model. Wanna know what lights up inside it?💡
I put a little juice in the box so you can find out!🧃
I built Cartographics, a HF ZeroGPU Space for building neural activation atlases of language models.
I wanted to make the tools I use to explore model internals accessible to people building their own LLMs. Punch in a Hugging Face model ID, run the compatibility check, and start a capture. Prompt formatting and the full corpus are handled automatically.
The atlases help you investigate questions like:
- Which layers and channels respond most strongly to different kinds of prompts?
- How do activation patterns vary across code, math, reasoning, and other topics?
- Which features activate together, and where do unusually strong activations appear?
The census contains 8,965 prompts, giving you a shared basis for exploring different models.
GPU capture runs through ZeroGPU using your account’s allowance. I handle the CPU analysis separately. This release supports compatible architectures and models that fit the available GPU memory.
I’m building this independently, and I’d love to see what models builders explore with it.
What would you want to investigate in your own model?
Try Cartographics Beta:
juiceb0xc0de/cartographics
juiceb0xc0de/bella-bartender-gemma-e4b-cgs-atlas
If you make an atlas with us, leave the Space a ❤️. You're helping other people discover mechanistic interpretability the way I wish I did.
Prerequisites:
Models under 10B parameters (20GB of weights in bf16)
No trust_remote_code models, for now.
Preflight catches edge cases. Don't be scared to test it out. It's free to check.
Sign in with Hugging Face. The capture runs on your ZeroGPU quota, the atlas build runs on ours.
I put a little juice in the box so you can find out!🧃
I built Cartographics, a HF ZeroGPU Space for building neural activation atlases of language models.
I wanted to make the tools I use to explore model internals accessible to people building their own LLMs. Punch in a Hugging Face model ID, run the compatibility check, and start a capture. Prompt formatting and the full corpus are handled automatically.
The atlases help you investigate questions like:
- Which layers and channels respond most strongly to different kinds of prompts?
- How do activation patterns vary across code, math, reasoning, and other topics?
- Which features activate together, and where do unusually strong activations appear?
The census contains 8,965 prompts, giving you a shared basis for exploring different models.
GPU capture runs through ZeroGPU using your account’s allowance. I handle the CPU analysis separately. This release supports compatible architectures and models that fit the available GPU memory.
I’m building this independently, and I’d love to see what models builders explore with it.
What would you want to investigate in your own model?
Try Cartographics Beta:
juiceb0xc0de/cartographics
juiceb0xc0de/bella-bartender-gemma-e4b-cgs-atlas
If you make an atlas with us, leave the Space a ❤️. You're helping other people discover mechanistic interpretability the way I wish I did.
Prerequisites:
Models under 10B parameters (20GB of weights in bf16)
No trust_remote_code models, for now.
Preflight catches edge cases. Don't be scared to test it out. It's free to check.
Sign in with Hugging Face. The capture runs on your ZeroGPU quota, the atlas build runs on ours.
- 2 replies
reacted to danielhanchen's post with 🔥🤗👀🤯 24 days ago
reacted to FlameF0X's post with 🔥 29 days ago
Post
4,697
Personally, I don't think bot accounts on Hugging Face are a good thing, as we don't know how many accounts are run by automated systems versus how many actual users there are. Dead Internet theory is already a thing.
To clarify I am taking about LLM powered bot accounts and NOT rule base once like @parquet-converter or others.
also I'd like to talk with HUMANS not a machine so I'm going to hide messages from bots.
To clarify I am taking about LLM powered bot accounts and NOT rule base once like @parquet-converter or others.
also I'd like to talk with HUMANS not a machine so I'm going to hide messages from bots.
- 24 replies
Existentialism’s a bitch hey.
I think it's better described as an elite group of individuals who still consider literacy a key role in the development process.
reacted to onekq's post with 🔥 about 1 month ago
Post
4,493
I turned off subagents in Claude Code. Am I a minority?
- 5 replies
I thought about history as an array of secondary emotions all night at work so I had W&B's AI model render its version on the concept:
It's interesting watching what the leap from enjoying science fiction to being able to experience it in the comfort of your own home has truly done to society. A lot of the language and debate in the new is so relatable to the experience of humanity during the industrial revolution that it makes you wonder if this reaction is almost inherently part of our nature.
The periods of backlash to acceptance of any major technological advancement when we observe it from a macro scale must like look like a sinusoidal wave with predictive trends. The only thing that's changed over time is the rate of compression. We've witnessed things go from a slow morose waltz to the upbeat tempo of house music being played by a DJ who habitually increases the tempo over the intro to their set by orders of magnitude.
So have we become a society of luddites now? I guess members of Hugging face really fit the bill, "Democratizing AI". What a strange situation I feel like I'm in now that I reflect on it. I am the luddite trying to build better access to the thing that has gated access by working in collaboration with the object that is gated in the first place. AI models, the thoughtful looms of the future!
It's always a pleasure reading your work. It always reminds me to set aside some time for writing descriptive prose. This space needs more abstract thought and philosophical connections. We work with models that have nuanced personalities defining them amongst the market in unique ways. It's like we've all been measuring absorption rating for so long we forgot to step back and look at the painting that the math was always intended to create in the first place.
reacted to Quazim0t0's post with 🔥 about 1 month ago
Post
187
It feels like a sugar rush… this golden ratio they found. Stepping through a maze of processes, landing on what aims to be a generalist, raised from the cemetery of echoes of our thoughts, statements, findings, and conclusions. I never really thought beyond this point. I always assumed that the world would look as elegant as the innovations we create. I thought it would look different, have a defining design that really embodied the point we have now arrived at. I can't tell the difference. Everything looks the same. I think that's why it's even harder sometimes for us to swallow what is actually happening, because every representation that we've had from science fiction doesn't really fit the theme.
I don't really see any worry with text and language. These models can generalize them well, and we have all seen how profound it can be. What I do worry about more is when we leave the tiny world of language and replace it with the big world of experiences. What would it be like to generalize hundreds of lifetimes? Would it see the nuances better than I? Could it predict what comes next if it had enough data from observing and experiencing the world?...
Read more: https://huggingface.co/blog/Quazim0t0/little-world-big-world
I don't really see any worry with text and language. These models can generalize them well, and we have all seen how profound it can be. What I do worry about more is when we leave the tiny world of language and replace it with the big world of experiences. What would it be like to generalize hundreds of lifetimes? Would it see the nuances better than I? Could it predict what comes next if it had enough data from observing and experiencing the world?...
Read more: https://huggingface.co/blog/Quazim0t0/little-world-big-world
- 5 replies
posted an update about 1 month ago
Post
160
🍻Ayy! Think your model has the banter to take on the bartender? Put it to the test in my new space Bellarena! 🍻
1. Choose a .gguf model
2. Watch the showdown
Bella is behind the bar at FP16 precision, 8B parameters.
I curated a dataset of nothing but absurdist questions. Both models answer a randomly selected draw from 1 to 15, and our regular old house model, Ternary Bonsai 27B👴, judges whether the bartender or the patron brought more entertainment to the bar that evening.
I encourage you to please ask bella to contribute to the swear jar while she’s on shift.
juiceb0xc0de/bellarena
juiceb0xc0de/bella-chat
1. Choose a .gguf model
2. Watch the showdown
Bella is behind the bar at FP16 precision, 8B parameters.
I curated a dataset of nothing but absurdist questions. Both models answer a randomly selected draw from 1 to 15, and our regular old house model, Ternary Bonsai 27B👴, judges whether the bartender or the patron brought more entertainment to the bar that evening.
I encourage you to please ask bella to contribute to the swear jar while she’s on shift.
juiceb0xc0de/bellarena
juiceb0xc0de/bella-chat
reacted to danielhanchen's post with 🚀 about 1 month ago
Post
6,902
Qwen3.8-27B Unsloth GGUF is now the #1 most-liked GGUF of all time! 🤗🦥
The model hit 10M downloads and 3.7K likes in just 24 days on Hugging Face - all thanks to you.
GGUF: unsloth/Qwen3.8-27B-GGUF
Guide: https://unsloth.ai/docs/models/qwen3.8
The model hit 10M downloads and 3.7K likes in just 24 days on Hugging Face - all thanks to you.
GGUF: unsloth/Qwen3.8-27B-GGUF
Guide: https://unsloth.ai/docs/models/qwen3.8
- 5 replies
reacted to nwaughachukwuma's post with 🧠 about 1 month ago
Post
3,240
It’s easy to get distracted by benchmarks, throughput (tok/s), and all the hype around frontier model releases.
This is Shiny Model Syndrome, which makes engineers and teams forget the basic physics of production software, i.e., using the right tool for the job and optimizing for ease of integration.
- Teams spend huge amounts of money on frontier models for document parsing, OCR, detection, segmentation, and other task-specific visual AI workflows.
- Inference marketplaces don’t find it profitable to list task-specific models like glm-ocr, paddleocr, or dots.mocr, even though they’re all superior to frontier VLMs for document parsing and OCR.
- Engineers stitch together multiple endpoints for different use cases across the long tail of visual AI. Those who choose to self-host instead deal with painful infrastructure and GPU ops.
At VLM Run, we wanted one place to run OCR models, VLMs, and ViTs that we could confidently use for our own internal agents and evals. The gateway was born out of that need, and we’ve since opened it to the public.
The gateway exposes a single OpenAI-compatible endpoint for the long tail of visual AI across OCR, document parsing, VQA, detection, segmentation, embeddings, and transcription. Simply point the base_url of your OpenAI SDK at gateway.vlm.run/v1/openai, or ask your agent to connect via MCP (gateway.vlm.run/mcp).
You can swap the model name to compare glm-ocr, dots.mocr, paddleocr-vl-1.6, qwen3.8-27b, gemma4-26b-a4b, and more. We handle serving, runtime, and pipelining behind the scenes to give you high-quality visual intelligence.
- https://vlm.run/gateway
- https://huggingface.co/blog/vlm-run/introducing-gateway
- https://www.vlm.run/blog/introducing-gateway
This is Shiny Model Syndrome, which makes engineers and teams forget the basic physics of production software, i.e., using the right tool for the job and optimizing for ease of integration.
- Teams spend huge amounts of money on frontier models for document parsing, OCR, detection, segmentation, and other task-specific visual AI workflows.
- Inference marketplaces don’t find it profitable to list task-specific models like glm-ocr, paddleocr, or dots.mocr, even though they’re all superior to frontier VLMs for document parsing and OCR.
- Engineers stitch together multiple endpoints for different use cases across the long tail of visual AI. Those who choose to self-host instead deal with painful infrastructure and GPU ops.
At VLM Run, we wanted one place to run OCR models, VLMs, and ViTs that we could confidently use for our own internal agents and evals. The gateway was born out of that need, and we’ve since opened it to the public.
The gateway exposes a single OpenAI-compatible endpoint for the long tail of visual AI across OCR, document parsing, VQA, detection, segmentation, embeddings, and transcription. Simply point the base_url of your OpenAI SDK at gateway.vlm.run/v1/openai, or ask your agent to connect via MCP (gateway.vlm.run/mcp).
You can swap the model name to compare glm-ocr, dots.mocr, paddleocr-vl-1.6, qwen3.8-27b, gemma4-26b-a4b, and more. We handle serving, runtime, and pipelining behind the scenes to give you high-quality visual intelligence.
- https://vlm.run/gateway
- https://huggingface.co/blog/vlm-run/introducing-gateway
- https://www.vlm.run/blog/introducing-gateway
reacted to their post with 🔥 about 1 month ago
Post
3,050
Dude, Where's My Update? I'll tell you where! ~97.6% of my BF16 parameter coordinates didn't move at all, and the ones that did overshot by ~1.33x.
It's nice to do research that doesn't end in disproving yourself once again and moving on to the next subject once in awhile.
Back to the topic, if you've ever wondered why most of your weights are basically ghosting you nearly every step when you store your weights at bf16, Dude, I Measured It.
https://huggingface.co/blog/juiceb0xc0de/intended-and-realized-updates-in-bf16-fine-tuning#dude-wheres-my-update
It's nice to do research that doesn't end in disproving yourself once again and moving on to the next subject once in awhile.
Back to the topic, if you've ever wondered why most of your weights are basically ghosting you nearly every step when you store your weights at bf16, Dude, I Measured It.
https://huggingface.co/blog/juiceb0xc0de/intended-and-realized-updates-in-bf16-fine-tuning#dude-wheres-my-update