Stable Diffusion v1.5
๐จ
2
Text-to-image with Stable Diffusion v1.5
Doron Adler PRO
Norod78
AI & ML interests
Fooling around with Generative machine learning models.
Recent Activity
liked a Space 7 days ago
mediasynthesismuseum/stable-diffusion-v1-5 liked a Space 7 days ago
mediasynthesismuseum/vqgan-clip updated a Space 11 days ago
Norod78/Hebrew-FakypediaOrganizations
liked 2
Spaces 7 days ago
updated 2
Spaces 11 days ago
updated a
Space 13 days ago
liked a
model 13 days ago
reacted to eabdullin's post with ๐ฅ๐ค about 1 month ago
Post
5730
Iโm doing a PhD in AI, which sounds impressive until you realize it mostly means I spend three years trying to make a computer say something slightly less stupid than it said yesterday.
People hear "AI researcher" and they think Iโm building the future. No. Iโm in a basement at 2 a.m. Googling, "CUDA error what the f**k does this mean."
And the worst part about AI research now is compute. You donโt even ask, "Is this idea good?" anymore. You ask, "Can I afford for this idea to be wrong?"
My advisor comes to me one day and says, "I think we should fine-tune our own language model."
I said, "Professor, with what money? Iโm a PhD student. I have two bank accounts: checking and emotionally checking."
He goes, "Donโt worry. We have compute."
Now, in academia, "donโt worry" is never the beginning of a good sentence.
I said, "What do you mean we have compute?"
He said, "My friend knows the cluster admin. He can get us on the GPUs."
I said, "Okayโฆ what do we have to do?"
He goes, "Nothing crazy. Just be very grateful in the acknowledgements."
I said, "How grateful?"
He said, "Maybe put him as co-author."
I said, "Co-author? Are we using the cluster, or is the cluster using us?"
Because at that point, thatโs not a favor. Thatโs academic child support.
So I go to the server room, and the cluster admin walks up to me and goes, "So youโre the NLP student."
And in my head Iโm like, "No, tonight youโre the principal investigator. Youโre the provider. Iโm just a little token waiting to be attended to."
Because whoever controls the GPUs controls the relationship. Thatโs lab romance.
He starts setting things up, and Iโm trying to act casual, but I donโt understand any of the numbers heโs saying.
Heโs like, "Yeah, I can probably give you four H100s for the weekend."
Iโm nodding like, "Mmm. Four. Weekend. H. One hundred. Absolutely."
Inside Iโm like, "Is that good? Is that prison time? Why did he say it like he was offering me organs?"
[Continue in comments...]
People hear "AI researcher" and they think Iโm building the future. No. Iโm in a basement at 2 a.m. Googling, "CUDA error what the f**k does this mean."
And the worst part about AI research now is compute. You donโt even ask, "Is this idea good?" anymore. You ask, "Can I afford for this idea to be wrong?"
My advisor comes to me one day and says, "I think we should fine-tune our own language model."
I said, "Professor, with what money? Iโm a PhD student. I have two bank accounts: checking and emotionally checking."
He goes, "Donโt worry. We have compute."
Now, in academia, "donโt worry" is never the beginning of a good sentence.
I said, "What do you mean we have compute?"
He said, "My friend knows the cluster admin. He can get us on the GPUs."
I said, "Okayโฆ what do we have to do?"
He goes, "Nothing crazy. Just be very grateful in the acknowledgements."
I said, "How grateful?"
He said, "Maybe put him as co-author."
I said, "Co-author? Are we using the cluster, or is the cluster using us?"
Because at that point, thatโs not a favor. Thatโs academic child support.
So I go to the server room, and the cluster admin walks up to me and goes, "So youโre the NLP student."
And in my head Iโm like, "No, tonight youโre the principal investigator. Youโre the provider. Iโm just a little token waiting to be attended to."
Because whoever controls the GPUs controls the relationship. Thatโs lab romance.
He starts setting things up, and Iโm trying to act casual, but I donโt understand any of the numbers heโs saying.
Heโs like, "Yeah, I can probably give you four H100s for the weekend."
Iโm nodding like, "Mmm. Four. Weekend. H. One hundred. Absolutely."
Inside Iโm like, "Is that good? Is that prison time? Why did he say it like he was offering me organs?"
[Continue in comments...]
- 1 reply
reacted to matteospanio's post with ๐ about 1 month ago
Post
7079
๐ถ Released mule-torch โ an unofficial PyTorch port of MULE (SF-NFNet-F0), SiriusXM/Pandora's music-audio embedding model (McCallum et al., ISMIR 2022).
No retraining: I re-implemented the architecture in pure PyTorch and transferred the original TensorFlow weights, then checked it layer by layer against the genuine TF pipeline.
โ End-to-end clip-embedding cosine 0.9999999 vs the original
โ ONNX backbone parity < 1e-6
โ 62.35M params (paper: ~62.4M)
โ Batched, GPU-native, ONNX-exportable โ none of which the original
๐ค Weights: matteospanio/mule
๐ป Code: https://github.com/matteospanio/mule-torch
๐ฆ PyPI: https://pypi.org/project/mule-torch/
The fun bug: parity was perfect through every conv but the block output was anti-correlated (cos = โ1). Cause: the learnable skip-init gains couldn't be mapped by layer name (Keras scrambles the order) โ they had to be recovered from the graph.
โ ๏ธ Unofficial, community port โ not affiliated with or endorsed by the original authors. All credit to them; please cite the paper. Weights inherit CC-BY-NC-4.0.
No retraining: I re-implemented the architecture in pure PyTorch and transferred the original TensorFlow weights, then checked it layer by layer against the genuine TF pipeline.
โ End-to-end clip-embedding cosine 0.9999999 vs the original
โ ONNX backbone parity < 1e-6
โ 62.35M params (paper: ~62.4M)
โ Batched, GPU-native, ONNX-exportable โ none of which the original
Analysis pipeline doespip install mule-torchfrom mule_torch import MuleModel
emb = MuleModel.from_pretrained()(waveform) # (B, T)@16kHz -> (B, 1728)๐ค Weights: matteospanio/mule
๐ป Code: https://github.com/matteospanio/mule-torch
๐ฆ PyPI: https://pypi.org/project/mule-torch/
The fun bug: parity was perfect through every conv but the block output was anti-correlated (cos = โ1). Cause: the learnable skip-init gains couldn't be mapped by layer name (Keras scrambles the order) โ they had to be recovered from the graph.
โ ๏ธ Unofficial, community port โ not affiliated with or endorsed by the original authors. All credit to them; please cite the paper. Weights inherit CC-BY-NC-4.0.
liked 6
Spaces about 1 month ago
TIPSv2 Feature Explorer
๐
22
Explore TIPSv2 features, segmentations, depth and normals
Ideogram 4
๐จ
227
Ideogram 4 state of the art open weights
Matcha TTS Hebrew
๐
1
Generate Hebrew speech from text or phonemes
TripoSplat
๐
291
Generate 3D Gaussian models from a single image
Cosmos3-Nano
๐
59
NVIDIA Cosmos3-Nano โ text/image to video + audio
Bonsai Image GPU
๐จ
85
Run Bonsai-Image-4B models on GPU
upvoted a collection about 1 month ago
liked a
Space about 1 month ago
liked a
Space about 2 months ago
reacted to sergiopaniego's post with ๐ about 2 months ago
Post
10021
Harness, Scaffold, Context Engineering, Agent... do you actually know what they mean?
We wrote an AI agent glossary and tried to make sense of it all with simple definitions and real examples
โ go read it โ
https://huggingface.co/blog/agent-glossary
We wrote an AI agent glossary and tried to make sense of it all with simple definitions and real examples
โ go read it โ
https://huggingface.co/blog/agent-glossary
- 2 replies
liked a
model about 2 months ago
