VLM API Tester
🦆
Sanity-test a deployed SmolVLM chatbot API
David Bertoin PRO
Bertoin
AI & ML interests
None yet
Recent Activity
updated a Space 2 days ago
Bertoin/vlm-api-tester published a Space 2 days ago
Bertoin/vlm-api-tester new activity 2 days ago
Photoroom/prx-512-t2i-sft-distilled:Image2Image?Organizations
updated a
Space 2 days ago
published a
Space 2 days ago
Image2Image?
1
#2 opened 8 days ago
by 
MinhNH232331M
more info please ? it based on what ?
1
#2 opened 12 days ago
by 
pikkaa
the NOTICE file is wrong
👍 1
1
#1 opened 12 days ago
by 
dummy9996
Turbo / Distilled model in the works?
1
#3 opened 10 days ago
by 
saftle
liked 2
Spaces 9 days ago
updated a
model 9 days ago
[hf staff] UI suggestion - stream the denoising process
2
#1 opened 10 days ago
by 
linoyts
updated a
collection 12 days ago
published a
model 12 days ago
published a
Space 12 days ago
updated a
Space 13 days ago
CFG is 1.0 for distilled models
1
#1 opened about 1 month ago
by 
catwell
CFG is 1.0 for distilled models
1
#1 opened about 1 month ago
by catwell
liked a
dataset 3 months ago
upvoted a collection 4 months ago
commented on PRX Part 3 — Training a Text-to-Image Model in 24h! 4 months ago
I don’t think we really observed such artifacts, or at least not due to LPIPS. We did see something similar without LPIPS when using JIT on undertrained models, but I don’t think we observe them anymore. It might simply be because the model had not converged yet.
One situation where we started to observe checkerboard patterns was during reward alignment with GRPO or SRPO on one of these models. In that case the artifacts were much more noticeable, but I don’t think they were related to LPIPS since we were not using it at the time.
What type of model were you using?
upvoted an article 4 months ago
Article
PRX Part 3 — Training a Text-to-Image Model in 24h!
Photoroom
• • 67