ddjohnson

8 ms·

1. 1.▲
   by ddjohnson 1y ago

   One of the blog post authors here! We evaluated o3 through the API, where the model does not have access to any specific built-in tools (although it does have the capability to use tools, and allows you to provide your own tools). This is d
2. 2.▲
   by ddjohnson 1y ago

   One of the blog post authors here! I think this finding is pretty surprising at the purely behavioral level, without needing to anthropomorphize the models. Two specific things I think are surprising: - This appears to be a regression relat
3. 3.▲
   by ddjohnson 2y ago

   I'm curious what interop difficulties you've run into in JAX? In my experience, the JAX ecosystem is quite modular and most JAX libraries work pretty well together. Penzai's core visualization tooling should work for most JAX
4. 4.▲
   by ddjohnson 2y ago

   Thanks! This is one of the more experimental design choices I made in designing Penzai, but so far I've found it to be quite useful. The effect system does come with a few sharp edges at the moment if you want to use JAX transformation
5. 5.▲
   by ddjohnson 2y ago

   Author of Penzai here! In idiomatic Penzai usage, you should always discharge all effects before running your model. While it's true you can't do `discharge_effect(jax.grad(your_model_here))`, you can still do `jax.grad(discharge_
6. 6.▲
   by ddjohnson 2y ago

   Author of Penzai here. I think the answer is a bit more nuanced (and closer to "yes") than this: - If you want to use the treescope pretty-printer or the pz.select tree manipulation utility, those should work out-of-the-box with b
7. 7.▲
   by ddjohnson 2y ago

   Author here! I didn't know PyTorch had its own pytree system. It looks like it's separate from JAX's pytree registry, though, so Penzai's tooling probably won't work with PyTorch out of the box.