vdutor / neural-diffusion-processes Goto Github PK

HI,
When I tried to install this repo regression test, I couldn't test with gpu.
You may need to change the poetry.lock file like this
dm-haiku==0.0.9 -> dm-haiku==0.0.10

Hi,
thank you for your repository and contributions.
I am trying to generate a dataset using regression/generate_data.py
and I encountered an error on line 102.

#####################################
File "/home/coldyoung/neural-diffusion-processes/experiments/regression/generate_data.py", line 30, in
batch = jitted_get_batch(bkey)
File "/home/coldyoung/neural-diffusion-processes/experiments/regression/generate_data.py", line 25, in
jitted_get_batch = jax.jit(lambda k: get_batch(k, batch_size=BATCH_SIZE, name=dataset, task=task, input_dim=input_dim))
File "/home/coldyoung/neural-diffusion-processes/experiments/regression/data.py", line 251, in get_batch
y = jax.vmap(sample_func)(keys, x)
File "/home/coldyoung/neural-diffusion-processes/experiments/regression/data.py", line 113, in sample
f = self.prior.predict(self.params)(x).sample(seed=key, sample_shape=()).reshape(x.shape[0], 1)
File "/home/coldyoung/anaconda3/envs/dpn/lib/python3.9/site-packages/gpjax/gps.py", line 221, in predict
mx = self.mean_function(x)
File "/home/coldyoung/anaconda3/envs/dpn/lib/python3.9/site-packages/gpjax/mean_functions.py", line 151, in call
return jnp.ones((x.shape[0], 1)) * self.constant
AttributeError: 'dict' object has no attribute 'shape'
#####################################
I am unable to determine why this code is using self.params as input in the predict function.

Would you be able to make the necessary changes to the code?
The GPjax library's predict function only accepts array types and not dictionaries.

Could you please resolve this issue?
Thank you.

Hi, I encountered problems with dependencies trying to run the code. Initially I tried to follow the instructions in the README.md
with poetry, the result running the main.py file in regression folder with this command (python main.py --config.dataset=se --config.input_dim=1 --config.restore=trained_models/May26_014440_rcva/) was an error due to conflicts with gpjax (version 0.6.0 and 0.5.9).
I have also tried to download the requirements in requirements_Sept2023.txt with the pip install -r command but so many conflicts emerged.
Please help, I'd like to contribute to the research topic.