How to implement this Python class?

[AntonLobach]

Hello, I would like to translate this Python code to Basilisp, but I'm stuck.

Basically, I'm not sure how to correctly translate the init method and how to invoke call methods.

I've checked the test suite and the docs, but couldn't come up with anything working based off of it.

import torch.nn as nn
import torch.nn.functional as F

class Model(nn.Module):
    def __init__(self) -> None:
        super().__init__()
        self.conv1 = nn.Conv2d(1, 20, 5)
        self.conv2 = nn.Conv2d(20, 20, 5)

    def forward(self, x):
        x = F.relu(self.conv1(x))
        return F.relu(self.conv2(x))

My attempt:

(ns core
  (:import
   [torch.nn :as nn]
   [torch.nn.functional :as F]))
   
 (deftype MyModule
    [conv1 conv2]
  ^:abstract nn/Module
  ;; How to implement __init__?
  (forward [self x]
    (-> x
        conv1 ;; Doesn't work, how to call a class with a __call__ method?
        F/relu
        conv2
        F/relu)))

[ikappaki]

Hi @AntonLobach,

The following approach seems to have worked for me: I directly implemented the __init__ function and called conv1 as the first argument in the threading macro:

(ns module
  (:import
   torch
   [torch.nn :as nn]
   [torch.nn.functional :as F]))


(deftype MyModule [^:mutable conv1 ^:mutable conv2]
  ^:abstract nn/Module
  (__init__ [self]
    (.__init__ (super))
    (set! conv1 (nn/Conv2d 1 20 5))
    (set! conv2 (nn/Conv2d 20 20 5))
    nil)
  (forward [self x]
    (try 
      (-> x
          (conv1)
          (F/relu)
          (conv2)
          (F/relu))
      (catch Exception e
        (println :error e)))))


(def model (MyModule))

(def x (torch/randn 4 1 28 28))

(println  (model x))
;; => tensor([[[[2.2859e-02, 0.0000e+00, 0.0000e+00,  ..., 1.5598e-02,
;; =>      0.0000e+00, 0.0000e+00],
;; =>      ...   

[AntonLobach]

Hi @ikappaki,

Ah, I see, I was actually so close to correctly implementing it, thanks a lot!