L2R Learner

Learner for Learning to Rank Applications

This module contains a specialized version of fastai’s full fledged Learner. Every functionality here can also be achieved with fastai’s Learner. The purpose of re-creating a learner was purely educational.

---

source

L2RLearner

 L2RLearner (model, dls, grad_func, loss_func, lr, cbs, opt_func=<function
             SGD>, path=None, moms:tuple=(0.95, 0.08, 0.95))

Initialize self. See help(type(self)) for accurate signature.

Serializing

---

source

L2RLearner.save

 L2RLearner.save (file, with_opt=True, pickle_protocol=2)

Save model and optimizer state (if ‘with_opt’) to self.path/file

---

source

L2RLearner.load

 L2RLearner.load (file, device=None, with_opt=True, strict=True)

Load model and optimizer state (if with_opt) from self.path/file using device

---

source

L2RLearner.show_results

 L2RLearner.show_results (device=None, k=None)

Produces the ranking for 100 random labels

Learner convenience functions

---

source

get_learner

 get_learner (model, dls, grad_fn=<function rank_loss3>, loss_fn=<function
              loss_fn2>, lr=1e-05, cbs=None,
              opt_func=functools.partial(<function SGD at 0x7f99c2c4a1f0>,
              mom=0.9), lambrank=False, **kwargs)