Make models use bptt parameter

ulmfit/__main__.py

@@ -50,7 +50,8 @@ class ULMFiT:
         return FireView(train=params.train_cls, validate_cls=params.validate_cls)
 
 
-    def eval_noise_resistance(self, lang="de", size=1, prefix_name="", model="sp15k/qrnn_nl4.m"):
+    def eval_noise_resistance(self, lang="de", size=1, prefix_name="", model="sp15k/qrnn_nl4.m",
+                              num_cls_epochs=8, bs=18, lr_sched="1cycle", label_smoothing_eps=0.0):
         def first_or_default(l, default=None):
             l = list(l)
             if l:
@@ -63,9 +64,10 @@ class ULMFiT:
                           name=f"nl4_{prefix_name}{noise}",
                           noise=noise/100,
                           dataset_template='${lang}-'+str(size),
-                          num_cls_epochs=8,
-                          bs=18,
-                          lr_sched="1cycle")
+                          num_cls_epochs=num_cls_epochs,
+                          bs=bs,
+                          lr_sched=lr_sched,
+                          label_smoothing_eps=label_smoothing_eps)
             val = first_or_default(d.values(), default=-1)
             results.append((noise/100, val))
         df = pd.DataFrame(results, columns=["noise", "accuracy"])

ulmfit/pretrain_lm.py

@@ -175,7 +175,7 @@ class LMHyperParams:
         self.model_dir.mkdir(exist_ok=True, parents=True)
         data_lm = self.load_wiki_data(bs=bs) if data_lm is None else data_lm
         learn = self.create_lm_learner(data_lm, drop_mult=drop_mult, label_smoothing_eps=label_smoothing_eps)
-
+        print("Bptt", data_lm.bptt)
         learn.true_wd = true_wd
         if num_epochs > 0:
             if self.pretrained_fnames or self.pretrained_model:
@@ -255,6 +255,7 @@ class LMHyperParams:
                           classes=None,
                           bs=bs,
                           text_cols='texts',
+                          bptt=self.bptt,
                           **args)
 
         itos, stoi, trn_path = data_lm.vocab.itos, data_lm.vocab.stoi, data_lm.path

ulmfit/train_clas.py

@@ -112,7 +112,7 @@ class CLSHyperParams(LMHyperParams):
         trn_args=dict(bptt=self.bptt, clip=self.clip)
         trn_args.update(kwargs)
         learn = text_classifier_learner(data_clas, AWD_LSTM, config=config,
-            pretrained=False, path=self.model_dir.parent, model_dir=self.model_dir.name, **trn_args)
+            pretrained=False, path=self.model_dir.parent, model_dir=self.model_dir.name, bptt=self.bptt, **trn_args)
 
         if self.pretrained_model is not None:
             print("Loading pretrained model")
@@ -151,7 +151,7 @@ class CLSHyperParams(LMHyperParams):
     def merge_cols(self, df):
         if len(df.columns) <= 2:
             return df
-        ndf = df[[0,1]].copy()
+        ndf = df[[0,1]].copy().fillna(" ")
         for i in range(2, len(df.columns)):
             ndf[1] += ("\n" + FLD + "\n") + df[i].fillna(" ")
 
@@ -213,7 +213,8 @@ class CLSHyperParams(LMHyperParams):
             cls_name = f'{cls_name}noise{noise}tv'
 
         args = self.tokenizer_to_fastai_args(sp_data_func=lambda: trn_df[1], use_moses=use_moses)
-        data_lm = self.lm_databunch('lm', train_df=lm_trn_df, valid_df=lm_val_df, bs=bs, force=force, **args)
+        lm_suffix = self.bptt if self.bptt != 70 else ""
+        data_lm = self.lm_databunch(f'lm{lm_suffix}', train_df=lm_trn_df, valid_df=lm_val_df, bs=bs, force=force, bptt=self.bptt, **args)
         args['vocab'] = data_lm.vocab
         data_cls = self.cls_databunch(cls_name, train_df=trn_df, valid_df=val_df, bs=bs, force=force, **args)
         data_tst = self.cls_databunch('tst', train_df=val_df, valid_df=tst_df, bs=bs, force=force, **args) # Hack to load test dataset with labels