Bug fixes

ontolearner/learner/taxonomy_discovery/alexbek.py

@@ -560,7 +560,7 @@ def lr_lambda(step):
                 if step < warmup_steps:
                     return (step + 1) / (warmup_steps + 1)
                 else:
-                    progress = (step - warmup_steps) / (total_steps - warmup_steps)
+                    progress = (step - warmup_steps) / max(1,(total_steps - warmup_steps))
                     return max(0.1, 0.5 * (1 + math.cos(math.pi * progress)))
 
             scheduler = torch.optim.lr_scheduler.LambdaLR(optimizer, lr_lambda=lr_lambda)

ontolearner/learner/taxonomy_discovery/rwthdbis.py

@@ -1078,7 +1078,7 @@ def tokenize_batch(batch: Dict[str, List[str]]):
             )
 
         tokenized_dataset = dataset_dict.map(
-            tokenize_batch, batched=True, remove_columns=dataset_dict["train"].column_names
+            tokenize_batch, batched=True, remove_columns=["text"]
         )
         data_collator = DataCollatorWithPadding(self.tokenizer)
 
@@ -1109,14 +1109,13 @@ def tokenize_batch(batch: Dict[str, List[str]]):
             fp16=self.fp16,
             bf16=self.bf16,
             report_to="none",
-            save_safetensors=True,
         )
 
         trainer = Trainer(
             model=self.model,
             args=training_args,
             train_dataset=tokenized_dataset["train"],
-            tokenizer=self.tokenizer,
+            processing_class=self.tokenizer,
             data_collator=data_collator,
         )
         trainer.train()

ontolearner/learner/taxonomy_discovery/skhnlp.py

@@ -181,7 +181,6 @@ def __init__(
             per_device_eval_batch_size=per_device_eval_batch_size,
             warmup_steps=warmup_steps,
             weight_decay=weight_decay,
-            logging_dir=logging_dir,
             logging_steps=logging_steps,
             eval_strategy=eval_strategy,
             save_strategy=save_strategy,

ontolearner/learner/term_typing/alexbek.py

@@ -823,7 +823,7 @@ def load(
 
         # Tokenizer
         self.tokenizer = AutoTokenizer.from_pretrained(
-            self.cfg["llm_model_id"], padding_side="left", token=self.cfg["token"]
+            self.cfg["llm_model_id"], padding_side="left", token=self.cfg["token"] or None
         )
         if self.tokenizer.pad_token is None:
             self.tokenizer.pad_token = self.tokenizer.eos_token
@@ -833,7 +833,7 @@ def load(
             self.cfg["llm_model_id"],
             device_map=device_map,
             torch_dtype=torch_dtype,
-            token=self.cfg["token"],
+            token=self.cfg["token"] or None,
         )
 
         # Deterministic decoding defaults

ontolearner/learner/term_typing/rwthdbis.py

@@ -1027,7 +1027,7 @@ def tokenize_batch(batch: Dict[str, List[str]]):
             model=self.model,
             args=training_args,
             train_dataset=tokenized["train"],
-            tokenizer=self.tokenizer,
+            processing_class=self.tokenizer,
             data_collator=data_collator,
         )