Use max_length instead of sequence_len since they are the same values

src/axolotl/prompt_strategies/completion.py

@@ -53,8 +53,8 @@ def tokenize_prompt(self, prompt):
  tokenized_full_prompt = self._tokenize(full_prompt)
 
  for key, val in tokenized_full_prompt.items():
- for i in range(0, len(val), self.sequence_len):
- res[key].append(val[i : i + self.sequence_len])
+ for i in range(0, len(val), self.max_length):
+ res[key].append(val[i : i + self.max_length])
 
  return dict(res)
 

src/axolotl/prompt_strategies/metharme.py

@@ -31,7 +31,7 @@ def _tokenize(
  result = self.tokenizer(
  prompt,
  truncation=True,
- max_length=self.sequence_len,
+ max_length=self.max_length,
  padding=False,
  return_tensors=None,
  )
@@ -43,7 +43,7 @@ def _tokenize(
 
  if num_eos_tokens > 0 and add_eos_token and len(result["input_ids"]) > 0:
  for _ in range(num_eos_tokens):
- if len(result["input_ids"]) < self.sequence_len:
+ if len(result["input_ids"]) < self.max_length:
  result["input_ids"].append(self.tokenizer.eos_token_id)
  result["attention_mask"].append(1)