Update README.md

README.md

@@ -38,73 +38,6 @@ paraphraser.paraphrase("Hey, can yuo hepl me cancel my last order?", tone="witty
 # "Hey, I need your help with my last order. Can you wave your magic wand and make it disappear?"
 ```
 
-OR use directly with transformers
-
-```python
-from transformers import AutoModelForCausalLM, AutoTokenizer, StoppingCriteria, StoppingCriteriaList
-
-model_id = "llm-toys/RedPajama-INCITE-Base-3B-v1-paraphrase-tone"
-DEVICE = "cuda"
-EOC_FORMAT = "\n\n### END"
-PARAPHRASE_PREDICT_FORMAT = (
-    "### Instruction:\nGenerate a paraphrase for the following Input sentence.\n\n"
-    "### Input:\n{input_text}\n\n### Response:\n"
-)
-TONE_CHANGE_PREDICT_FORMAT = (
-    "### Instruction:\nChange the tone of the following Input sentence to {tone}.\n\n"
-    "### Input:\n{input_text}\n\n### Response:\n"
-)
-
-tokenizer = AutoTokenizer.from_pretrained(model_id)
-model = AutoModel.from_pretrained(mode_id).to(DEVICE)
-
-class StoppingCriteriaSub(StoppingCriteria):
-  """Helps in stopping the generation when a certain sequence of tokens is generated."""
-  
-  def __init__(self, stops: list = []):
-      super().__init__()
-      self.stops = stops
-  
-  def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor) -> bool:
-      return input_ids[0][-len(self.stops) :].tolist() == self.stops
-
-
-stopping_criteria = StoppingCriteriaList(
-  [StoppingCriteriaSub(stops=tokenizer(EOC_FORMAT)["input_ids"])]
-)
-
-def predict(input_text: str) -> str:
-  tokenized = tokenizer(
-      input_text,
-      max_length=self.max_length,
-      padding=True,
-      truncation=True,
-      return_tensors="pt",
-  )
-
-  with torch.no_grad():
-      out = model.generate(
-          input_ids=tokenized["input_ids"].to(DEVICE),
-          attention_mask=tokenized["attention_mask"].to(DEVICE),
-          pad_token_id=self.tokenizer.eos_token_id,
-          max_new_tokens=max_new_tokens,
-          num_return_sequences=num_return_sequences,
-          do_sample=True,
-          temperature=temperature,
-          top_p=top_p,
-          stopping_criteria=self.stopping_criteria,
-      )
-
-  out_texts = [tokenizer.decode(o, skip_special_tokens=True) for o in out]
-  return out_texts
-
-print("Paraphrasing:")
-print(predict(PARAPHRASE_PREDICT_FORMAT.format(input_text="If you have any further questions, feel free to ask.")))
-
-print("Tone change:")
-print(predict(TONE_CHANGE_PREDICT_FORMAT.format(input_text="If you have any further questions, feel free to ask.", tone="professional")))
-```
-
 ## Sample training data
 
 ```json