fix: prompt formation bug

[kumaranant1]

PREGO/step_anticipation/src/models/llm_hf.py

Lines 147 to 156 in 3078571

	# ! Don't know why it was inside the loop
	# >>>
	prompt_builder = load_data(f"{CONTEXT_PROMPT_PATH}/context_prompt.json")
	init = prompt_builder[prompt_context]["init"]
	if remove_toySequence:
	prompt_ = f"{prompt}{init} {toy_class}\n"
	else:
	prompt_ = f"{prompt}{init} {toy}\n"
	# <<<

Right now, the prompts formed are getting appended to the previous prompts, leading to this behaviour:

For the anticipation of the first symbol, the LLM gets this prompt:

<in-context-prompts>
Sequence type: crane
Input Sequence:
 -1
Next Symbol:

But for the anticipation of the second symbol, the LLM gets this prompt:

<in-context-prompts>
Sequence type: crane
Input Sequence:
 -1
Next Symbol:
Input Sequence:
 -1, attach-wheel
Next Symbol:

However, the expected behaviour while anticipating the second symbol should be this:

<in-context-prompts>
Sequence type: crane
Input Sequence:
 -1, attach-wheel
Next Symbol:

Subsequently, while anticipating the third symbol, the LLM gets this prompt:

<in-context-prompts>
Sequence type: crane
Input Sequence:
 -1
Next Symbol:
Input Sequence:
 -1, attach-wheel
Next Symbol:
Input Sequence:
 -1, attach-wheel, attach-grill
Next Symbol:

However, the expected behaviour while anticipating the third symbol should be this:

<in-context-prompts>
Sequence type: crane
Input Sequence:
 -1, attach-wheel, attach-grill
Next Symbol: