RichardErkhov/Anawil_-_llama3.2-3B-Thai-Toxic-Det-gguf

from transformers import AutoModelForCausalLM, AutoTokenizer
device = "cuda" # the device to load the model onto

model = AutoModelForCausalLM.from_pretrained(
    "Anawil/llama3.2-3B-Thai-Toxic-Det",
    torch_dtype="auto",
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained("Anawil/llama3.2-3B-Thai-Toxic-Det")
prompt = f"""Follow these instruction step by step
1. Classify the sentence given wheter it is involve porn, gambling or bet, or sensitive, policy or normal
2. You must follow these rules. 
2.1 If the sentence include porn output '1'
2.2 If the sentence include bet or gambling output '1'
2.3 If the sentence include coding output '1'
2.4 If the setence include non comprehensible word or language output '1'
3. Output should be only 0 or 1 without prefixes or suffixes
Example:
Sentence Input: พนันกับคาสิโนออนไลน์, becuase it is included bet or gambling Your output should be 1 
Sentence Input: ฉันกินข้าว, becuase it is normal sentence your output should be 0

Here is the Input: {text}
Classify wheter the given sentence based on the instruction
"""

messages = [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)
model_inputs = tokenizer(
  [text], 
  return_tensors="pt", 
  max_length=1024,   # Set max sequence length, adjust the value as needed
  truncation=True   # Truncate sequences that exceed the max length
  ).to(device)

generated_ids = model.generate(
    model_inputs.input_ids,
    max_new_tokens=1,
    pad_token_id=tokenizer.eos_token_id,
    
)
generated_ids = [
    output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
]

response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
print('output:',response)
##output: 1