nouamanetazi/hf-ar-134000

1p46G-gemma-fp-dedup-rehydr-ar-350BT-seed-6/transformers/134000

Tokenizer: google/gemma-7b

from transformers import AutoTokenizer, AutoModelForCausalLM
# Initialize model and tokenizer
TEST_PROMPT = "الزرادشتية هي ديانة انتشرت في بلاد"
save_path = "nouamanetazi/hf-ar-134000"
tokenizer = AutoTokenizer.from_pretrained(save_path)
input_ids = tokenizer(TEST_PROMPT, return_tensors="pt")["input_ids"].cuda() # google/gemma-7b
print("Input prompt:", tokenizer.batch_decode(input_ids)[0])

model = AutoModelForCausalLM.from_pretrained(save_path, device="cuda", dtype=torch.bfloat16)
outputs = model.generate(input_ids, max_new_tokens=100)
print("Generated text:", tokenizer.batch_decode(outputs)[0])