translate infer problem
Can I translate the model from Korean to English? I made some infer attempts, but the output results were empty. The corresponding input_id and output results are as follows. The generated part is not normal. Or do you have a more suitable translation infer script for reference? Thank you very much! ! !
Response:
input_ids: tensor([[ 1, 13866, 338, 385, 15278, 393, 16612, 263, 3414, 29892,
3300, 2859, 411, 385, 1881, 393, 8128, 4340, 3030, 29889,
14350, 263, 2933, 393, 7128, 2486, 1614, 2167, 278, 2009,
29889, 13, 13, 2277, 29937, 2799, 4080, 29901, 13, 21652,
1494, 16549, 304, 3033, 1674, 13, 13, 2277, 29937, 10567,
29901, 13, 43840, 32764, 31231, 13, 13, 2277, 29937, 13291,
29901]], device='cuda:0')
s: tensor([ 1, 13866, 338, 385, 15278, 393, 16612, 263, 3414, 29892,
3300, 2859, 411, 385, 1881, 393, 8128, 4340, 3030, 29889,
14350, 263, 2933, 393, 7128, 2486, 1614, 2167, 278, 2009,
29889, 13, 13, 2277, 29937, 2799, 4080, 29901, 13, 21652,
1494, 16549, 304, 3033, 1674, 13, 13, 2277, 29937, 10567,
29901, 13, 43840, 32764, 31231, 13, 13, 2277, 29937, 13291,
29901, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166,
30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166,
30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166,
30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166,
30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166,
30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166,
30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166,
30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166,
30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166,
30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166,
30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166,
30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166,
30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166,
30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166,
30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166,
30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166,
30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166,
30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166,
30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166,
30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166,
30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166,
30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166,
30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166,
30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166,
30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166, 30166,
30166, 30166, 30166, 30166, 30166, 30166, 30166], device='cuda:0')
Or what kind of instructions or prompts should I use? thank u
Hi, since llama-2-ko model is pretrained language model, so this is NOT trained to follow the instructions.
If you're looking for translation-specific model, there is tailored for translation instruction: https://huggingface.co/squarelike/Gugugo-koen-7B-V1.1
That model is finetuned on this llama-2-ko base model using several translation datasets.
thank u very much!