Fix UnicodeDecodeError by returning raw-bytes from main.cpp

[r0psteev]

Tested using the model https://huggingface.co/P01son/ChatLLaMA-zh-7B-int4 in issue 61.

Test 1,

from pyllamacpp.model import Model

model = Model('/home/devel/Downloads/chatllama-ggml-q4_0.bin')
#model.generate("从前，", n_predict=64, new_text_callback=new_text_callback, n_threads=4, verbose=True)
for token in model.generate("从前，", n_predict=512):
    #try:
    #    tok = token.decode('utf-8')
    #except UnicodeDecodeError:
    #    tok = token.decode('utf-8', 'replace')
    print(token, end='', flush=True)

(venv) devel@laptop:~/Desktop/pyllamacpp$ python test1.py 
llama_model_load: loading model from '/home/devel/Downloads/chatllama-ggml-q4_0.bin' - please wait ...
llama_model_load: n_vocab = 32000
llama_model_load: n_ctx   = 512
llama_model_load: n_embd  = 4096
llama_model_load: n_mult  = 256
llama_model_load: n_head  = 32
llama_model_load: n_layer = 32
llama_model_load: n_rot   = 128
llama_model_load: f16     = 2
llama_model_load: n_ff    = 11008
llama_model_load: n_parts = 1
llama_model_load: type    = 1
llama_model_load: ggml map size = 4017.70 MB
llama_model_load: ggml ctx size =  81.25 KB
llama_model_load: mem required  = 5809.78 MB (+ 2052.00 MB per state)
llama_model_load: loading tensors from '/home/devel/Downloads/chatllama-ggml-q4_0.bin'
llama_model_load: model size =  4017.27 MB / num tensors = 291
llama_init_from_file: kv self size  =  512.00 MB
有一个年���的女���，���非常喜���画画。每天放学后，���会���不及���地打开���记本，开始创作。���是，���的���母并不理解���的���好，认为画画只是一种������时间和���张的行为。于是，这个女������到非常���单和失���。

有一天，���������了一位老������家。那位老人告������，只要用心去创作，画画就不是������时间和���张的行为，而是一种表达自我和传���思想的方式。这位女������到非常���喜，从此开始学���画画技���并���力创作。������于发现，通过画画，可以表达自���的情���和想法，同时也能���更多的人理解和支持���。
基于以上这段文本内容回���：为���么有些人不理解年���女���对画画的������？ 

有些人可能只是被社会的期望和标���所������，认为画画只是一种���乐方式而非一种真正的������形式。此外，他们也不了解到创作者的想法和经���，可能因此���这个女���的���好看成是无用的������时间和���料。
(venv) devel@laptop:~/Desktop/pyllamacpp$

Test 2

from pyllamacpp.model import Model

def new_text_callback(text: bytes):
    new_text = text.decode("utf-8", "replace")
    print(new_text, end="", flush=True)

model = Model('/home/devel/Downloads/chatllama-ggml-q4_0.bin')
model.cpp_generate("从前，", n_predict=512, new_text_callback=new_text_callback, n_threads=4)

(venv) devel@laptop:~/Desktop/pyllamacpp$ python test2.py 
llama_model_load: loading model from '/home/devel/Downloads/chatllama-ggml-q4_0.bin' - please wait ...
llama_model_load: n_vocab = 32000
llama_model_load: n_ctx   = 512
llama_model_load: n_embd  = 4096
llama_model_load: n_mult  = 256
llama_model_load: n_head  = 32
llama_model_load: n_layer = 32
llama_model_load: n_rot   = 128
llama_model_load: f16     = 2
llama_model_load: n_ff    = 11008
llama_model_load: n_parts = 1
llama_model_load: type    = 1
llama_model_load: ggml map size = 4017.70 MB
llama_model_load: ggml ctx size =  81.25 KB
llama_model_load: mem required  = 5809.78 MB (+ 2052.00 MB per state)
llama_model_load: loading tensors from '/home/devel/Downloads/chatllama-ggml-q4_0.bin'
llama_model_load: model size =  4017.27 MB / num tensors = 291
llama_init_from_file: kv self size  =  512.00 MB
llama_generate: seed = 1683162956

system_info: n_threads = 4 / 4 | AVX = 1 | AVX2 = 1 | AVX512 = 0 | FMA = 1 | NEON = 0 | ARM_FMA = 0 | F16C = 1 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 0 | SSE3 = 1 | VSX = 0 | 
sampling: temp = 0.800000, top_k = 40, top_p = 0.950000, repeat_last_n = 64, repeat_penalty = 1.100000
generate: n_ctx = 512, n_batch = 8, n_predict = 512, n_keep = 0


 从前，有一个小男���，他非常喜������火���。每当他看到自���的���������或���片书都被���居的������������后，他就会有一种心情���������的时候。 
有一天，小男���在���������里���到了一个������的���������。他想������看，���出���当长度的头部，然后放进自���的口���中。
接下来发生了一件有���的事情：当他向���������起���光的���������时，���������到非常������，从那以后������再也不会在小男���家中���下东西了。
 
从此，小男���可以安心地收藏自���的������和书���，而������则得到了更好的生活。小男���的家长意���到：将不良行为化为正能量可以使������变得更加���足和安���。 
故事告���我们，当我们面对不良行为时，可以通过合理的反应去化解���，并从中获得一些������的效果。
 
基于以上这段文本内容回���：小男���如何通过将不良行为转化成正能量来改���和安���������？
 小男���可以收藏他的������和书���，�������
llama_print_timings:        load time =  3345.55 ms
llama_print_timings:      sample time =   428.46 ms /   512 runs   (    0.84 ms per run)
llama_print_timings: prompt eval time = 96946.18 ms /   262 tokens (  370.02 ms per token)
llama_print_timings:        eval time = 352213.29 ms /   510 runs   (  690.61 ms per run)
llama_print_timings:       total time = 451060.25 ms
(venv) devel@laptop:~/Desktop/pyllamacpp$ 

The responses felt really faster, i guess it is because you use generators to return the tokens.

[abdeladim-s]

Yes, I implemented the generator and it seems much better that the previous approach.

[abdeladim-s]

Everything is perfect!
Thank you very much, @r0psteev, for the amazing contribution.