You can deploy this model and start earning money today!

RichardErkhov/
mesolitica_-_malaysian-mistral-7b-32k-instructions-gguf

Name: mesolitica_-_malaysian-mistral-7b-32k-instructions-gguf
Availability: InStock

endpoints_compatible

gguf

template

Views

No views yet

Quantization made by Richard Erkhov.

Github

Discord

Request more models

malaysian-mistral-7b-32k-instructions - GGUF

Model creator: https://huggingface.co/mesolitica/
Original model: https://huggingface.co/mesolitica/malaysian-mistral-7b-32k-instructions/

Name	Quant method	Size
malaysian-mistral-7b-32k-instructions.Q2_K.gguf	Q2_K	2.53GB
malaysian-mistral-7b-32k-instructions.IQ3_XS.gguf	IQ3_XS	2.81GB
malaysian-mistral-7b-32k-instructions.IQ3_S.gguf	IQ3_S	2.96GB
malaysian-mistral-7b-32k-instructions.Q3_K_S.gguf	Q3_K_S	2.95GB
malaysian-mistral-7b-32k-instructions.IQ3_M.gguf	IQ3_M	3.06GB
malaysian-mistral-7b-32k-instructions.Q3_K.gguf	Q3_K	3.28GB
malaysian-mistral-7b-32k-instructions.Q3_K_M.gguf	Q3_K_M	3.28GB
malaysian-mistral-7b-32k-instructions.Q3_K_L.gguf	Q3_K_L	3.56GB
malaysian-mistral-7b-32k-instructions.IQ4_XS.gguf	IQ4_XS	3.67GB
malaysian-mistral-7b-32k-instructions.Q4_0.gguf	Q4_0	3.83GB
malaysian-mistral-7b-32k-instructions.IQ4_NL.gguf	IQ4_NL	3.87GB
malaysian-mistral-7b-32k-instructions.Q4_K_S.gguf	Q4_K_S	3.86GB
malaysian-mistral-7b-32k-instructions.Q4_K.gguf	Q4_K	4.07GB
malaysian-mistral-7b-32k-instructions.Q4_K_M.gguf	Q4_K_M	4.07GB
malaysian-mistral-7b-32k-instructions.Q4_1.gguf	Q4_1	4.24GB
malaysian-mistral-7b-32k-instructions.Q5_0.gguf	Q5_0	4.65GB
malaysian-mistral-7b-32k-instructions.Q5_K_S.gguf	Q5_K_S	4.65GB
malaysian-mistral-7b-32k-instructions.Q5_K.gguf	Q5_K	4.78GB
malaysian-mistral-7b-32k-instructions.Q5_K_M.gguf	Q5_K_M	4.78GB
malaysian-mistral-7b-32k-instructions.Q5_1.gguf	Q5_1	5.07GB
malaysian-mistral-7b-32k-instructions.Q6_K.gguf	Q6_K	5.53GB
malaysian-mistral-7b-32k-instructions.Q8_0.gguf	Q8_0	7.17GB

Original model description:

language:

Full Parameter Finetuning 7B 32768 context length Mistral on Malaysian instructions dataset

README at https://github.com/mesolitica/malaya/tree/5.1/session/mistral#instructions-7b-16384-context-length

We use exact Mistral Instruct chat template.

WandB, https://wandb.ai/mesolitica/fpf-mistral-7b-hf-instructions-16k?workspace=user-husein-mesolitica

how-to

python

from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig
import torch
import json

def parse_mistral_chat(messages, function_call = None):

    user_query = messages[-1]['content']

    users, assistants = [], []
    for q in messages[:-1]:
        if q['role'] == 'user':
            users.append(q['content'])
        elif q['role'] == 'assistant':
            assistants.append(q['content'])

    texts = ['<s>']
    
    if function_call:
        fs = []
        for f in function_call:
            f = json.dumps(f, indent=4)
            fs.append(f)
        fs = '\n\n'.join(fs)
        texts.append(f'\n[FUNCTIONCALL]\n{fs}\n')
        
    for u, a in zip(users, assistants):
        texts.append(f'[INST] {u.strip()} [/INST] {a.strip()}</s>')

    texts.append(f'[INST] {user_query.strip()} [/INST]')
    prompt = ''.join(texts).strip()
    return prompt

TORCH_DTYPE = 'bfloat16'
nf4_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_quant_type='nf4',
    bnb_4bit_use_double_quant=True,
    bnb_4bit_compute_dtype=getattr(torch, TORCH_DTYPE)
)

tokenizer = AutoTokenizer.from_pretrained('mesolitica/malaysian-mistral-7b-32k-instructions')
model = AutoModelForCausalLM.from_pretrained(
    'mesolitica/malaysian-mistral-7b-32k-instructions',
    use_flash_attention_2 = True,
    quantization_config = nf4_config
)

messages = [
    {'role': 'user', 'content': 'kwsp tu apa'}
]
prompt = parse_mistral_chat(messages)
inputs = tokenizer([prompt], return_tensors='pt', add_special_tokens=False).to('cuda')
generate_kwargs = dict(
    inputs,
    max_new_tokens=1024,
    top_p=0.95,
    top_k=50,
    temperature=0.9,
    do_sample=True,
    num_beams=1,
)
r = model.generate(**generate_kwargs)
tokenizer.decode(r[0])

<s> [INST] kwsp tu apa [/INST]KWSP bermaksud Kumpulan Wang Simpanan Pekerja. Ia adalah sebuah institusi simpanan persaraan yang ditubuhkan oleh Kementerian Kewangan Malaysia untuk tujuan mengumpul simpanan ahli untuk dibayar pada umur persaraan, penuh atau penuh persaraan penuh. KWSP ditubuhkan pada tahun 1951 dan mula beroperasi pada tahun 1952. KWSP adalah salah satu institusi simpanan persaraan terbesar di dunia, dengan pangkalan ahli sekitar 14 juta ahli.</s>

python

messages = [
    {'role': 'user', 'content': 'camne nak kill pid dalam linux'}
]
prompt = parse_mistral_chat(messages)
inputs = tokenizer([prompt], return_tensors='pt', add_special_tokens=False).to('cuda')
generate_kwargs = dict(
    inputs,
    max_new_tokens=1024,
    top_p=0.95,
    top_k=50,
    temperature=0.9,
    do_sample=True,
    num_beams=1,
)
r = model.generate(**generate_kwargs)
print(tokenizer.decode(r[0]))

text

<s> [INST] camne nak kill pid dalam linux [/INST]Untuk membunuh pid dalam Linux, ikuti langkah-langkah berikut:

Cari pid menggunakan perintah `ps aux | grep 'pid'`. Pilih pid yang anda ingin bunuh.

Gunakan perintah `kill` diikuti dengan nombor pid. Sebagai contoh, jika anda ingin membunuh pid dengan nombor 1234, gunakan perintah `kill -1234`.

Sahkan pembunuhan dengan menjalankan perintah `kill -id`. Jika perintah tidak mengembalikan sebarang ralat, maka pid telah dibunuh dengan berjaya.

Anda juga boleh menggunakan perintah `kill -s` untuk membunuh semua pid dengan nama atau atribut tertentu.</s>

python

f = {
    'name': 'parse_entities',
    'description': 'extract entities from the text',
    'parameters': {
        'type': 'object',
        'properties': {
            'drink': {
                'type': 'string',
                'description': 'drink name',
            },
            'event': {
                'type': 'string',
                'description': 'event name',
            },
            'person_name': {
                'type': 'string',
                'description': 'person name',
            }
        },
        'required': [
            'drink',
            'event',
            'person_name'
        ]
    }
}
messages = [
    {'role': 'user', 'content': 'nama saya husein bin zolkepli, saya sekarang berada di jomheboh 2023 sambil minum starbucks'}
]
prompt = parse_mistral_chat(messages, function_call = [f])
inputs = tokenizer([prompt], return_tensors='pt', add_special_tokens=False).to('cuda')
generate_kwargs = dict(
    inputs,
    max_new_tokens=128,
    top_p=0.95,
    top_k=50,
    temperature=0.9,
    do_sample=True,
    num_beams=1,
)
r = model.generate(**generate_kwargs)
print(tokenizer.decode(r[0]))

text

<s> 
[FUNCTIONCALL]
{
    "name": "parse_entities",
    "description": "extract entities from the text",
    "parameters": {
        "type": "object",
        "properties": {
            "drink": {
                "type": "string",
                "description": "drink name"
            },
            "event": {
                "type": "string",
                "description": "event name"
            },
            "person_name": {
                "type": "string",
                "description": "person name"
            }
        },
        "required": [
            "drink",
            "event",
            "person_name"
        ]
    }
}
[INST] nama saya husein bin zolkepli, saya sekarang berada di jomheboh 2023 sambil minum starbucks [/INST] <functioncall> {"name": "parse_entities", "arguments": '{
  "drink": "Starbucks",
  "event": "Jom Heboh 2023",
  "person_name": "Husein Bin Zolkepli"
}'}</s>

RichardErkhov/mesolitica_-_malaysian-mistral-7b-32k-instructions-gguf

Original model description:

Full Parameter Finetuning 7B 32768 context length Mistral on Malaysian instructions dataset

how-to

RichardErkhov/
mesolitica_-_malaysian-mistral-7b-32k-instructions-gguf