malaysian-mistral-1.1B-4096-GGUF/README.md

base_model, inference, language, model_creator, model_name, pipeline_tag, quantized_by, tags

base_model	inference	language	model_creator	model_name	pipeline_tag	quantized_by	tags
mesolitica/malaysian-mistral-1.1B-4096	false	ms	mesolitica	malaysian-mistral-1.1B-4096	text-generation	afrideva	gguf ggml quantized q2_k q3_k_m q4_k_m q5_k_m q6_k q8_0

mesolitica/malaysian-mistral-1.1B-4096-GGUF

Quantized GGUF model files for malaysian-mistral-1.1B-4096 from mesolitica

Name	Quant method	Size
malaysian-mistral-1.1b-4096.fp16.gguf	fp16	2.25 GB
malaysian-mistral-1.1b-4096.q2_k.gguf	q2_k	491.42 MB
malaysian-mistral-1.1b-4096.q3_k_m.gguf	q3_k_m	561.96 MB
malaysian-mistral-1.1b-4096.q4_k_m.gguf	q4_k_m	682.68 MB
malaysian-mistral-1.1b-4096.q5_k_m.gguf	q5_k_m	799.13 MB
malaysian-mistral-1.1b-4096.q6_k.gguf	q6_k	922.87 MB
malaysian-mistral-1.1b-4096.q8_0.gguf	q8_0	1.19 GB

Original Model Card:

Pretrain 1.1B 4096 context length Mistral on Malaysian text

README at https://github.com/mesolitica/malaya/tree/5.1/pretrained-model/mistral

• Dataset gathered at https://github.com/malaysia-ai/dedup-text-dataset/tree/main/pretrain-llm
• We use Ray cluster to train on 5 nodes of 4x A100 80GB, https://github.com/malaysia-ai/jupyter-gpu/tree/main/ray

WandB, https://wandb.ai/mesolitica/pretrain-mistral-1.1b?workspace=user-husein-mesolitica

WandB report, https://wandb.ai/mesolitica/pretrain-mistral-3b/reports/Pretrain-Larger-Malaysian-Mistral--Vmlldzo2MDkyOTgz

how-to

from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig
import torch

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-1.1B-4096', model_input_names = ['input_ids'])
model = AutoModelForCausalLM.from_pretrained(
    'mesolitica/malaysian-mistral-1.1B-4096',
    use_flash_attention_2 = True,
    quantization_config = nf4_config
)
prompt = '<s>nama saya'
inputs = tokenizer([prompt], return_tensors='pt', add_special_tokens=False).to('cuda')

generate_kwargs = dict(
    inputs,
    max_new_tokens=512,
    top_p=0.95,
    top_k=50,
    temperature=0.9,
    do_sample=True,
    num_beams=1,
    repetition_penalty=1.05,
)
r = model.generate(**generate_kwargs)