qwen3-4b-uzbek-v2-bnb-4bit

bitsandbytes nf4 4-bit quant (~3.4 gb) of inspirebek/qwen3-4b-uzbek-v2. nvidia gpu only; easiest hf-native 4-bit load.

usage

from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
import torch

bnb = BitsAndBytesConfig(
 load_in_4bit=True,
 bnb_4bit_quant_type="nf4",
 bnb_4bit_compute_dtype=torch.bfloat16,
 bnb_4bit_use_double_quant=True,
)
tok = AutoTokenizer.from_pretrained("inspirebek/qwen3-4b-uzbek-v2-bnb-4bit")
model = AutoModelForCausalLM.from_pretrained(
 "inspirebek/qwen3-4b-uzbek-v2-bnb-4bit",
 quantization_config=bnb,
 device_map="auto",
)

quantization

• method: bitsandbytes nf4 (4-bit normalfloat)
• double quantization: enabled
• compute dtype: bfloat16

datasets

stage a — fluency (continued pretraining):

• yakhyo/uz-wiki · MIT
• tahrirchi/uz-books-v2 · MIT
• tahrirchi/uz-crawl · Apache-2.0

stage b — instruct (sft):

• saillab/alpaca_uzbek_taco · CC-BY-NC-4.0
• behbudiy/alpaca-cleaned-uz · CC-BY-4.0
• UAzimov/uzbek-instruct-llm · Apache-2.0
• CohereLabs/aya_collection_language_split · Apache-2.0
• med-alex/qa_mt_ru_to_uzn · unspecified
• med-alex/qa_mt_tr_to_uzn · unspecified

⚠️ licensing note: saillab/alpaca_uzbek_taco is cc-by-nc-4.0, which restricts commercial use of derivative models. downstream users who need a fully permissive license should retrain without that subset.

sibling formats

• inspirebek/qwen3-4b-uzbek-v2
• inspirebek/qwen3-4b-uzbek-v2-lora
• inspirebek/qwen3-4b-uzbek-v2-bnb-4bit
• inspirebek/qwen3-4b-uzbek-v2-awq
• inspirebek/qwen3-4b-uzbek-v2-GGUF

Base model

Qwen/Qwen3-4B-Base

Qwen/Qwen3-4B

inspirebek/qwen3-4b-uzbek-v2

(3)

all distribution formats of the uzbek fine-tune of qwen3-4b: merged bf16, lora adapter, bnb nf4, awq, and the gguf suite. • 5 items • Updated Apr 20 • 1