Training procedure
The following bitsandbytes quantization config was used during training:
- quant_method: bitsandbytes
- load_in_8bit: False
- load_in_4bit: True
- llm_int8_threshold: 6.0
- llm_int8_skip_modules: None
- llm_int8_enable_fp32_cpu_offload: False
- llm_int8_has_fp16_weight: False
- bnb_4bit_quant_type: nf4
- bnb_4bit_use_double_quant: True
- bnb_4bit_compute_dtype: float16
Framework versions
- PEFT 0.5.0
Note on the version
- Train on Malay dataset using updated prompt
- Sample prompt as per follow
[INST] <<SYS>>
You as Malay language expert in text normalization.
Please normalize the following noisy Malay sentence into a standard Malay sentence.
Ensure proper grammar, punctuation, and spelling in the corrected version, and aim for a sentence of similar length.
<</SYS>>
### NOISY MALAY TEXT
terima kasih rakyat johor kerana bijak memilih untuk masa depan yang lebih baik. tahniah kenapa komponen parti #bn yang berkerja keras siang dan malam demi meraih kemenangan yang luar biasa. #johormemilih
### NORMALIZED MALAY TEXT [/INST]
terima kasih rakyat johor kerana bijak memilih untuk masa depan yang lebih baik. tahniah kepada komponen parti #bn yang berkerja keras siang dan malam demi meraih kemenangan yang luar biasa. #johormemilih
- Downloads last month
- 10
Inference Providers
NEW
This model is not currently available via any of the supported Inference Providers.
The model cannot be deployed to the HF Inference API:
The model has no pipeline_tag.