Model Card for Model ID
<!-- Provide a quick summary of what the model is/does. -->
Model Details
Version 1 of 71, qLora finetune of the Mistral-7B model using 1/71 of the GPT4 based part of the Orca Dataset, using approx ~14k records from a total of 1 million records.
Model Description
- Developed by: Marlind Maksuti @ StochastX
- Model type: Mistral-7B
- Language(s) (NLP): Albanian, Shqip
- Finetuned from model: Mistral-7B-SlimOrca
Model Sources
- Repository: https://huggingface.co/Open-Orca/Mistral-7B-SlimOrca
Uses
Text generation in Albanian.
Bias, Risks, and Limitations
Model is just version v0.1, outputs are still not optimal.
Training Hyperparameters
- Training regime: [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
Training procedure
The following bitsandbytes quantization config was used during training:
- quant_method: QuantizationMethod.BITS_AND_BYTES
- load_in_8bit: True
- load_in_4bit: False
- 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 0.6.0.dev0
Training procedure
The following bitsandbytes quantization config was used during training:
- quant_method: QuantizationMethod.BITS_AND_BYTES
- load_in_8bit: True
- load_in_4bit: False
- 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 0.6.0.dev0
