metadata
language:
- en
license: llama2
library_name: peft
datasets:
- b-mc2/sql-create-context
pipeline_tag: question-answering
widget:
- question (string): How many heads of the departments are older than 56 ?
context (string): CREATE TABLE head (age INTEGER)
answer (string): SELECT COUNT(*) FROM head WHERE age > 56
- question (string): What are the maximum and minimum budget of the departments?
context (string): CREATE TABLE department (budget_in_billions INTEGER)
answer (string): SELECT MAX(budget_in_billions), MIN(budget_in_billions) FROM department
base_model: codellama/CodeLlama-13b-Instruct-hf
Training procedure
The following bitsandbytes quantization config was used during training:
- quant_method: QuantizationMethod.BITS_AND_BYTES
- 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: bfloat16
Framework versions
- PEFT 0.4.0.dev0