| ## OVERVIEW | |
| PEFT model trained on a strided, raw text dataset. | |
| Full training parameters are available in training_parameters.json. | |
| The most salient model feature is a training context window of 1536 tokens, the maximum possible while holding all previous training parameters constant. | |
| Each training window is overlapped with the last 512 tokens of the previous window. | |
| This overlap ratio is less unusual than the 'C' type model and produces less objectionable results in eval. | |
| --- | |
| library_name: peft | |
| --- | |
| ## Training procedure | |
| The following `bitsandbytes` quantization config was used during training: | |
| - 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: fp4 | |
| - bnb_4bit_use_double_quant: False | |
| - bnb_4bit_compute_dtype: float32 | |
| The following `bitsandbytes` quantization config was used during training: | |
| - 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: fp4 | |
| - bnb_4bit_use_double_quant: False | |
| - bnb_4bit_compute_dtype: float32 | |
| The following `bitsandbytes` quantization config was used during training: | |
| - 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: fp4 | |
| - bnb_4bit_use_double_quant: False | |
| - bnb_4bit_compute_dtype: float32 | |
| ### Framework versions | |
| - PEFT 0.5.0.dev0 | |
| - PEFT 0.5.0.dev0 | |
| - PEFT 0.5.0.dev0 | |