metadata
license: mit
library_name: peft
tags:
- llama-factory
- lora
- generated_from_trainer
base_model: cognitivecomputations/dolphin-2_6-phi-2
model-index:
- name: dolphin-2_6-phi-2-sft-glaive-function-calling-v2-ep1-lora
results: []
dolphin-2_6-phi-2-sft-glaive-function-calling-v2-ep1-lora
This model is a fine-tuned version of cognitivecomputations/dolphin-2_6-phi-2 on the simple-function-calling-v2_convert dataset that I converted for llama_factory https://huggingface.co/datasets/Yhyu13/glaive-function-calling-v2-llama-factory-convert, but with a subset of only the first 1000 data entries. It achieves the following results on the evaluation set:
- Loss: 0.3524
Training script is availbale at ./scripts/local_ft_phi2_fn.sh)
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
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: float16
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 5e-05
- train_batch_size: 1
- eval_batch_size: 1
- seed: 42
- distributed_type: multi-GPU
- num_devices: 2
- gradient_accumulation_steps: 4
- total_train_batch_size: 8
- total_eval_batch_size: 2
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- num_epochs: 1.0
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 0.3453 | 1.0 | 376 | 0.3524 |
Framework versions
- PEFT 0.7.0
- Transformers 4.36.2
- Pytorch 2.1.1+cu121
- Datasets 2.14.7
- Tokenizers 0.15.0