<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
codellama-13b-personal-copilot-fa2-patch
This model is a fine-tuned version of codellama/CodeLlama-13b-hf on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.3810
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0003
- train_batch_size: 4
- eval_batch_size: 4
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 30
- training_steps: 2000
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 0.4724 | 0.05 | 100 | 0.4802 |
| 0.4237 | 0.1 | 200 | 0.4669 |
| 0.4344 | 0.15 | 300 | 0.4486 |
| 0.8711 | 0.2 | 400 | 0.4419 |
| 0.4341 | 0.25 | 500 | 0.4355 |
| 0.3828 | 0.3 | 600 | 0.4329 |
| 0.5729 | 0.35 | 700 | 0.4256 |
| 0.3516 | 0.4 | 800 | 0.4176 |
| 0.4342 | 0.45 | 900 | 0.4026 |
| 0.4593 | 0.5 | 1000 | 0.3957 |
| 0.2775 | 0.55 | 1100 | 0.3915 |
| 0.3277 | 0.6 | 1200 | 0.3867 |
| 0.301 | 0.65 | 1300 | 0.3834 |
| 0.6374 | 0.7 | 1400 | 0.3837 |
| 0.3582 | 0.75 | 1500 | 0.3822 |
| 0.2869 | 0.8 | 1600 | 0.3814 |
| 0.3958 | 0.85 | 1700 | 0.3812 |
| 0.3048 | 0.9 | 1800 | 0.3810 |
| 0.2705 | 0.95 | 1900 | 0.3810 |
| 0.3555 | 1.0 | 2000 | 0.3810 |
Framework versions
- Transformers 4.35.0.dev0
- Pytorch 2.1.0+cu118
- Datasets 2.14.5
- Tokenizers 0.14.0
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: bfloat16
Framework versions
- PEFT 0.6.0.dev0
