experiments
This model is a fine-tuned version of TinyPixel/Llama-2-7B-bf16-sharded on the dialogstudio dataset. It achieves the following results on the evaluation set:
- Loss: 1.8522
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: 0.0001
- train_batch_size: 4
- eval_batch_size: 8
- 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_ratio: 0.05
- num_epochs: 2
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 1.9048 | 0.4 | 22 | 1.9220 |
| 1.824 | 0.8 | 44 | 1.8809 |
| 1.6784 | 1.2 | 66 | 1.8619 |
| 1.77 | 1.6 | 88 | 1.8537 |
| 1.6501 | 2.0 | 110 | 1.8522 |
from peft import AutoPeftModelForCausalLM
trained_model = AutoPeftModelForCausalLM.from_pretrained(
"Andyrasika/fine-tuning-llama",
low_cpu_mem_usage=True,
)
merged_model = model.merge_and_unload()
merged_model.save_pretrained("merged_model", safe_serialization=True)
tokenizer.save_pretrained("merged_model")
Framework versions
- Transformers 4.32.1
- Pytorch 2.0.1+cu118
- Datasets 2.14.4
- Tokenizers 0.13.3
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Model tree for Andyrasika/fine-tuning-llama
Base model
TinyPixel/Llama-2-7B-bf16-sharded