metadata
library_name: transformers
license: llama3
base_model: meta-llama/Meta-Llama-3-8B-Instruct
tags:
- alignment-handbook
- generated_from_trainer
datasets:
- >-
simonycl/Meta-Llama-3-8B-Instruct_ultrafeedback-Meta-Llama-3-8B-Instruct-annotate-start-0-end-1.0-judge-5
model-index:
- name: llama-3-8b-instruct-agg-judge
results: []
llama-3-8b-instruct-agg-judge
This model is a fine-tuned version of meta-llama/Meta-Llama-3-8B-Instruct on the simonycl/Meta-Llama-3-8B-Instruct_ultrafeedback-Meta-Llama-3-8B-Instruct-annotate-start-0-end-1.0-judge-5 dataset. It achieves the following results on the evaluation set:
- Loss: 0.6390
- Rewards/chosen: -1.0532
- Rewards/rejected: -1.3037
- Rewards/accuracies: 0.6057
- Rewards/margins: 0.2506
- Logps/rejected: -280.7787
- Logps/chosen: -256.8969
- Logits/rejected: -1.4905
- Logits/chosen: -1.5260
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: 5e-07
- train_batch_size: 1
- eval_batch_size: 2
- seed: 42
- distributed_type: multi-GPU
- num_devices: 4
- gradient_accumulation_steps: 16
- total_train_batch_size: 64
- total_eval_batch_size: 8
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 1
Training results
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.6265 | 0.4264 | 400 | 0.6455 | -0.7831 | -0.9487 | 0.6504 | 0.1655 | -245.2767 | -229.8961 | -1.3679 | -1.4091 |
| 0.6053 | 0.8529 | 800 | 0.6390 | -1.0532 | -1.3037 | 0.6057 | 0.2506 | -280.7787 | -256.8969 | -1.4905 | -1.5260 |
Framework versions
- Transformers 4.45.1
- Pytorch 2.4.0+cu121
- Datasets 2.21.0
- Tokenizers 0.20.0