<!-- 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. -->
reddit_gen_final
This model is a fine-tuned version of microsoft/DialoGPT-small on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 2.5050
- Rouge: {'rouge1': 0.5318334167452433, 'rouge2': 0.3266503490464716, 'rougeL': 0.4940196552424935, 'rougeLsum': 0.49965823775029017}
- Perplexity: 810.2161
- Bleu: {'bleu': 0.3233116246700081, 'precisions': [0.5456588886510291, 0.3399931653275477, 0.273607307447275, 0.2384403661808989], 'brevity_penalty': 0.9747575251310703, 'length_ratio': 0.975070821529745, 'translation_length': 130796, 'reference_length': 134140}
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: 0.001
- train_batch_size: 1024
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 32
- total_train_batch_size: 32768
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 100
- training_steps: 1077
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Rouge | Perplexity | Bleu |
|---|---|---|---|---|---|---|
| 3.9872 | 18.02 | 320 | 3.1407 | {'rouge1': 0.4236605668840905, 'rouge2': 0.17676647154634773, 'rougeL': 0.37369585565755137, 'rougeLsum': 0.38069226779493875} | 1007.2803 | {'bleu': 0.1795660368641941, 'precisions': [0.4476203532341719, 0.1996849620846328, 0.12947506323794772, 0.097952801903731], 'brevity_penalty': 0.9786100068791068, 'length_ratio': 0.9788355449530342, 'translation_length': 131301, 'reference_length': 134140} |
| 3.0112 | 37.02 | 640 | 2.6693 | {'rouge1': 0.5006690963461402, 'rouge2': 0.2845737029774397, 'rougeL': 0.4598926127632702, 'rougeLsum': 0.46623659707701914} | 891.6387 | {'bleu': 0.28259351848586683, 'precisions': [0.5153005174673647, 0.2977358252901072, 0.22869830241856198, 0.19400129812455164], 'brevity_penalty': 0.9838352619991267, 'length_ratio': 0.9839645146861488, 'translation_length': 131989, 'reference_length': 134140} |
| 2.5776 | 56.02 | 960 | 2.5050 | {'rouge1': 0.5318334167452433, 'rouge2': 0.3266503490464716, 'rougeL': 0.4940196552424935, 'rougeLsum': 0.49965823775029017} | 810.2161 | {'bleu': 0.3233116246700081, 'precisions': [0.5456588886510291, 0.3399931653275477, 0.273607307447275, 0.2384403661808989], 'brevity_penalty': 0.9747575251310703, 'length_ratio': 0.975070821529745, 'translation_length': 130796, 'reference_length': 134140} |
Framework versions
- Transformers 4.28.1
- Pytorch 1.13.1+cu117
- Datasets 2.10.1
- Tokenizers 0.13.2
