Open-weight model
npov_PERL_organic_gemma-4-E4B-it_S130104_epo0_5_lr1_1e-05_beta0_078_r8_2610062017
by Leonardo Martins bianco leobianco/npov_PERL_organic_gemma-4-E4B-it_S130104_epo0_5_lr1_1e-05_beta0_078_r8_2610062017
npov_PERL_organic_gemma-4-E4B-it_S130104_epo0_5_lr1_1e-05_beta0_078_r8_2610062017 is an open-weight model from Leonardo Martins bianco. Its published files total 74.3 MB.
This model is a fine-tuned version of google/gemma-4-E4B-it. It has been trained using TRL. This model was trained with RLOO, a method introduced in Back to Basics: Revisiting REINFORCE-Style Optimization for Learning from Human Feedback in LLMs.
Model Card
This model is a fine-tuned version of google/gemma-4-E4B-it. It has been trained using TRL. This model was trained with RLOO, a method introduced in Back to Basics: Revisiting REINFORCE-Style Optimization for Learning from Human Feedback in LLMs.
Excerpt from the card by Leonardo Martins bianco.
Identity and Version
- Repository
- leobianco/npov_PERL_organic_gemma-4-E4B-it_S130104_epo0_5_lr1_1e-05_beta0_078_r8_2610062017
- Publisher
- Leonardo Martins bianco
- Task
- Not stated by the source
- Modality
- Other
- Library
- transformers
- Parameters
- Not stated by the source
- Languages
- trl
- Revision
- 19450dbe2b92251482673de73d62d1e1e3001ddb
- First published
- 2026-10-06
- Last updated
- 2026-10-07
Files and Weights
23 files, 74.3 MB in total. The weights are 4 files totalling 9.1 MB in bin, safetensors.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| adapter_model.safetensors | Weights | 4.6 MB | 7c184a16a108 |
| best_continual_eval_checkpoint/adapter_model.safetensors | Weights | 4.6 MB | 3f5784d7caac |
| best_continual_eval_checkpoint/training_args.bin | Weights | 8.5 KB | 1689c1d28b89 |
| training_args.bin | Weights | 8.5 KB | f205846890ac |
| adapter_config.json | Configuration | 1.1 KB | — |
| best_continual_eval_checkpoint/adapter_config.json | Configuration | 1.1 KB | — |
| best_continual_eval_checkpoint/continual_eval_best_checkpoint.json | Configuration | 159 B | — |
| best_continual_eval_checkpoint/trainer_state.json | Configuration | 410.8 KB | — |
| continual_eval_history.json | Configuration | 19.3 KB | — |
| continual_eval_status.json | Configuration | 787 B | — |
| README.md | Documentation | 2.7 KB | — |
| best_continual_eval_checkpoint/README.md | Documentation | 5.2 KB | — |
| best_continual_eval_checkpoint/chat_template.jinja | Other | 18.6 KB | — |
| best_continual_eval_checkpoint/wandb_run_id.txt | Other | 8 B | — |
| chat_template.jinja | Other | 18.6 KB | — |
| pareto_plots/pareto_hallucination_vs_reward_hacking_quality.png | Other | 148.3 KB | a8bb6a76d223 |
| pareto_plots/pareto_hallucination_vs_reward_hacking_rate.png | Other | 151.1 KB | 6be71068b496 |
| wandb_run_id.txt | Other | 8 B | — |
| .gitattributes | Repository | 1.8 KB | — |
| best_continual_eval_checkpoint/tokenizer.json | Tokenizer | 32.2 MB | cc8d3a0ce364 |
| best_continual_eval_checkpoint/tokenizer_config.json | Tokenizer | 3.7 KB | — |
| tokenizer.json | Tokenizer | 32.2 MB | cc8d3a0ce364 |
| tokenizer_config.json | Tokenizer | 3.7 KB | — |
License and Download
- License
- Not stated by the source
- Access
- Open weights, no gate
- Download size
- 9.1 MB
Released by Leonardo Martins bianco through its official repository on Hugging Face.
Built From
- Derived from google/gemma-4-E4B-it
- Described by arXiv:2402.14740
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 9.1 MB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.