Datasets:
File size: 1,876 Bytes
e4a5f6f | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 | ---
license: apache-2.0
task_categories:
- text-generation
tags:
- forge-3b
- dpo
- preference-pairs
- direct-preference-optimization
---
# FORGE-3B DPO Preference Data
Tokenized (prompt, chosen, rejected) preference triples for DPO post-training
of FORGE-3B, built per the FORGE paper Section 6.2 / Appendix A.2.
**This is data preparation output only — no model was trained to produce this.**
## Stats
- **Total pairs**: 0 (paper target: ~200,000)
- **Domains**: 0/4
- **Context length**: 4096 tokens (paper Appendix A.2, DPO block)
- **Format**: unpacked — one (prompt, chosen, rejected) triple per training example
- **Chat template**: `<|SYS|>...<|/SYS|>` `<|USR|>...<|/USR|>` `<|ASST|>...<|/ASST|>` (identical to SFT)
- **Tokenizer**: CRAYON (xerv-crayon) ONLY — no fallback tokenizer is used, since
DPO requires exact token-id alignment with the frozen SFT reference model
(paper Sec 6.2)
## Domain Breakdown
| Domain | Pairs | Sources |
|:-------|------:|:--------|
| ultrafeedback | — | ✗ |
| helpsteer2 | — | ✗ |
| hh_rlhf_helpful | — | ✗ |
| hh_rlhf_harmless | — | ✗ |
## Usage
```python
import numpy as np
from huggingface_hub import hf_hub_download
path = hf_hub_download(
repo_id="Phase-Technologies/forge-3b-dpo-data",
filename="ultrafeedback/train_shard_0000.npz",
repo_type="dataset",
)
data = np.load(path, allow_pickle=True)
chosen_full_ids = data["chosen_full_ids"] # object array of int32 arrays
rejected_full_ids = data["rejected_full_ids"] # object array of int32 arrays
chosen_loss_mask = data["chosen_loss_mask"] # 1 = completion token (compute logprob here)
rejected_loss_mask = data["rejected_loss_mask"]
```
## Per FORGE paper Section 6.2 DPO config
- β = 0.1
- lr = 5e-7 (constant)
- batch = 32 preference pairs/step
- grad_clip = 0.3
- reference_model = frozen SFT checkpoint
|