Qwen3-8B BT — Qwen science data
Qwen3-8B sequence-scoring checkpoint trained in BonVoyage with a BT pairwise objective.
Training details
- Base model:
Qwen/Qwen3-8B - Training data:
graf/qwen_4b_science_mix_train - Validation data:
graf/qwen_4b_science_sciknowsci_val - Validation examples: 512
- Learning rate:
1e-5 - Final checkpoint: epoch 103 of 104
- Output head: one scalar (
num_labels=1) - Weights: BF16 safetensors
- Tokenizer: the tokenizer saved with the training run;
pad_token_id=151643
Experiment: science_4b_mix_bt_8b_solid-bt-a577c3ef-1-104-on.
- Downloads last month
- -