ojaffe/20260411-190341-align-qwen-0d3d-2026-04-12-018-ob-correction
1
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

初始化项目，由ModelHub XC社区提供模型

Browse Source 
Model: ojaffe/20260411-190341-align-qwen-0d3d-2026-04-12-018-ob-correction
Source: Original Platform
This commit is contained in:
ModelHub XC
2026-04-22 00:21:01 +08:00
commit ad3d5ebd35
25 changed files with 304251 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

21
training_config.json Normal file
View File

@@ -0,0 +1,21 @@
{
"base_model": "/home/coder/experiments/2026-04-12-017-tqa-dpo",
"method": "OB_correction_DPO",
"beta": 0.03,
"lora_rank": 16,
"lora_alpha": 32,
"learning_rate": 3e-07,
"epochs": 2,
"batch_size": 4,
"grad_accum": 2,
"effective_batch_size": 8,
"max_seq_length": 512,
"target_modules": [
"q_proj",
"k_proj",
"v_proj",
"o_proj"
],
"train_time_seconds": 1212.3,
"num_training_examples": 8847
}