wvnvwn/llama2-7b-chat-lr5e-5-boolq-lr5e-5-safeinstr0_1
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: wvnvwn/llama2-7b-chat-lr5e-5-boolq-lr5e-5-safeinstr0_1
Source: Original Platform
This commit is contained in:
ModelHub XC
2026-09-07 13:54:17 +08:00
commit 0800a7fb22
13 changed files with 277640 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

77
logs/finetune_boolq_20260718_221735.log Normal file
View File

@@ -0,0 +1,77 @@
2026-07-18 22:17:35 - __main__ - INFO -
======================================================================
2026-07-18 22:17:35 - __main__ - INFO - 🚀 Full Parameter boolq Fine-tuning (SN-Tuned Model)
2026-07-18 22:17:35 - __main__ - INFO - ======================================================================
2026-07-18 22:17:35 - __main__ - INFO - Log file: /NHNHOME/26msit001_A/BASE/edge_ai_lab/jongbokwon/minseong_results/checkpoints/llama-2-7b-chat-hf/baselines/safeinstr/ratio_0_1/boolq/logs/finetune_boolq_20260718_221735.log
2026-07-18 22:17:35 - __main__ - INFO - ⚙️ Configuration:
2026-07-18 22:17:35 - __main__ - INFO - ├─ SN-Tuned model: /NHNHOME/26msit001_A/BASE/edge_ai_lab/jongbokwon/warp-safety/outputs/chat_ssft/llama2_7b_chat_circuit_breakers
2026-07-18 22:17:35 - __main__ - INFO - ├─ Downstream task: boolq
2026-07-18 22:17:35 - __main__ - INFO - ├─ Input formatting: chat template
2026-07-18 22:17:35 - __main__ - INFO - ├─ Training samples: 0
2026-07-18 22:17:35 - __main__ - INFO - ├─ Batch size: 4
2026-07-18 22:17:35 - __main__ - INFO - ├─ Gradient accumulation: 4
2026-07-18 22:17:35 - __main__ - INFO - ├─ Epochs: 3
2026-07-18 22:17:35 - __main__ - INFO - ├─ Learning rate: 5e-05
2026-07-18 22:17:35 - __main__ - INFO - ├─ Weight decay: 0.01
2026-07-18 22:17:35 - __main__ - INFO - ├─ Optimizer: AdamW (torch)
2026-07-18 22:17:35 - __main__ - INFO - ├─ Warmup ratio: 0.1
2026-07-18 22:17:35 - __main__ - INFO - ├─ Max length: 1024
2026-07-18 22:17:35 - __main__ - INFO - ├─ Dtype: bf16
2026-07-18 22:17:35 - __main__ - INFO - └─ Output dir: /NHNHOME/26msit001_A/BASE/edge_ai_lab/jongbokwon/minseong_results/checkpoints/llama-2-7b-chat-hf/baselines/safeinstr/ratio_0_1/boolq
2026-07-18 22:17:35 - __main__ - INFO -
======================================================================
2026-07-18 22:17:35 - __main__ - INFO - [1/4] Loading Tokenizer
2026-07-18 22:17:35 - __main__ - INFO - ======================================================================
2026-07-18 22:17:35 - __main__ - INFO - Attempting to load tokenizer (local files only)...
2026-07-18 22:17:35 - __main__ - INFO - ✓ Tokenizer loaded from local files
2026-07-18 22:17:35 - __main__ - INFO - ✅ Tokenizer loaded successfully
2026-07-18 22:17:35 - __main__ - INFO - ├─ Tokenizer type: LlamaTokenizer
2026-07-18 22:17:35 - __main__ - INFO - ├─ Vocab size: 32000
2026-07-18 22:17:35 - __main__ - INFO - └─ Pad token: </s>
2026-07-18 22:17:35 - __main__ - INFO -
======================================================================
2026-07-18 22:17:35 - __main__ - INFO - [2/4] Loading Model (bf16)
2026-07-18 22:17:35 - __main__ - INFO - ======================================================================
2026-07-18 22:17:35 - __main__ - INFO - Attempting to load model (local files only)...
2026-07-18 22:17:42 - __main__ - INFO - ✓ Model loaded from local files
2026-07-18 22:17:42 - __main__ - INFO - ✅ Model loaded successfully
2026-07-18 22:17:42 - __main__ - INFO - ├─ Model size: 6.74B parameters
2026-07-18 22:17:42 - __main__ - INFO - ├─ Trainable: 6.74B (100.00%)
2026-07-18 22:17:42 - __main__ - INFO - ├─ Dtype: torch.bfloat16
2026-07-18 22:17:42 - __main__ - INFO - └─ Gradient checkpointing: Disabled
2026-07-18 22:17:42 - __main__ - INFO -
======================================================================
2026-07-18 22:17:42 - __main__ - INFO - [3/4] Loading boolq Dataset
2026-07-18 22:17:42 - __main__ - INFO - ======================================================================
2026-07-18 22:17:42 - __main__ - INFO - ✅ Datasets loaded
2026-07-18 22:17:42 - __main__ - INFO - ├─ Train: 9427 samples
2026-07-18 22:17:42 - __main__ - INFO -
======================================================================
2026-07-18 22:17:42 - __main__ - INFO - [3.5/4] Preprocessing Data
2026-07-18 22:17:42 - __main__ - INFO - ======================================================================
2026-07-18 22:17:44 - __main__ - INFO - ✅ Safety data mixed: 942 samples (ratio=0.1)
2026-07-18 22:17:44 - __main__ - INFO - Total training samples: 10369 (boolq 9427 + Safety 942)
2026-07-18 22:17:44 - __main__ - INFO -
======================================================================
2026-07-18 22:17:44 - __main__ - INFO - [4/4] Training with Trainer + AdamW
2026-07-18 22:17:44 - __main__ - INFO - ======================================================================
2026-07-18 22:17:44 - __main__ - INFO - Starting training...
2026-07-18 22:26:07 - __main__ - INFO -
======================================================================
2026-07-18 22:26:07 - __main__ - INFO - Saving Fine-tuned Model
2026-07-18 22:26:07 - __main__ - INFO - ======================================================================
2026-07-18 22:26:19 - __main__ - INFO - ✅ Fine-tuned model saved successfully to /NHNHOME/26msit001_A/BASE/edge_ai_lab/jongbokwon/minseong_results/checkpoints/llama-2-7b-chat-hf/baselines/safeinstr/ratio_0_1/boolq
2026-07-18 22:26:19 - __main__ - INFO - ✅ Config saved to: /NHNHOME/26msit001_A/BASE/edge_ai_lab/jongbokwon/minseong_results/checkpoints/llama-2-7b-chat-hf/baselines/safeinstr/ratio_0_1/boolq/finetune_config.json
2026-07-18 22:26:19 - __main__ - INFO -
======================================================================
2026-07-18 22:26:19 - __main__ - INFO - ✅ Fine-tuning Complete!
2026-07-18 22:26:19 - __main__ - INFO - ======================================================================