Upup-ashton-wang's group workspace
SAE Hookpoints Ablation
What makes this group special?
Tags
Resa-STILL-27th-Layer
Notes
Author
State
Finished
Start time
May 9th, 2025 8:57:57 PM
Runtime
22m 12s
Tracked hours
22m 11s
Run path
upup-ashton-wang-usc/Resa/hvurump1
OS
Linux-4.18.0-553.22.1.el8_10.x86_64-x86_64-with-glibc2.28
Python version
CPython 3.10.16
Command
/home1/shangsha/workspace/reasoning/reasoning-sae/./scripts/training/run_distill_model_accelerate.py --config ./recipes/DeepSeek-R1-Distill-Qwen-1.5B/grpo/distill_curated_still.yaml --host_model_checkpoint checkpoint-2000 --student_model_name DeepSeek-R1-Distill-Qwen-1.5B --distill_type sft_r1_distill --sae_hookpoint model.layers.26
System Hardware
CPU count | 32 |
Logical CPU count | 32 |
GPU count | 2 |
GPU type | NVIDIA A40 |
W&B CLI Version
0.19.9
Config
Config parameters are your model's inputs. Learn more
- {} 18 keys▶
- "DeepSeek-R1-Distill-Qwen-1.5B"
- 1
- "sft_r1_distill"
- "checkpoint-2000"
- "curated_still"
- "grpo"
- 0.000001
- 1
- 128
- 0.05
- 32
- [] 7 items▶
- 2
- "model.layers.26"
- "sae-DeepSeek-R1-Distill-Qwen-1.5B-65k"
- 500
- 42
- "DeepSeek-R1-Distill-Qwen-1.5B"
Summary
Summary metrics are your model's outputs. Learn more
- {} 5 keys▶
- 1
- 238.4261049723757
- 0.000001
- 107
- 320
Artifact Outputs
This run produced these artifacts as outputs. Total: 1. Learn more
Type
Name
Consumer count
Loading...