Sparse Coders - a dimred Collection

dimred 's Collections

updated Jul 26

A collection of sparse coders trained on Pythia + some others. Notes are structured as: {model}_{coder-type}_{coder-arch}_{width}

Upvote

EleutherAI/Pythia-160m-ST-k128-4k

Updated Jul 25 • 30

Note pythia160m_st_topk_4k
EleutherAI/sae-llama-3.1-8b-64x

Updated Jul 22 • 60 • 17

Note llama3.1_sae_topk_64x
EleutherAI/sae-llama-3-8b-32x

Updated Jul 9, 2024 • 42

Note llama3_sae_topk_32x
EleutherAI/sae-llama-3-8b-32x-v2

Updated Jul 16, 2024 • 17

Note llama3_sae_topk_32x
EleutherAI/sae-pythia-160m-32k

Updated Aug 5, 2024 • 1

Note pythia160m_sae_topk_32k
EleutherAI/sae-llama-3.1-8b-32x

Updated Aug 6, 2024 • 13

Note llama3.1_sae_topk_32x
EleutherAI/sae-pythia-70m-32k

Updated Aug 5, 2024

Note pythia70m_sae_topk_32k
EleutherAI/sae-pythia-410m-65k

Updated Aug 5, 2024

Note pythia410m_sae_topk_32k
EleutherAI/sae-pythia-70m-deduped-32k

Updated Aug 15, 2024

Note pythia70m_sae_topk_32k
EleutherAI/sae-pythia-160m-deduped-32k

Updated Aug 15, 2024

Note pythia160m_sae_topk_32k
EleutherAI/sae_overlap

Updated Jan 20
EleutherAI/sae-DeepSeek-R1-Distill-Qwen-1.5B-65k

Updated Jan 26 • 37 • 7

Note qwen1.5b_sae_topk_65k
EleutherAI/sae-Llama-3.2-1B-131k

Updated Jan 26 • 30 • 1

Note llama3.1_sae_topk_131k
EleutherAI/sae-SmolLM2-135M-64x

Updated Feb 3 • 38 • 1

Note smollm2_sae_topk_64x
EleutherAI/sae-SmolLM2-135M-64x-random

Updated Feb 3 • 33
EleutherAI/Pythia-160m-SAE-k64-131k

Updated Jul 25 • 30

Note pythia160m_sae_topk_131k
EleutherAI/Pythia-160m-SAE-k64-65k

Updated Jul 25 • 28

Note pythia160m_sae_topk_65k
EleutherAI/Pythia-160m-SAE-k64-32k

Updated Jul 25 • 31

Note pythia160m_sae_topk_32k
EleutherAI/Pythia-160m-SAE-k128-32k

Updated Jul 25 • 36

Note pythia160m_sae_topk_32k
EleutherAI/Pythia-160m-SAE-k64-4k

Updated Jul 25 • 31

Note pythia160m_sae_topk_4k
EleutherAI/Pythia-160m-SAE-k128-768

Updated Jul 25 • 29

Note pythia160m_sae_topk_768
EleutherAI/Pythia-160m-SAE-k64-768

Updated Jul 25 • 30

Note pythia160m_sae_topk_768
EleutherAI/Pythia-160m-SAE-k32-768

Updated Jul 25 • 30

Note pythia160m_sae_topk_768
EleutherAI/Pythia-160m-SST-k32-768

Updated Jul 25 • 33

Note pythia160m_sst_topk_768
EleutherAI/Pythia-160m-SST-k32-32k

Updated Jul 25 • 33

Note pythia160m_sst_topk_32k
EleutherAI/Pythia-160m-SST-k64-4k

Updated Jul 25 • 35

Note pythia160m_sst_topk_4k
EleutherAI/Pythia-160m-SST-k64-32k

Updated Jul 25 • 36

Note pythia160m_sst_topk_32k
EleutherAI/Pythia-160m-SST-k64-65k

Updated Jul 25 • 35

Note pythia160m_sst_topk_65k
EleutherAI/Pythia-160m-SST-k128-32k

Updated Jul 25 • 30

Note pythia160m_sst_topk_32k
EleutherAI/Pythia-160m-SST-k32-65k

Updated Jul 25 • 33

Note pythia160m_sst_topk_65k
EleutherAI/Pythia-160m-SST-k32-131k

Updated Jul 25 • 32

Note pythia160m_sst_topk_131k
EleutherAI/Pythia-160m-SST-k128-131k

Updated Jul 25 • 35

Note pythia160m_sst_topk_131k
EleutherAI/Pythia-160m-ST-k128-32k

Updated Jul 25 • 35

Note pythia160m_st_topk_32k
EleutherAI/Pythia-160m-ST-k32-131k

Updated Jul 25 • 31

Note pythia160m_st_topk_131k
EleutherAI/Pythia-160m-ST-k32-4k

Updated Jul 25 • 32

Note pythia160m_st_topk_4k
EleutherAI/Pythia-160m-ST-k32-768

Updated Jul 25 • 32

Note pythia160m_st_topk_768
EleutherAI/Pythia-160m-ST-k64-768

Updated Jul 25 • 33

Note pythia160m_st_topk_768
EleutherAI/Pythia-160m-ST-k128-131k

Updated Jul 25 • 31

Note pythia160m_st_topk_131k
EleutherAI/Pythia-160m-ST-k64-131k

Updated Jul 25 • 30

Note pythia160m_st_topk_131k
EleutherAI/Pythia-160m-ST-k64-4k

Updated Jul 25 • 31

Note pythia160m_st_topk_4k
EleutherAI/Pythia-160m-ST-k64-65k

Updated Jul 25 • 32

Note pythia160m_st_topk_65k
canrager/saebench_gemma-2-2b_width-2pow12_date-1109

Updated Nov 9, 2024

Note gemma2_sae_mult_4k
canrager/saebench_gemma-2-2b_width-2pow14_date-1109

Updated Nov 10, 2024

Note gemma2_sae_mult_16k
canrager/saebench_gemma-2-2b_width-2pow16_date-1109

Updated Jan 9

Note gemma2_sae_mult_65k
canrager/saebench_gemma-2-2b_width-2pow14_date-1205

Updated Dec 6, 2024

Note gemma2_sae_mult_16k
canrager/saebench_gemma-2-2b_width-2pow14_date-1205_buffer

Updated Dec 7, 2024
canrager/saebench_gemma-2-2b_width-2pow14_date-0107

Updated Feb 4 • 2

Note gemma2_sae_mult_16k
canrager/saebench_gemma-2-2b_width-2pow12_date-0107

Updated Jan 13 • 1

Note gemma2_sae_mult_4k
canrager/saebench_gemma-2-2b_width-2pow16_date-0107

Updated Feb 4 • 1

Note gemma2_sae_mult_65k
adamkarvonen/saebench_pythia-160m-deduped_width-2pow12_date-0108

Feature Extraction • Updated Mar 14

Note pythia160m_sae_mult_4k
adamkarvonen/saebench_pythia-160m-deduped_width-2pow14_date-0108

Feature Extraction • Updated Mar 14

Note pythia160m_sae_mult_16k
adamkarvonen/saebench_pythia-160m-deduped_width-2pow16_date-0108

Feature Extraction • Updated Jun 9

Note pythia160m_sae_mult_65k
adamkarvonen/saebench_gemma-2-2b_width-2pow12_date-0108

Updated Feb 4

Note gemma2_sae_mult_4k

Upvote