You can deploy this model and start earning money today!

siddharthmb/
2026.TA.features_2026.TA.gemma2_2b_tc8192_decb_l1w0.001_tarbb_lb2.0_ln1_dr20000_lr_ha43539c2ed71

Name: 2026.TA.features_2026.TA.gemma2_2b_tc8192_decb_l1w0.001_tarbb_lb2.0_ln1_dr20000_lr_ha43539c2ed71
Availability: InStock

circuit-tracer

transcoder-adapters

feature-cache

siddharthmb/2026.transcoder-adapters.lmsys-chat-1m-splits

science-of-finetuning/fineweb-1m-sample

siddharthmb/2026.TA.gemma2_2b_tc8192_decb_l1w0.001_tarbb_lb2.0_ln1_dr20000_lr8e-04_bs4_sl14793860

finetune

Views

No views yet

2026.TA.features_2026.TA.gemma2_2b_tc8192_decb_l1w0.001_tarbb_lb2.0_ln1_dr20000_lr_ha43539c2ed71

Packed circuit-tracer feature cache produced by analysis.features.collect_feature_activations.

Status

Complete.

Circuit-Tracer Usage

Use this repo ID as the scan name: siddharthmb/2026.TA.features_2026.TA.gemma2_2b_tc8192_decb_l1w0.001_tarbb_lb2.0_ln1_dr20000_lr_ha43539c2ed71. The frontend reads packed features from features/index.json.gz and features/layer_N.bin.

Collection Inputs

Model: siddharthmb/2026.TA.gemma2_2b_tc8192_decb_l1w0.001_tarbb_lb2.0_ln1_dr20000_lr8e-04_bs4_sl14793860
Config: feature_collection_config.json
Datasets:
- siddharthmb/2026.transcoder-adapters.lmsys-chat-1m-splits
- science-of-finetuning/fineweb-1m-sample

Summary

Total tokens: 109689219
Tokens per domain: {'fineweb': 62706474, 'chat': 46982745}

per_feature_json.tar.zst

Archive of the 181,192 legacy per-feature dashboard JSON files (the local features/ directory, one JSON per feature). features/ in this repo is the packed circuit-tracer cache used by the overlay/dashboard; the archive preserves the older per-feature format. Extract with tar -I zstd -xf per_feature_json.tar.zst.