Oct 15 20 days ago 19:59 Add speculators training note 00:11 Add USE_POLICY.md and license name/link to the model card USE_POLICY.md copied verbatim from meta-models/Muse-Spark-Open-816B-A42B@1a31ebd1371f16c90e705abe3481e99a479c9527. 00:06 Add LICENSE.md from meta-models/Muse-Spark-Open-816B-A42B Copied verbatim from meta-models/Muse-Spark-Open-816B-A42B@1a31ebd1371f16c90e705abe3481e99a479c9527. Oct 14 21 days ago 17:46 Update README.md 05:50 Update acceptance rates for the rc4 target (e519b25b) Re-measured against meta-models/Muse-Spark-open-816B-A42B-NVFP4@e519b25b (rc4 weights, default xhigh reasoning effort) with full GSM8K/MBPP/SWE-bench Pro and SPEED-Bench low-entropy; AA-LCR not re-measured and removed. Draft weights unchanged. 00:04 Update README.md Oct 13 22 days ago 00:47 Update README.md Sep 30 00:46 Remove internal MQA reference from README 00:46 Remove MQA accuracy line from README 00:45 Trim README: drop training details, condense architecture and acceptance notes 00:39 Add brief README: DSpark architecture and step-3600 acceptance results Sep 25 17:30 fresh_fp32_20260920 training3 final draft: step 3600 (from scratch, 3600 updates), trained against Muse-Spark-1.2-816B-A42B-NVFP4-open; full-block acceptance PASSED (job 11304) Sep 20 18:50 Remove provenance/docs files; keep config.json and model.safetensors only 18:39 warm3605 training3 final draft: step 2324 (4 epochs), trained against Muse-Spark-1.2-816B-A42B-NVFP4-open; full-block acceptance PASSED (job 11141) Sep 13 17:09 Upload train_command.txt with huggingface_hub 17:09 Upload export_manifest.json with huggingface_hub 17:09 Muse Spark DSpark ce_scale10 final step 3605 (four epochs); full block-verification acceptance PASSED job 10804 Sep 11 02:33 Upload verified Muse Spark DSpark step-2000 checkpoint Sep 9 23:54 Upload verified Muse Spark DSpark step-900 checkpoint 23:51 initial commit 
