Downloads · 30 days
0
oracle_baseline_d24_fp8
kaushikreddyxyz/oracle_baseline_d24_fp8
oracle_baseline_d24_fp8 is a machine learning model from kaushikreddyxyz. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for pytorch.
What: nanochat d24 base-model checkpoints (every 2000 steps + final step 8352; gzipped model.pt.gz + meta.json, optimizer state included) — negative control, no oracle injection, value embeddings LEARNED (stock nanoch…
Access
Public
Updated Jul 8, 2026
Repo size
43.2 GB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.gz43.2 GB · 100%
At a glance
- Library
- pytorch
- Access
- Public
- Created
- Jun 26, 2026
- Updated
- Jul 8, 2026
- SHA
- b8006646
Model card
From the Hugging Face model README
oracle_baseline_d24_fp8
What: nanochat d24 base-model checkpoints (every 2000 steps + final step 8352; gzipped model_*.pt.gz + meta_*.json, optimizer state included) — negative control, no oracle injection, value embeddings LEARNED (stock nanochat).
Trained on: karpathy/climbmix-400b-shuffle (first ~274 shards), 8.758B tokens = compute-optimal ratio 12, seed 1337, bf16 + fp8 GEMM (tensorwise), 8xH100. Script: nanochat/runs/oracle_runs/baseline.sh in oracle-encodings.
Result: CORE 0.2777 at step 8352.
Used for: baseline pair (VE arm) for oracle-encoding injection experiments; injected runs share seed/data order for matched comparison. Sibling no-VE control: oracle_baseline_noVE_d24_fp8 (the Stage-7 comparison arm, since value embeddings confound injection).
Status: active (frozen reference).
- Library
- pytorch
- Created
- Jun 26, 2026
- Updated
- Jul 8, 2026