ChrisMcCormick
/

decoderstack-d24

Model card Files Files and versions

decoderstack-d24

53.1 GB

Ctrl+K

Ctrl+K

1 contributor

History: 19 commits

ChrisMcCormick's picture

ladder-fmt05 r233: the fix-ladder's working RL recipe

5913dab verified about 16 hours ago

base_checkpoints
meta_005568.json: add user_config so chat_sft inherits the run's LRs 1 day ago
chatrl_checkpoints
ladder-fmt05 r233: the fix-ladder's working RL recipe about 16 hours ago
chatsft_checkpoints
Tool-free SFT of d24_decoderstack step 5568 (462 steps, val bpb 0.2732, ChatCORE 0.2527); model + meta + w1 optim shard about 21 hours ago
checkpoints
Upload folder using huggingface_hub 1 day ago
code
Add model card, meta.json, logs, tokenizer, training source, and nanochat converter 1 day ago
logs
Add model card, meta.json, logs, tokenizer, training source, and nanochat converter 1 day ago
runs
ep1-ppr32: full-epoch stock-algorithm RL — collapse by r140; rollouts + metrics about 20 hours ago
tokenizer
tokenizer: keep the pre-fix token_bytes so published bpb numbers stay reproducible about 23 hours ago
.gitattributes

1.52 kB
initial commit 1 day ago
README.md

17.1 kB
Model card: correct the recalibration — leaderboard is best-fit only, so the like-for-like gap is +0.019; both checkpoints now measured under both packings about 23 hours ago
convert_ckpt_to_nanochat.py

26.3 kB
Converter now emits nanochat optimizer shards for an SFT warm-start 1 day ago
test_convert_ckpt_to_nanochat.py

13.9 kB
Converter now emits nanochat optimizer shards for an SFT warm-start 1 day ago