CoolFace
Modelpublic

mlnomad/yatnmn-softplus-ca-d12-chinchilla-261M-pytorch

sourceHugging Faceapache-2.0updated 5mo agoView on Hugging Face
0likes32downloads
12 commits on main
0074b2b5mo ago

Fix: lazy non-persistent RoPE buffers — HF from_pretrained's meta-init was leaving persistent rope_cos/rope_sin uninitialised, producing NaN logits. Buffers are now computed on first forward.

mlnomad
d169b065mo ago

Fix: lazy non-persistent RoPE buffers — HF from_pretrained's meta-init was leaving persistent rope_cos/rope_sin uninitialised, producing NaN logits. Buffers are now computed on first forward.

mlnomad
65ddab75mo ago

Fix: lazy non-persistent RoPE buffers — HF from_pretrained's meta-init was leaving persistent rope_cos/rope_sin uninitialised, producing NaN logits. Buffers are now computed on first forward.

mlnomad
0d05db35mo ago

Fix: lazy non-persistent RoPE buffers — HF from_pretrained's meta-init was leaving persistent rope_cos/rope_sin uninitialised, producing NaN logits. Buffers are now computed on first forward.

mlnomad
5f5b0c25mo ago

Fix: lazy non-persistent RoPE buffers — HF from_pretrained's meta-init was leaving persistent rope_cos/rope_sin uninitialised, producing NaN logits. Buffers are now computed on first forward.

mlnomad
623c2ec5mo ago

Fix: lazy non-persistent RoPE buffers — HF from_pretrained's meta-init was leaving persistent rope_cos/rope_sin uninitialised, producing NaN logits. Buffers are now computed on first forward.

mlnomad
5884b225mo ago

Fix: lazy non-persistent RoPE buffers — HF from_pretrained's meta-init was leaving persistent rope_cos/rope_sin uninitialised, producing NaN logits. Buffers are now computed on first forward.

mlnomad
5ad3af95mo ago

Fix: lazy non-persistent RoPE buffers — HF from_pretrained's meta-init was leaving persistent rope_cos/rope_sin uninitialised, producing NaN logits. Buffers are now computed on first forward.

mlnomad
aeb243f5mo ago

Fix: lazy non-persistent RoPE buffers — HF from_pretrained's meta-init was leaving persistent rope_cos/rope_sin uninitialised, producing NaN logits. Buffers are now computed on first forward.

mlnomad
73d1b525mo ago

Upload yatnmn_gpt.py with huggingface_hub

mlnomad
83bd6f75mo ago

yatnmn-softplus-ca-d12-chinchilla-261M-pytorch — constant alpha ablation

mlnomad
6a895dd5mo ago

initial commit

mlnomad