Downloads · 30 days
17
24% of all-time downloads
pakphum/babylm2026-pl-ru-baseline
babylm2026-pl-ru-baseline is a machine learning model from pakphum. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as mit.
LLaMA-style causal language model trained for BabyLM 2026 with IPA phonological feature augmentation.
Downloads · 30 days
17
24% of all-time downloads
All-time downloads
70
Public
Parameters
51.8M
207 MB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors207 MB · 99%
From the Hugging Face model README
LLaMA-style causal language model trained for BabyLM 2026 with IPA phonological feature augmentation.
baseline — gated fusion of panphon IPA feature vectors at the embedding layer| Variant | Description |
|---|---|
baseline | Token embeddings only |
ipa_add | embed + W·ipa |
ipa_gate | embed + σ(W_g·ipa) ⊙ (W·ipa) |
ipa_full | Gated fusion + MSE auxiliary reconstruction loss |
from transformers import AutoConfig, AutoModelForCausalLM
import torch
config = AutoConfig.from_pretrained("pakphum/babylm2026-pl-ru-baseline", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("pakphum/babylm2026-pl-ru-baseline", trust_remote_code=True)
model.eval()
# input_ids from your tokenizer
input_ids = torch.tensor([[1, 2, 3, 4]])
# Optional: supply IPA vectors [B, T, ipa_dim=24]. Zeros = no phonological signal.
ipa_vectors = torch.zeros(1, 4, config.ipa_dim)
with torch.no_grad():
out = model(input_ids, ipa_vectors=ipa_vectors)
# out.logits: [B, T, vocab_size]
BabyLM 2026 shared task.