timorobrecht/babylm-qwen2-33m-hybrid-softmax

Pinyin-Code Causal LM

This repository contains a custom Transformers causal language model. External evaluation repositories should load it with trust_remote_code=True and use the causal backend.

Dependencies

Install the runtime dependencies before loading the model:

pip install torch transformers safetensors pypinyin jieba

This export uses the repository's hybrid tokenizer and does not require SentencePiece. pypinyin is required for raw Mandarin-to-pinyin preprocessing. jieba is required when use_jieba is true; this export was created with use_jieba=true.

Loading

from transformers import (
    AutoConfig,
    AutoModel,
    AutoModelForCausalLM,
    AutoModelForSequenceClassification,
    AutoTokenizer,
)

model_path = "PATH_OR_REPO_ID"

config = AutoConfig.from_pretrained(model_path, trust_remote_code=True)
tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
base_model = AutoModel.from_pretrained(model_path, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(model_path, trust_remote_code=True)
classifier = AutoModelForSequenceClassification.from_pretrained(
    model_path,
    trust_remote_code=True,
    num_labels=3,
)

Evaluation

Configure external evaluators with:

• —model path: this local folder or Hugging Face repo ID
• —backend: causal
• —trust remote code: enabled

The hybrid tokenizer accepts preprocessed pinyin-code text through standard calls such as tokenizer(text), tokenizer(text, add_special_tokens=False), and tokenizer(texts, padding=True, truncation=True, return_tensors="pt"). The model supports output_hidden_states=True for representation extraction tasks.

This export sets patch_pathlib_utf8_open=true in config.json. When loaded with trust_remote_code=True, the config installs a narrow Windows compatibility shim so later text-mode Path.open("r") calls without an explicit encoding default to UTF-8. Set PINYIN_CODE_DISABLE_UTF8_OPEN_PATCH=1 before loading the model to disable that shim.

Export metadata:

• —tokenizer_kind: hybrid
• —transliteration: pinyin-code
• —use_jieba: true