Zero-Shot Classification
Transformers
Safetensors
PyTorch
Turkish
English
tau
feature-extraction
turkish
decision-model
system-one
text-classification
natural-language-inference
calibrated-probabilities
long-context
qwen3.5
gated-deltanet
hybrid-attention
custom_code
Eval Results (legacy)
Instructions to use cisimcik/tau-4b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use cisimcik/tau-4b with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("zero-shot-classification", model="cisimcik/tau-4b", trust_remote_code=True)# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("cisimcik/tau-4b", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
Download configuration_tau.py from cisimcik/tau-4b: direct link, hf CLI and curl.
- Browser
- Download file 1.05 kB
-
https://huggingface.co/cisimcik/tau-4b/resolve/main/configuration_tau.py
- Command line
-
hf download hf://cisimcik/tau-4b/configuration_tau.py
-
curl -L -o configuration_tau.py https://huggingface.co/cisimcik/tau-4b/resolve/main/configuration_tau.py
1.05 kB
| """tau-4b configuration: a Qwen3.5 text trunk plus the System One readout (special tokens, readout block, heads).""" | |
| from transformers import PretrainedConfig | |
| class TauConfig(PretrainedConfig): | |
| model_type = "tau" | |
| def __init__( | |
| self, | |
| trunk: dict | None = None, | |
| special_base: int = 248_320, | |
| special_tokens: list[str] | None = None, | |
| readout_layers: int = 2, | |
| readout_heads: int = 8, | |
| pointer_dim: int = 256, | |
| max_tokens: int = 32_768, | |
| **kwargs, | |
| ): | |
| self.trunk = trunk or {} # Qwen3_5TextConfig as a dict | |
| self.special_base = special_base # id of the first control token (= trunk vocabulary size) | |
| self.special_tokens = special_tokens or [] | |
| self.readout_layers = readout_layers | |
| self.readout_heads = readout_heads | |
| self.pointer_dim = pointer_dim | |
| self.max_tokens = max_tokens # state + question; longer states are cut at the end | |
| super().__init__(**kwargs) | |