zakarth/violet-1b4-chat

Text Generation • 1B • Updated Feb 2 • 13 • 7

Note 1.41B GPT-NeoX chat model — Miss Violet Hartwell of Kensington, London, 1899. Original safetensors weights.

zakarth/violet-1b4-chat-gguf

Text Generation • 1B • Updated 8 days ago • 345 • 1

Note GGUF builds of violet-1b4-chat (F16 / Q8_0 / Q5_K_M / Q4_K_M) with embedded Jinja chat template — for llama.cpp, Ollama, LM Studio, KoboldCpp.

zakarth/violet-1b4-chat-onnx

Text Generation • Updated Feb 2 • 5

Note ONNX INT8 build of violet-1b4-chat for in-browser inference via Transformers.js / WebGPU.

zakarth/violet-1b4

Text Generation • 1B • Updated Feb 2 • 5

Note 1.41B GPT-NeoX base/completion model trained on the 1800–1899 corpus. No chat protocol — raw Victorian text continuation.

zakarth/violet-1b4-onnx

Text Generation • Updated Feb 2 • 1

Note ONNX build of the violet-1b4 completion model.

zakarth/violet-160m-chat

Text Generation • 0.2B • Updated Feb 3 • 8

Note 160M chat variant — smaller, faster, suitable for very constrained devices.

zakarth/violet-160m-chat-onnx

Text Generation • Updated Feb 3 • 3

Note ONNX build of the 160M chat variant for WebGPU.

zakarth/violet-160m

Text Generation • 0.2B • Updated Feb 2 • 2

Note 160M completion variant — smallest base model in the family.

zakarth/violet-160m-onnx

Text Generation • Updated Feb 2 • 1

Note ONNX build of the 160M completion variant.

Note Live in-browser demo of Violet running on WebGPU via Transformers.js — no data leaves your device.

zakarth/violet-160m-chat-gguf

Text Generation • 0.2B • Updated 8 days ago • 207

Note GGUF builds of violet-160m-chat (F16 / Q8_0 / Q5_K_M / Q4_K_M, 100–366 MB) with embedded Jinja chat template — for llama.cpp, Ollama, LM Studio.