Text Generation
MLX
Safetensors
qwen3_5
transliteration
pinyin
romaji
hinglish
arabizi
ime
qwen3.5
v2.1
conversational
6-bit
Instructions to use UnimeType/Transliteration-2.0-4B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use UnimeType/Transliteration-2.0-4B with MLX:
# Make sure mlx-lm is installed # pip install --upgrade mlx-lm # Generate text with mlx-lm from mlx_lm import load, generate model, tokenizer = load("UnimeType/Transliteration-2.0-4B") prompt = "Write a story about Einstein" messages = [{"role": "user", "content": prompt}] prompt = tokenizer.apply_chat_template( messages, add_generation_prompt=True ) text = generate(model, tokenizer, prompt=prompt, verbose=True) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Pi
How to use UnimeType/Transliteration-2.0-4B with Pi:
Start the MLX server
# Install MLX LM: uv tool install mlx-lm # Start a local OpenAI-compatible server: mlx_lm.server --model "UnimeType/Transliteration-2.0-4B"
Configure the model in Pi
# Install Pi: npm install -g @earendil-works/pi-coding-agent # Add to ~/.pi/agent/models.json: { "providers": { "mlx-lm": { "baseUrl": "http://localhost:8080/v1", "api": "openai-completions", "apiKey": "none", "models": [ { "id": "UnimeType/Transliteration-2.0-4B" } ] } } }Run Pi
# Start Pi in your project directory: pi
- MLX LM
How to use UnimeType/Transliteration-2.0-4B with MLX LM:
Generate or start a chat session
# Install MLX LM uv tool install mlx-lm # Interactive chat REPL mlx_lm.chat --model "UnimeType/Transliteration-2.0-4B"
Run an OpenAI-compatible server
# Install MLX LM uv tool install mlx-lm # Start the server mlx_lm.server --model "UnimeType/Transliteration-2.0-4B" # Calling the OpenAI-compatible server with curl curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "UnimeType/Transliteration-2.0-4B", "messages": [ {"role": "user", "content": "Hello"} ] }' - Hermes Agent
How to use UnimeType/Transliteration-2.0-4B with Hermes Agent:
Start the MLX server
# Install MLX LM: uv tool install mlx-lm # Start a local OpenAI-compatible server: mlx_lm.server --model "UnimeType/Transliteration-2.0-4B"
Configure Hermes
# Install Hermes: curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash hermes setup # Point Hermes at the local server: hermes config set model.provider custom hermes config set model.base_url http://127.0.0.1:8080/v1 hermes config set model.default UnimeType/Transliteration-2.0-4B
Run Hermes
hermes
- Atomic Chat
- OpenClaw
How to use UnimeType/Transliteration-2.0-4B with OpenClaw:
Start the MLX server
# Install MLX LM: uv tool install mlx-lm # Start a local OpenAI-compatible server: mlx_lm.server --model "UnimeType/Transliteration-2.0-4B"
Configure OpenClaw
# Install OpenClaw: npm install -g openclaw@latest # Register the local server and set it as the default model: openclaw onboard --non-interactive --mode local \ --auth-choice custom-api-key \ --custom-base-url http://127.0.0.1:8080/v1 \ --custom-model-id "UnimeType/Transliteration-2.0-4B" \ --custom-provider-id mlx-lm \ --custom-compatibility openai \ --custom-text-input \ --accept-risk \ --skip-health
Run OpenClaw
openclaw agent --local --agent main --message "Hello from Hugging Face"
Download PROVENANCE.json from UnimeType/Transliteration-2.0-4B: direct link, hf CLI and curl.
- Browser
- Download file 4.92 kB
-
https://huggingface.co/UnimeType/Transliteration-2.0-4B/resolve/main/PROVENANCE.json
- Command line
-
hf download hf://UnimeType/Transliteration-2.0-4B/PROVENANCE.json
-
curl -L -o PROVENANCE.json https://huggingface.co/UnimeType/Transliteration-2.0-4B/resolve/main/PROVENANCE.json
4.92 kB
| { | |
| "release_version": "2.0-v4", | |
| "format": "MLX 6-bit affine, group size 64", | |
| "base_model": "Qwen/Qwen3.5-4B", | |
| "base_snapshot_revision": "851bf6e806efd8d0a36b00ddf55e13ccb7b8cd0a", | |
| "base_source_sha256": { | |
| "model.safetensors-00001-of-00002.safetensors": "26a93f066e1916adb13453dae5a0c707c0fbc71299ed98779571a907b8e74c61", | |
| "model.safetensors-00002-of-00002.safetensors": "cb544bd9bfae93dc59b0f22b292f5933573854a7f9b97835c67060d7d910e188" | |
| }, | |
| "lora_adapter_sha256": "fbc38ca87cfb89381882ab2d503edd651a759de4ebd978cf13b8924ccea14f9e", | |
| "training_method": "Two-Stage LoRA SFT + Reference-Anchored DPO (r=16, alpha=32, all 32 layers) on native BF16 Qwen/Qwen3.5-4B with multi-clause long-sentence curriculum", | |
| "native_unquantized_base_precision": true, | |
| "training_dataset_public": false, | |
| "standard_weight_sha256": { | |
| "model-00001-of-00002.safetensors": "c2f53be3784006c53a0c406bf47f851c70396807a924b977c723c7ba42fb6f9a", | |
| "model-00002-of-00002.safetensors": "e1f4969fe87d5564eba3ed06b58e8d5aa111c9b43667046a3dc8e5a8eccce294" | |
| }, | |
| "published_weight_sha256": { | |
| "model.safetensors": "6b9667173d636acffd2dcbae81c91c5a45fd0dce6f6b5f8ec4e909faf04c31b3" | |
| }, | |
| "evaluation": { | |
| "protocol": "unimetype-convert-short-v4-long", | |
| "winning_checkpoint": "sft_best", | |
| "short_414_suite": { | |
| "strict_reference_matches": 366, | |
| "strict_rate": 0.8841, | |
| "expanded_reference_matches": 366, | |
| "expanded_rate": 0.8841, | |
| "total": 414, | |
| "by_language_strict": { | |
| "chinese": { | |
| "passed": 195, | |
| "total": 207, | |
| "rate": 0.942 | |
| }, | |
| "japanese": { | |
| "passed": 40, | |
| "total": 52, | |
| "rate": 0.7692 | |
| }, | |
| "hindi": { | |
| "passed": 61, | |
| "total": 71, | |
| "rate": 0.8592 | |
| }, | |
| "arabic": { | |
| "passed": 70, | |
| "total": 84, | |
| "rate": 0.8333 | |
| } | |
| }, | |
| "by_language_expanded": { | |
| "chinese": { | |
| "passed": 195, | |
| "total": 207, | |
| "rate": 0.942 | |
| }, | |
| "japanese": { | |
| "passed": 40, | |
| "total": 52, | |
| "rate": 0.7692 | |
| }, | |
| "hindi": { | |
| "passed": 61, | |
| "total": 71, | |
| "rate": 0.8592 | |
| }, | |
| "arabic": { | |
| "passed": 70, | |
| "total": 84, | |
| "rate": 0.8333 | |
| } | |
| } | |
| }, | |
| "long_160_suite": { | |
| "tag": "4b-v4-sft_best", | |
| "model_path": "/Users/losdwind/CodeProjects2026/MixedIME/outputs/2026-10-30-long-sentence-v4/release-4b/mlx-6bit", | |
| "total": 160, | |
| "strict_matches": 46, | |
| "strict_rate": 0.2875, | |
| "normalized_matches": 47, | |
| "normalized_rate": 0.2938, | |
| "by_language": { | |
| "chinese": { | |
| "strict": 9, | |
| "norm": 9, | |
| "total": 40, | |
| "strict_rate": 0.225, | |
| "norm_rate": 0.225, | |
| "char_accuracy": 0.9261 | |
| }, | |
| "japanese": { | |
| "strict": 9, | |
| "norm": 9, | |
| "total": 40, | |
| "strict_rate": 0.225, | |
| "norm_rate": 0.225, | |
| "char_accuracy": 0.9098 | |
| }, | |
| "hindi": { | |
| "strict": 25, | |
| "norm": 26, | |
| "total": 40, | |
| "strict_rate": 0.625, | |
| "norm_rate": 0.65, | |
| "char_accuracy": 0.9809 | |
| }, | |
| "arabic": { | |
| "strict": 3, | |
| "norm": 3, | |
| "total": 40, | |
| "strict_rate": 0.075, | |
| "norm_rate": 0.075, | |
| "char_accuracy": 0.9118 | |
| } | |
| }, | |
| "by_slice": { | |
| "multi_clause_spaced": { | |
| "strict": 4, | |
| "norm": 4, | |
| "total": 32, | |
| "strict_rate": 0.125, | |
| "norm_rate": 0.125 | |
| }, | |
| "multi_clause_unspaced_or_grouped": { | |
| "strict": 7, | |
| "norm": 7, | |
| "total": 32, | |
| "strict_rate": 0.2188, | |
| "norm_rate": 0.2188 | |
| }, | |
| "multi_clause_codeswitch_tech": { | |
| "strict": 17, | |
| "norm": 17, | |
| "total": 32, | |
| "strict_rate": 0.5312, | |
| "norm_rate": 0.5312 | |
| }, | |
| "multi_clause_protected_multiline": { | |
| "strict": 9, | |
| "norm": 10, | |
| "total": 32, | |
| "strict_rate": 0.2812, | |
| "norm_rate": 0.3125 | |
| }, | |
| "multi_clause_reconvert_and_lexicon": { | |
| "strict": 9, | |
| "norm": 9, | |
| "total": 32, | |
| "strict_rate": 0.2812, | |
| "norm_rate": 0.2812 | |
| } | |
| }, | |
| "contract_violations": { | |
| "rejected_output_repeated": 2, | |
| "retained_latin_changed": 1 | |
| }, | |
| "elapsed_seconds": 125.83, | |
| "mean_char_accuracy": 0.9321 | |
| } | |
| }, | |
| "repository": "UnimeType/Transliteration-2.0-4B", | |
| "version": "2.1", | |
| "model_version": "2.1 (2.0-v4)", | |
| "release_tag": "v2.1", | |
| "protocol_version": "unimetype-convert-short-v3", | |
| "source_safetensors_repo": "UnimeType/Transliteration-2.0-4B-Safetensors" | |
| } | |