Modelpublic

lemuralabs/DeepSeek-V4-Flash-TQ-Q4.5-MLX

sourceHugging Facemitupdated 2mo agoView on Hugging Face

1likes198downloads

Model Card

DeepSeek-V4-Flash-TQ-Q4.5-MLX

lemuralabs/DeepSeek-V4-Flash-TQ-Q4.5-MLX is an Apple-Silicon MLX TurboQuant/JANGTQ quantization of `deepseek-ai/DeepSeek-V4-Flash`.

No fine-tuning, distillation, or retraining was applied. The official mixed FP4/FP8 source weights were converted locally, the MTP head was dropped because it is not used for normal decode, and router/mHC/control tensors were preserved rather than aggressively quantized.

Model Details

Property	Value
Base model	`deepseek-ai/DeepSeek-V4-Flash`
Architecture	DeepSeek-V4 Flash MoE, 284B total / 13B active, 1M context
Local profile	`JANGTQ-Q4.5`
Bundle size	164.25 GB
Layout	Pre-stacked MLX `switch_mlp` layout
MTP head	Dropped
Validation	Safetensors header/index validation, metadata validation

Required Sidecar

This is a JANGTQ/TurboQuant bundle and requires jangtq_runtime.safetensors from this repository. The sidecar stores the deterministic codebooks and Hadamard rotation signs used to decode the .tq_packed expert weights. If it is missing, re-download the full repository or fetch that file explicitly:

bash

hf download lemuralabs/DeepSeek-V4-Flash-TQ-Q4.5-MLX jangtq_runtime.safetensors --local-dir <your-model-dir>

Quantization Recipe

Tensor class	Codec	Bits / handling
Routed experts	TurboQuant MXTQ	113 routed layer/projection groups at 4-bit MXTQ and 16 at 8-bit MXTQ
Routed effective bits	MXTQ	4.4961 bits
Attention, shared experts, compressor, indexer, embed, lm head	MLX affine	8-bit, group size 32
Norms, router, mHC, sinks, integer routing tables	passthrough	source precision preserved

The fractional target is implemented as a power-of-two lane mix because the current JANGTQ vectorized packer is stable on 2/4/8-bit lanes for DeepSeek-V4 expert dimensions.

Use

Install the JANG loader/runtime and MLX LM:

bash

pip install mlx-lm jang-tools

Example:

python

from jang_tools.load_jangtq import load_jangtq_model
from mlx_lm import generate

model, tokenizer = load_jangtq_model("lemuralabs/DeepSeek-V4-Flash-TQ-Q4.5-MLX")
prompt = "Write a short note about MLX quantization."
text = generate(model, tokenizer, prompt=prompt, verbose=True)
print(text)

Files

—model-*.safetensors: pre-stacked JANGTQ/MLX shards
—model.safetensors.index.json: shard index
—jangtq_runtime.safetensors: required TurboQuant runtime sidecar
—config.json, jang_config.json: MLX/JANGTQ metadata
—encoding/: upstream DeepSeek-V4 prompt encoding reference

Notes

This upload follows the same broad model-card shape as the public OsaurusAI/JANGQ DeepSeek-V4-Flash JANGTQ uploads: a sidecar warning, an explicit recipe table, and minimal reproducible loading instructions.

License

MIT, following the upstream DeepSeek-V4-Flash release.