Index-Echo-S2ST-2B-FP4

Official NVFP4 (W4A4) quantization of IndexTeam/Index-Echo-S2ST-2B, part of the Index-Echo speech-to-speech translation (S2ST) model family by bilibili.

This repository mirrors the original checkpoint layout; only the text LLM backbone (stlm_llm/) is quantized to NVFP4 - the audio tower, connector, and all speech-synthesis components remain in BF16. Use it exactly like the original repo (same infer.py / configs).

Quantization

• Scheme: NVFP4 (W4A4: 4-bit floating-point weights with per-group-16 scales, 4-bit floating-point activations with calibrated per-tensor global scales), produced with llm-compressor (quantization_scheme recorded in recipe.yaml). Calibrated on a small bilingual translation corpus.
• All Linear layers of the language model backbone (stlm_llm/) are quantized; the audio tower, connector, lm_head, embeddings and all other pipeline components are kept in BF16.
• Format: compressed-tensors nvfp4-pack-quantized safetensors - load directly with vLLM (quantization="compressed-tensors") or transformers.

Consistency validation

Measured on an NVIDIA A100 (weight-dequantized execution) against the original BF16 checkpoint (greedy decoding, official translation prompt):

Metric	BF16	FP4	Delta
Perplexity (fixed corpus)	5.9332	6.4980	+9.52%
zh->en generation identical	-	-	yes
en->zh generation identical	-	-	yes

Usage

Identical to the original checkpoint - clone this repo and follow the README / infer.py of the base model (IndexTeam/Index-Echo-S2ST-2B). The quantized LLM backbone loads via compressed-tensors; make sure compressed-tensors (or vLLM / a recent transformers) is installed.

Hardware note: full NVFP4 (W4A4) acceleration requires an NVIDIA Blackwell GPU (SM100+, e.g. B200 / RTX 50 series). On older GPUs vLLM falls back to weight-only dequantization, which still reduces memory but gives no FP4 speedup.

Quantized and published by the Index team, 2026-10-04.