Hakureirm/rwkv7-sglang-w4gptq-1.5b

RWKV-7 G1 1.5B — int4 GPTQ for rwkv-sglang

Hand-written weight-only int4 (GPTQ-calibrated, symmetric, group 64) quantization of BlinkDL's RWKV-7 "Goose" G1 1.5B, for the rwkv-sglang serving overlay.

• Accuracy (perplexity-style): GPTQ (wikitext-calibrated) lambada 0.639 vs 0.672 fp16 (−3.34pt), recovering +1.6pt over calibration-free RTN; kernel output is bit-identical to the offline dequant.
• Speed: faster than fp16 at every batch size ≤ 32 on an RTX 3090 (1.03–1.56× decode), via a hand-written int4 GEMV / small-M GEMM / tensor-core GEMM family (JIT, Turing→Blackwell).
• VRAM: checkpoint 1.2 GB vs 2.9 GB fp16 (~2.4×); serve VRAM −950 MiB at bsz1.

⚠️ Accuracy warning — multi-step reasoning (read before use)

Perplexity-style metrics understate int4's damage to multi-step reasoning at this model size. On MATH500 (avg@64, 32,000 rollouts) this checkpoint scores 14.98% vs fp16's 40.60% (−25.6pt) — the quantized model tends to lose the thread mid-derivation and run to the token cap (57.7% truncation vs fp16's 14.2%). This is a 1.5B-specific fragility, not a property of the scheme itself: the identical symmetric GPTQ at 7.2B costs only −3.1pt on the same ruler (see rwkv7-sglang-w4gptq-7.2b).

Recommendation: treat this checkpoint as a memory-footprint tool for non-reasoning workloads. For reasoning-heavy use at 1.5B, use the int8 w8g64 tier instead (greedy-exact, no measurable accuracy cost) — int4 is not the lossless tier at this size.

Where this fits

This checkpoint loads through the rwkv-sglang overlay. Native RWKV-7 support is being upstreamed into SGLang — see sgl-project/sglang#30115; once that lands the overlay is no longer required.

If you want RWKV-7 without quantization or a custom runtime, the same base weights are also published in standard HuggingFace layout — plain safetensors, ordinary config.json, no trust_remote_code:

0.1B	rwkv7-0.1b-hf
0.4B	rwkv7-0.4b-hf
1.5B	rwkv7-1.5b-hf
2.9B	rwkv7-2.9b-hf
7.2B (G1)	rwkv7-g1h-7.2b-hf
Pile 168M	rwkv7-168m-pile-hf

Transformers support for the architecture itself is open as huggingface/transformers#47780.

Format & loading (important)

Not a drop-in HuggingFace checkpoint. Weights are group-wise (GROUP=64) symmetric int4 (.qweight + .scale); they load only through the rwkv-sglang overlay:

bash scripts/deploy.sh   # from github.com/Hakureirm/rwkv-sglang, built against sglang v0.5.10.post1 — newer releases need the overlay rebased
RWKV_W4=1 python -m sglang.launch_server --model-path <this-dir> --dtype float16 \
  --trust-remote-code --disable-radix-cache

LoRA/norm/embedding/head stay full precision. Base model © BlinkDL (Bo Peng), Apache-2.0.