Reza2kn/sapiens2-pose-0.8b-INT4-G128

sapiens2-pose-0.8b INT4-G128

Packed 4-bit derivative of facebook/sapiens2-pose-0.8b.

This artifact uses symmetric per-group INT4 packing with group size 128 for large floating-point weight tensors. Norms, biases, positional/rope tensors, and small tensors are kept in their source dtype. It is a storage/runtime-loader quant for the current official Sapiens2 code path, not an AWQ/GGUF/NVFP4 LLM artifact.

Files

• facebook__sapiens2-pose-0.8b-int4-g128.safetensors: packed INT4 safetensors artifact.
• load_sapiens2_int4.py: loader that reconstructs a PyTorch state dict for the official Sapiens2 model code.
• config.json and preprocessor_config.json: copied from the source repo.
• quantization_report.json: build report.

Quantization Report

• Source revision: 2d5be15b2cdc2bac5d9ea8a61849c3ba2c52f42a
• Group size: 128
• Source bytes: 3394819360
• Artifact bytes: 439862416
• Compression ratio: 7.7179x
• Tensors: 560
• Quantized tensors: 200
• Max tensor MAE during dequant smoke: 0.00429335

Fidelity Validation

The packed INT4 artifact was dequantized back to floating-point tensors and compared against the source checkpoint.

• Validation gate: global floating-tensor similarity >= 90.00%
• Result: PASS
• Global floating-tensor similarity: 99.425898%
• Minimum large-tensor cosine: 0.974259198

Loading

from load_sapiens2_int4 import load_state_dict

state_dict = load_state_dict("facebook__sapiens2-pose-0.8b-int4-g128.safetensors", device="cpu")
# Then instantiate the matching official Sapiens2 architecture and load:
# model.load_state_dict(state_dict, strict=True)

Limitations

This is a verified packed-weight artifact with a dequantizing loader. It does not claim native INT4 CUDA kernels for Sapiens2 yet. Runtime speedups require a Sapiens2-specific kernel/export path and should be benchmarked separately.