4.33 GB

Ctrl+K

1 contributor

History: 24 commits

F-Labs

Harden final logits against non-finite generation scores

d322fd1 verified about 1 month ago

.gitattributes

1.52 kB
initial commit about 2 months ago
ADAPTIVE_RELEASE.md

3.23 kB
Update model with adaptive calibration-aware mixed-precision release about 1 month ago
LICENSE

11.3 kB
Add Apache 2.0 LICENSE file about 2 months ago
README.md

4.24 kB
Contain non-finite KV-BSS scores during generation about 1 month ago
adaptive_activation_second_moments.pt

3.19 MB
xet

Update model with adaptive calibration-aware mixed-precision release about 1 month ago
adaptive_bifurcation_report.json

221 kB
Update model with adaptive calibration-aware mixed-precision release about 1 month ago
chat_template.jinja

9.06 kB
Release MiniCPM5-2B-Hadamard-GSQ: 2.03 GB (2.30x compression), DV-SSQ, KV-BSS, Zero-Compression Shield about 2 months ago
config.json

54.9 kB
Update model with adaptive calibration-aware mixed-precision release about 1 month ago
configuration_minicpm_hadamard.py

2.61 kB
Update model with adaptive calibration-aware mixed-precision release about 1 month ago
generation_config.json

213 Bytes
Release MiniCPM5-2B-Hadamard-GSQ: 2.03 GB (2.30x compression), DV-SSQ, KV-BSS, Zero-Compression Shield about 2 months ago
kv_bss.py

6.33 kB
Contain non-finite KV-BSS scores during generation about 1 month ago
model-00001-of-00003.safetensors

2.15 GB
xet

Update model with adaptive calibration-aware mixed-precision release about 1 month ago
model-00002-of-00003.safetensors

2.13 GB
xet

Update model with adaptive calibration-aware mixed-precision release about 1 month ago
model-00003-of-00003.safetensors

44 MB
xet

Update model with adaptive calibration-aware mixed-precision release about 1 month ago
model.safetensors.index.json

47.9 kB
Update model with adaptive calibration-aware mixed-precision release about 1 month ago
modeling_minicpm_hadamard.py

27.4 kB
Harden final logits against non-finite generation scores about 1 month ago
quantize_minicpm.py

14.5 kB
Rename colliding acronyms: group-wise INT4 + SRC (SVD Residual Compensation) about 2 months ago
quantize_minicpm_adaptive.py

23.8 kB
Update model with adaptive calibration-aware mixed-precision release about 1 month ago
special_tokens_map.json

551 Bytes
Release MiniCPM5-2B-Hadamard-GSQ: 2.03 GB (2.30x compression), DV-SSQ, KV-BSS, Zero-Compression Shield about 2 months ago
test_inference.py

969 Bytes
Add production source file test_inference.py about 2 months ago
tokenizer.json

9.89 MB
Release MiniCPM5-2B-Hadamard-GSQ: 2.03 GB (2.30x compression), DV-SSQ, KV-BSS, Zero-Compression Shield about 2 months ago
tokenizer_config.json

94.4 kB
Release MiniCPM5-2B-Hadamard-GSQ: 2.03 GB (2.30x compression), DV-SSQ, KV-BSS, Zero-Compression Shield about 2 months ago