COMMITS
October 22, 2025
Z
Updated CHANGELOG
Zifei Xu committed
T
version bump in pyproject
Tristan Webb committed
T
Version v0.1.11
Tristan Webb committed
October 21, 2025
X
Minor fix for backward compatibility with fx tracing (#1018)
xuzifei-dmatrix committed
October 19, 2025
H
removed raptor configs. pushed to main by mistake
hmostafa committed
H
added raptor configs
hmostafa committed
October 16, 2025
B
Added monitoring across gms to support monitoring Whisper encoder layers (#1013)
brendan-dmatrix committed
October 11, 2025
H
Merge pull request #1014 from d-matrix-ai/hesham/asym_bfp16_fix
Hesham Mostafa committed
H
fixed hardcoded value in asymmetric mantissa postprocessing
hmostafa committed
October 10, 2025
H
added fix for asymmetric mantissa in BFP16_64
hmostafa committed
October 7, 2025
T
Update README.md
Tristan Webb committed
September 26, 2025
T
update poetry lock and tighten Python version for graph viz
Tristan Webb committed
X
Adding support for quantized submod (#986)
xuzifei-dmatrix committed
October 2, 2025
X
Torch export Integration (#996)
xuzifei-dmatrix committed
September 30, 2025
H
Merge pull request #1007 from d-matrix-ai/sayeh/sdpa_operation_order
Hesham Mostafa committed
H
Merge pull request #1008 from d-matrix-ai/sayeh/fp16_subnormal_flush
Hesham Mostafa committed
“
flush fp16 subnormal values in when flushing is set
“sayehs” committed
“
changed order of operations in SDPA to match hardware behavior
“sayehs” committed
September 20, 2025
H
Merge pull request #985 from d-matrix-ai/hesham/expanded_simd_arguments
Hesham Mostafa committed
September 19, 2025
H
added more robust checking for the presence of the numerics module
hmostafa committed
September 18, 2025
H
formatted and added docs to approximate.py
hmostafa committed
September 8, 2025
H
Merge pull request #984 from d-matrix-ai/hesham/flexible_quantization
Hesham Mostafa committed
September 6, 2025
H
added support for vsimd wrappers and arguments
hmostafa committed
H
fixed test
hmostafa committed
H
added more flexible quantization options
hmostafa committed
W
Merge pull request #983 from d-matrix-ai/sayeh/to_fp8_eval
wanzin-dmatrix committed
“
added to_fp8_mode to dmx model.
“sayehs” committed
September 5, 2025
H
Merge pull request #982 from d-matrix-ai/sayeh/fp8_eval_mode
Hesham Mostafa committed
“
uncommented vanilla eval in benchmark_clip
“sayehs” committed
“
fix apply config rule
“sayehs” committed