Open riverliuintel opened 1 month ago
Request INT8 quantization (PT2E) feature on Linux. It requires, implement PT2E infrastructure for Intel GPU path, complete essential oneDNN, Triton quantized INT8 ops, pass benchmark models quantization testing.
And complete essential docs changes
No response
🚀 The feature, motivation and pitch
Request INT8 quantization (PT2E) feature on Linux. It requires, implement PT2E infrastructure for Intel GPU path, complete essential oneDNN, Triton quantized INT8 ops, pass benchmark models quantization testing.
And complete essential docs changes
Alternatives
No response
Additional context
No response