Search results

Add scale_i32_bf16 operator

Status: Open.

#99 In amd/IRON;

· albiol2004 opened on Apr 9, 2026

Add quantize operator (bf16->int8)

Status: Open.

#97 In amd/IRON;

· albiol2004 opened on Apr 9, 2026

Add dequantize operator (int32 → bfloat16)

Status: Open.

#95 In amd/IRON;

· albiol2004 opened on Apr 9, 2026

INT8 GEMM support

Status: Open.

#93 In amd/IRON;

· albiol2004 opened on Apr 9, 2026

To do: Softmax masking should be optimized

Status: Open.

#84 In amd/IRON;

· andrej opened on Mar 23, 2026

GEMM partition_N > 1 produces incorrect results (affects Llama final vocab projection)

Status: Open.

#83 In amd/IRON;

· tonyjie opened on Mar 20, 2026

Intermittent CI failures for llama on later iterations

Status: Open.

#54 In amd/IRON;

· andrej opened on Dec 10, 2025

GGML Operation Requirements

Status: Open.

#53 In amd/IRON;

· ypapadop-amd opened on Dec 10, 2025

Clean up SwiGLU reference code weight generation

Status: Open.

#52 In amd/IRON;

· andrej opened on Dec 10, 2025

Clean up test discovery and operator paths

Status: Open.

#42 In amd/IRON;

· andrej opened on Dec 2, 2025

Normalization Kernels Conflate tile_size and Batch Size

Status: Open.

#40 In amd/IRON;

· andrej opened on Dec 1, 2025

Fix or remove broken leaky ReLU kernel

Status: Open.

#36 In amd/IRON;

· andrej opened on Nov 25, 2025