text stringlengths 1.1k 6.49k | spec stringclasses 16
values | ok_variants int64 1 6 | n_variants int64 1 6 |
|---|---|---|---|
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton softmax_row | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [4096, 4096]}
KERNEL run(x) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write the... | softmax_row | 2 | 2 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton rmsnorm_row | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [2048, 4096], 'w': [4096]}
KERNEL run(x, w) -> torch.Tensor
OBJECTIVE minimize late... | rmsnorm_row | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton gemm_nn | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'a': [1024, 4096], 'b': [4096, 1024]}
KERNEL run(a, b) -> torch.Tensor
OBJECTIVE minimize la... | gemm_nn | 3 | 4 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton rowsum | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [4096, 4096]}
KERNEL run(x) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write the base... | rowsum | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton conv1d_causal_dw | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [8, 256, 8192], 'w': [256, 1, 4]}
KERNEL run(x, w) -> torch.Tensor
OBJECTIVE m... | conv1d_causal_dw | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton cumsum_row | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [256, 16384]}
KERNEL run(x) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write the ... | cumsum_row | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton layernorm_row | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [2048, 4096], 'w': [4096], 'b': [4096]}
KERNEL run(x, w, b) -> torch.Tensor
OBJEC... | layernorm_row | 2 | 2 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton add_rmsnorm | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [2048, 2048], 'r': [2048, 2048], 'w': [2048]}
KERNEL run(x, r, w) -> torch.Tensor
O... | add_rmsnorm | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton rope_pairwise | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'q': [2, 4, 2048, 128], 'cos': [2048, 64], 'sin': [2048, 64]}
KERNEL run(q, cos, sin) ... | rope_pairwise | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton causal_attn_small | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'q': [1, 4, 256, 64], 'k': [1, 4, 256, 64], 'v': [1, 4, 256, 64]}
KERNEL run(q, k,... | causal_attn_small | 2 | 2 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton topk_rows | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [1024, 4096]}
KERNEL run(x) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write the b... | topk_rows | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton embedding_gather | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'emb': [32768, 512], 'idx': [8192]}
KERNEL run(emb, idx) -> torch.Tensor
OBJECTIVE ... | embedding_gather | 2 | 2 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton transpose_2d | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [8192, 8192]}
KERNEL run(x) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write th... | transpose_2d | 2 | 2 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton adamw_fused | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'p': [1048576], 'g': [1048576], 'm': [1048576], 'v': [1048576]}
KERNEL run(p, g, m, v) -... | adamw_fused | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton matvec | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'a': [4096, 8192], 'b': [8192]}
KERNEL run(a, b) -> torch.Tensor
OBJECTIVE minimize latency_m... | matvec | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton swiglu_silu | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'g': [16384, 1024]}
KERNEL run(g) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write th... | swiglu_silu | 6 | 6 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton softmax_row | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [8192, 8192]}
KERNEL run(x) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write the... | softmax_row | 2 | 2 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton rmsnorm_row | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [1024, 2048], 'w': [2048]}
KERNEL run(x, w) -> torch.Tensor
OBJECTIVE minimize late... | rmsnorm_row | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton gemm_nn | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'a': [4096, 1024], 'b': [1024, 1024]}
KERNEL run(a, b) -> torch.Tensor
OBJECTIVE minimize la... | gemm_nn | 3 | 4 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton rowsum | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [4096, 16384]}
KERNEL run(x) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write the bas... | rowsum | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton conv1d_causal_dw | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [4, 256, 8192], 'w': [256, 1, 4]}
KERNEL run(x, w) -> torch.Tensor
OBJECTIVE m... | conv1d_causal_dw | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton cumsum_row | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [256, 8192]}
KERNEL run(x) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write the b... | cumsum_row | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton layernorm_row | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [2048, 4096], 'w': [4096], 'b': [4096]}
KERNEL run(x, w, b) -> torch.Tensor
OBJEC... | layernorm_row | 2 | 2 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton add_rmsnorm | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [1024, 4096], 'r': [1024, 4096], 'w': [4096]}
KERNEL run(x, r, w) -> torch.Tensor
O... | add_rmsnorm | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton rope_pairwise | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'q': [2, 4, 1024, 128], 'cos': [1024, 64], 'sin': [1024, 64]}
KERNEL run(q, cos, sin) ... | rope_pairwise | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton causal_attn_small | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'q': [1, 4, 256, 64], 'k': [1, 4, 256, 64], 'v': [1, 4, 256, 64]}
KERNEL run(q, k,... | causal_attn_small | 2 | 2 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton topk_rows | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [512, 4096]}
KERNEL run(x) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write the ba... | topk_rows | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton embedding_gather | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'emb': [32768, 512], 'idx': [8192]}
KERNEL run(emb, idx) -> torch.Tensor
OBJECTIVE ... | embedding_gather | 2 | 2 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton transpose_2d | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [2048, 2048]}
KERNEL run(x) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write th... | transpose_2d | 2 | 2 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton adamw_fused | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'p': [1048576], 'g': [1048576], 'm': [1048576], 'v': [1048576]}
KERNEL run(p, g, m, v) -... | adamw_fused | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton matvec | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'a': [4096, 4096], 'b': [4096]}
KERNEL run(a, b) -> torch.Tensor
OBJECTIVE minimize latency_m... | matvec | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton swiglu_silu | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'g': [16384, 1024]}
KERNEL run(g) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write th... | swiglu_silu | 6 | 6 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton softmax_row | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [2048, 2048]}
KERNEL run(x) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write the... | softmax_row | 2 | 2 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton rmsnorm_row | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [1024, 4096], 'w': [4096]}
KERNEL run(x, w) -> torch.Tensor
OBJECTIVE minimize late... | rmsnorm_row | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton gemm_nn | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'a': [4096, 1024], 'b': [1024, 4096]}
KERNEL run(a, b) -> torch.Tensor
OBJECTIVE minimize la... | gemm_nn | 3 | 4 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton rowsum | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [1024, 16384]}
KERNEL run(x) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write the bas... | rowsum | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton conv1d_causal_dw | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [8, 256, 8192], 'w': [256, 1, 4]}
KERNEL run(x, w) -> torch.Tensor
OBJECTIVE m... | conv1d_causal_dw | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton cumsum_row | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [512, 16384]}
KERNEL run(x) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write the ... | cumsum_row | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton layernorm_row | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [512, 2048], 'w': [2048], 'b': [2048]}
KERNEL run(x, w, b) -> torch.Tensor
OBJECT... | layernorm_row | 2 | 2 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton rope_pairwise | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'q': [2, 8, 2048, 128], 'cos': [2048, 64], 'sin': [2048, 64]}
KERNEL run(q, cos, sin) ... | rope_pairwise | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton causal_attn_small | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'q': [1, 8, 512, 64], 'k': [1, 8, 512, 64], 'v': [1, 8, 512, 64]}
KERNEL run(q, k,... | causal_attn_small | 2 | 2 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton topk_rows | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [1024, 4096]}
KERNEL run(x) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write the b... | topk_rows | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton embedding_gather | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'emb': [32768, 512], 'idx': [8192]}
KERNEL run(emb, idx) -> torch.Tensor
OBJECTIVE ... | embedding_gather | 2 | 2 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton transpose_2d | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [8192, 8192]}
KERNEL run(x) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write th... | transpose_2d | 2 | 2 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton adamw_fused | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'p': [1048576], 'g': [1048576], 'm': [1048576], 'v': [1048576]}
KERNEL run(p, g, m, v) -... | adamw_fused | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton matvec | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'a': [4096, 4096], 'b': [4096]}
KERNEL run(a, b) -> torch.Tensor
OBJECTIVE minimize latency_m... | matvec | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton swiglu_silu | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'g': [16384, 2048]}
KERNEL run(g) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write th... | swiglu_silu | 6 | 6 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton softmax_row | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [8192, 8192]}
KERNEL run(x) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write the... | softmax_row | 2 | 2 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton rmsnorm_row | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [2048, 2048], 'w': [2048]}
KERNEL run(x, w) -> torch.Tensor
OBJECTIVE minimize late... | rmsnorm_row | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton gemm_nn | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'a': [4096, 1024], 'b': [1024, 1024]}
KERNEL run(a, b) -> torch.Tensor
OBJECTIVE minimize la... | gemm_nn | 3 | 4 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton rowsum | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [1024, 16384]}
KERNEL run(x) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write the bas... | rowsum | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton conv1d_causal_dw | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [4, 128, 8192], 'w': [128, 1, 4]}
KERNEL run(x, w) -> torch.Tensor
OBJECTIVE m... | conv1d_causal_dw | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton cumsum_row | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [512, 8192]}
KERNEL run(x) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write the b... | cumsum_row | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton layernorm_row | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [1024, 4096], 'w': [4096], 'b': [4096]}
KERNEL run(x, w, b) -> torch.Tensor
OBJEC... | layernorm_row | 2 | 2 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton rope_pairwise | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'q': [2, 4, 1024, 128], 'cos': [1024, 64], 'sin': [1024, 64]}
KERNEL run(q, cos, sin) ... | rope_pairwise | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton causal_attn_small | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'q': [1, 4, 512, 64], 'k': [1, 4, 512, 64], 'v': [1, 4, 512, 64]}
KERNEL run(q, k,... | causal_attn_small | 2 | 2 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton topk_rows | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [1024, 4096]}
KERNEL run(x) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write the b... | topk_rows | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton embedding_gather | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'emb': [32768, 1024], 'idx': [8192]}
KERNEL run(emb, idx) -> torch.Tensor
OBJECTIVE... | embedding_gather | 2 | 2 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton transpose_2d | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [8192, 8192]}
KERNEL run(x) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write th... | transpose_2d | 2 | 2 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton adamw_fused | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'p': [4194304], 'g': [4194304], 'm': [4194304], 'v': [4194304]}
KERNEL run(p, g, m, v) -... | adamw_fused | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton matvec | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'a': [4096, 8192], 'b': [8192]}
KERNEL run(a, b) -> torch.Tensor
OBJECTIVE minimize latency_m... | matvec | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton swiglu_silu | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'g': [16384, 1024]}
KERNEL run(g) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write th... | swiglu_silu | 6 | 6 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton softmax_row | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [2048, 4096]}
KERNEL run(x) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write the... | softmax_row | 2 | 2 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton rmsnorm_row | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [4096, 2048], 'w': [2048]}
KERNEL run(x, w) -> torch.Tensor
OBJECTIVE minimize late... | rmsnorm_row | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton gemm_nn | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'a': [4096, 1024], 'b': [1024, 1024]}
KERNEL run(a, b) -> torch.Tensor
OBJECTIVE minimize la... | gemm_nn | 3 | 4 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton rowsum | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [2048, 8192]}
KERNEL run(x) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write the base... | rowsum | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton conv1d_causal_dw | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [8, 256, 8192], 'w': [256, 1, 4]}
KERNEL run(x, w) -> torch.Tensor
OBJECTIVE m... | conv1d_causal_dw | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton cumsum_row | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [512, 8192]}
KERNEL run(x) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write the b... | cumsum_row | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton layernorm_row | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [512, 4096], 'w': [4096], 'b': [4096]}
KERNEL run(x, w, b) -> torch.Tensor
OBJECT... | layernorm_row | 2 | 2 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton rope_pairwise | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'q': [2, 4, 1024, 128], 'cos': [1024, 64], 'sin': [1024, 64]}
KERNEL run(q, cos, sin) ... | rope_pairwise | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton causal_attn_small | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'q': [1, 4, 256, 64], 'k': [1, 4, 256, 64], 'v': [1, 4, 256, 64]}
KERNEL run(q, k,... | causal_attn_small | 2 | 2 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton topk_rows | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [512, 4096]}
KERNEL run(x) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write the ba... | topk_rows | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton embedding_gather | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'emb': [32768, 1024], 'idx': [8192]}
KERNEL run(emb, idx) -> torch.Tensor
OBJECTIVE... | embedding_gather | 2 | 2 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton transpose_2d | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [4096, 4096]}
KERNEL run(x) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write th... | transpose_2d | 2 | 2 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton adamw_fused | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'p': [1048576], 'g': [1048576], 'm': [1048576], 'v': [1048576]}
KERNEL run(p, g, m, v) -... | adamw_fused | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton matvec | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'a': [4096, 8192], 'b': [8192]}
KERNEL run(a, b) -> torch.Tensor
OBJECTIVE minimize latency_m... | matvec | 1 | 1 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton swiglu_silu | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'g': [8192, 1024]}
KERNEL run(g) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write the... | swiglu_silu | 6 | 6 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton softmax_row | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [4096, 4096]}
KERNEL run(x) -> torch.Tensor
OBJECTIVE minimize latency_ms
Write the... | softmax_row | 2 | 2 |
<|prompt>
You are a kernel optimization agent. Given an optimization task and the trajectory so far, write the next complete kernel candidate. Output code only.
TASK triton rmsnorm_row | gpu=mi350x arch=gfx950 | dtype=bf16 | shape={'x': [1024, 2048], 'w': [2048]}
KERNEL run(x, w) -> torch.Tensor
OBJECTIVE minimize late... | rmsnorm_row | 1 | 1 |
End of preview. Expand in Data Studio
No dataset card yet
- Downloads last month
- 27