-
Notifications
You must be signed in to change notification settings - Fork 5
Expand file tree
/
Copy pathconfig.yaml
More file actions
executable file
·104 lines (95 loc) · 3.67 KB
/
config.yaml
File metadata and controls
executable file
·104 lines (95 loc) · 3.67 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
# ============================================================================
# Agent & Task Configuration
# ============================================================================
# Agent Template Selection
# Choose ONE agent template that defines how the LLM interacts with the system
# Available options:
# - cursor
# - codex
# - geak_optimagentv2
# - single_llm_call
# - SWE-agent
# - claude_code
# - geak_hip
# - openevolve
# - geak_ourllm_kernel2kernel
# - task_validator
agent:
template: cursor
# Task Selection
# Specify task names to execute (use full path from tasks/ directory)
# Examples:
# - hip2hip # All tasks under hip2hip
# - triton2triton # All tasks under triton2triton
# - torch2hip # All tasks under torch2hip
# - hip2hip/silu # single task under hip2hip
# - all # Run ALL available tasks
tasks:
# hip2hip: 20 tasks (18 from gpumode, 2 from others)
- hip2hip/gpumode/CrossEntropyLossLabelSmoothing
- hip2hip/gpumode/Feedforward
- hip2hip/gpumode/FusedLeakyReLU
- hip2hip/gpumode/GateGRUSelectionLayer
- hip2hip/gpumode/Gather
- hip2hip/gpumode/GELU
- hip2hip/gpumode/InnerProd
- hip2hip/gpumode/ItemQueryAttention
- hip2hip/gpumode/KDLoss
- hip2hip/gpumode/layer_normalization
- hip2hip/gpumode/MaskedLanguageModel
- hip2hip/gpumode/MLP_model
- hip2hip/gpumode/MultiHeadAttention
- hip2hip/gpumode/NormalAttention_dot
- hip2hip/gpumode/NormalAttention_embedded_gaussian
- hip2hip/gpumode/PositionEmbedder
- hip2hip/gpumode/Sigmoid
- hip2hip/gpumode/SiLU
- hip2hip/others/assign_score_withk
- hip2hip/others/ball_query
# triton2triton/vllm: 10 tasks
- triton2triton/vllm/triton_awq_dequantize
- triton2triton/vllm/triton_awq_gemm
- triton2triton/vllm/triton_batched_moe
- triton2triton/vllm/triton_flash_prefill_attention
- triton2triton/vllm/triton_fused_moe
- triton2triton/vllm/triton_layernorm_gated
- triton2triton/vllm/triton_matmul_persistent
- triton2triton/vllm/triton_rms_norm
- triton2triton/vllm/triton_scaled_mm
- triton2triton/vllm/triton_topk_log_softmax
# triton2triton/rocmbench: 10 tasks
- triton2triton/rocmbench/easy/test_add_kernel
- triton2triton/rocmbench/easy/test_batched_vecmat
- triton2triton/rocmbench/easy/test_block_copy
- triton2triton/rocmbench/easy/test_kernel_dot
- triton2triton/rocmbench/medium/layernorm
- triton2triton/rocmbench/medium/naive_softmax
- triton2triton/rocmbench/medium/rmsnorm_fwd
- triton2triton/rocmbench/medium/softmax
- triton2triton/rocmbench/hard/gemm
- triton2triton/rocmbench/hard/moe_gemm
# torch2hip: 10 tasks
- torch2hip/gpumode/1001_NormalAttention_dot
- torch2hip/gpumode/10024_Feedforward
- torch2hip/gpumode/1003_NormalAttention_embedded_gaussian
- torch2hip/gpumode/10082_SoftmaxModule
- torch2hip/gpumode/10099_Gather
- torch2hip/gpumode/10190_FusedLeakyReLU
- torch2hip/gpumode/10456_MultiHeadAttention
- torch2hip/gpumode/11754_layer_normalization
- torch2hip/gpumode/14539_GELU
- torch2hip/gpumode/16636_SiLU
# instruction2triton/rocmbench: 10 tasks
- instruction2triton/rocmbench/gemm
- instruction2triton/rocmbench/layernorm
- instruction2triton/rocmbench/moe_gemm
- instruction2triton/rocmbench/multreduce_matmul_dot_kernel
- instruction2triton/rocmbench/naive_softmax
- instruction2triton/rocmbench/rmsnorm_bwd
- instruction2triton/rocmbench/rmsnorm_fwd
- instruction2triton/rocmbench/softmax
- instruction2triton/rocmbench/test_flashattention_fwd
- instruction2triton/rocmbench/test_gemm_fusion
target_gpu_model: MI300
log_directory: logs
workspace_directory_prefix: workspace