Shauray Singh shauray8
shauray8
/ rms_norm_per.sass
Created
March 23, 2026 11:19
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| code for sm_120a | |
| .target sm_120a | |
| Function : k_rms_16 | |
| .headerflags @"EF_CUDA_ACCELERATORS EF_CUDA_SM120 EF_CUDA_VIRTUAL_SM(EF_CUDA_SM120)" | |
| /*0000*/ LDC R1, c[0x0][0x37c] ; /* 0x0000df00ff017b82 */ | |
| /* 0x000fe20000000800 */ | |
| /*0010*/ S2R R0, SR_TID.X ; /* 0x0000000000007919 */ | |
| /* 0x000e220000002100 */ |
shauray8
/ rms_norm.py
Created
March 21, 2026 22:43
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| import cuda.tile as ct | |
| import torch | |
| from math import ceil | |
| def next_pow_of_2(n): | |
| p = 1 | |
| while p < n: p *= 2 | |
| return p | |
| @ct.kernel( |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| import cutlass | |
| import cutlass.cute as cute | |
| @cute.jit | |
| def compose_layouts(): | |
| # Define two layouts | |
| layout_a = cute.make_layout((6, 2), stride=(1, 7)) | |
| layout_b = cute.make_layout((3, 2), stride=(2, 3)) | |
| composed = cute.composition(layout_a, layout_b) |
shauray8
/ matmul_test.cpp
Created
May 1, 2025 13:13
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| #include <vector> | |
| #include <chrono> | |
| #include <iostream> | |
| // Function to get current time in seconds | |
| double get_time() { | |
| auto now = std::chrono::high_resolution_clock::now(); | |
| return std::chrono::duration<double>(now.time_since_epoch()).count(); | |
| } |
shauray8
/ compile_generated_kernel.py
Created
April 16, 2025 12:10
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| # AOT ID: ['0_inference'] | |
| from ctypes import c_void_p, c_long, c_int | |
| import torch | |
| import math | |
| import random | |
| import os | |
| import tempfile | |
| from math import inf, nan | |
| from torch._inductor.hooks import run_intermediate_hooks | |
| from torch._inductor.utils import maybe_profile |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| With tqdm took 0.698805 seconds | |
| 2062651 function calls (2061588 primitive calls) in 0.752 seconds | |
| Ordered by: internal time | |
| ncalls tottime percall cumtime percall filename:lineno(function) | |
| 1000002 0.350 0.000 0.355 0.000 loading_utils.py:41(update) | |
| 1000001 0.242 0.000 0.597 0.000 loading_utils.py:32(__iter__) | |
| 1 0.102 0.102 0.699 0.699 loading_utils.py:185(with_tqdm_loop) | |
| 58 0.005 0.000 0.005 0.000 {built-in method marshal.loads} |
shauray8
/ graph_break.txt
Created
March 26, 2025 11:01
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Graph break in user code at /workspace/Wan2GP/wan/modules/sage2_core.py:66 | |
| Reason: Unsupported: torch.* op returned non-Tensor int call_function <function device_count at 0x795533c18400> | |
| User code traceback: | |
| File "/workspace/Wan2GP/wan/modules/model.py", line 452, in forward | |
| y = self.self_attn( xlist, seq_lens, grid_sizes,freqs) | |
| File "/usr/local/lib/python3.11/dist-packages/torch/nn/modules/module.py", line 1750, in _call_impl | |
| return forward_call(*args, **kwargs) | |
| File "/usr/local/lib/python3.11/dist-packages/mmgp/offload.py", line 1926, in check_empty_cuda_cache | |
| return previous_method(*args, **kwargs) | |
| File "/workspace/Wan2GP/wan/modules/model.py", line 266, in forward |