Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[FP6-LLM] Port splitK map from DeepSpeed #283

Merged
merged 2 commits into from
May 29, 2024
Merged

[FP6-LLM] Port splitK map from DeepSpeed #283

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
7aad308
port splitK map from DeepSpeed
gau-nernst May 26, 2024
b89ec10
Merge branch 'main' into fp6_llm_splitk_map
msaroufim May 29, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
147 changes: 145 additions & 2 deletions torchao/quantization/fp6_llm.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,3 +1,4 @@
import math
from typing import Optional

import torch
Expand Down Expand Up @@ -111,6 +112,143 @@ def from_tc_float6_e3m2(tensor: Tensor, M: int, N: int, dtype: torch.dtype = tor
return from_float6_e3m2(tensor_fp6, no_bit_packing=True, dtype=dtype)


# https://github.com/microsoft/DeepSpeed/blob/3a3a6db3332e339cc9fd94efd4982f6d60635a3d/deepspeed/inference/v2/kernels/core_ops/cuda_linear/cuda_linear.py
_SPLIT_K_MAP = [
{ # tokens: [1, 64]
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

n00b q: what is meant by token counts here?

Also can these values be autotuned? I don't necesarily wanna merge in something that's only fast on A100

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think it's the batch size. Yea, it will be great if we can autotune CUDA kernel also.

3072: 18,
4096: 13,
5120: 10,
6144: 9,
8192: 6,
10240: 5,
14336: 7,
28672: 7,
57344: 7
},
{ # tokens: [65:128]
3072: 9,
4096: 6,
5120: 5,
6144: 9,
8192: 3,
10240: 5,
14336: 7,
28672: 7,
57344: 6
},
{ # tokens: [129:192]
3072: 6,
4096: 4,
5120: 7,
6144: 3,
8192: 2,
10240: 5,
14336: 5,
28672: 5,
57344: 4
},
{ # tokens: [193:256]
3072: 9,
4096: 3,
5120: 5,
6144: 2,
8192: 5,
10240: 4,
14336: 8,
28672: 6,
57344: 4
},
{ # tokens: [257:320]
3072: 7,
4096: 5,
5120: 2,
6144: 5,
8192: 4,
10240: 1,
14336: 3,
28672: 3,
57344: 4
},
{ # tokens: [321:384]
3072: 3,
4096: 2,
5120: 5,
6144: 3,
8192: 1,
10240: 8,
14336: 3,
28672: 4,
57344: 3
},
{ # tokens: [385:448]
3072: 5,
4096: 7,
5120: 3,
6144: 5,
8192: 7,
10240: 3,
14336: 1,
28672: 1,
57344: 3
},
{ # tokens: [449:512]
3072: 2,
4096: 5,
5120: 4,
6144: 1,
8192: 5,
10240: 2,
14336: 6,
28672: 4,
57344: 1
},
{ # tokens: [513:576]
3072: 2,
4096: 3,
5120: 1,
6144: 1,
8192: 3,
10240: 3,
14336: 3,
28672: 1,
57344: 1
},
{ # tokens: [577:640]
3072: 5,
4096: 4,
5120: 1,
6144: 4,
8192: 2,
10240: 1,
14336: 1,
28672: 1,
57344: 1
},
{ # tokens: [641:704]
3072: 3,
4096: 1,
5120: 2,
6144: 2,
8192: 1,
10240: 2,
14336: 1,
28672: 1,
57344: 1
},
{ # tokens: [705:768]
3072: 3,
4096: 1,
5120: 3,
6144: 2,
8192: 1,
10240: 1,
14336: 1,
28672: 1,
57344: 1
}
]


class Fp6LlmLinear(nn.Module):
"""FP6-LLM Linear layer as described in https://arxiv.org/pdf/2401.14112.
"""
Expand All @@ -124,12 +262,17 @@ def __init__(self, weight: Tensor, scales: Tensor, bias: Optional[Tensor] = None
self.in_features = weight.shape[1] * 16 // 3

def forward(self, x: Tensor) -> Tensor:
# TODO: splitK map
out = fp16act_fp6weight_linear(x.view(-1, self.in_features).half(), self.weight, self.scales, splitK=1)
splitK = self.get_split_k(math.prod(x.shape[:-1]), self.out_features)
out = fp16act_fp6weight_linear(x.view(-1, self.in_features).half(), self.weight, self.scales, splitK=splitK)
if self.bias is not None:
out = out + self.bias
return out.view(*x.shape[:-1], self.out_features).to(x.dtype)

@staticmethod
def get_split_k(bsize: int, out_dim: int) -> int:
# https://github.com/microsoft/DeepSpeed/blob/3a3a6db3332e339cc9fd94efd4982f6d60635a3d/deepspeed/inference/v2/kernels/core_ops/cuda_linear/cuda_linear.py
return _SPLIT_K_MAP[(bsize - 1) // 64].get(out_dim, 1) if bsize <= 768 else 1

@classmethod
def from_float(cls, linear: nn.Linear):
assert (linear.in_features % 64 == 0) and (linear.out_features % 256 == 0)
Expand Down
Loading