drbh
commited on
Commit
·
7a0d4b3
1
Parent(s):
906ccdf
fix: bump to v0.2.0
Browse files- README.md +1 -1
- compare_example.py +1 -1
- gpt_oss_backward.py +1 -1
- gpt_oss_match.py +1 -1
- perf_plot.py +1 -1
- readme_example.py +1 -1
README.md
CHANGED
|
@@ -60,7 +60,7 @@ torch.cuda.manual_seed_all(42)
|
|
| 60 |
torch.backends.cudnn.deterministic = True
|
| 61 |
torch.backends.cudnn.benchmark = False
|
| 62 |
|
| 63 |
-
yamoe = get_kernel("drbh/yamoe", revision="v0.
|
| 64 |
|
| 65 |
# Configuration
|
| 66 |
batch_size, seq_len, hidden_dim = 16, 256, 2880
|
|
|
|
| 60 |
torch.backends.cudnn.deterministic = True
|
| 61 |
torch.backends.cudnn.benchmark = False
|
| 62 |
|
| 63 |
+
yamoe = get_kernel("drbh/yamoe", revision="v0.2.0")
|
| 64 |
|
| 65 |
# Configuration
|
| 66 |
batch_size, seq_len, hidden_dim = 16, 256, 2880
|
compare_example.py
CHANGED
|
@@ -30,7 +30,7 @@ if load_method == 1:
|
|
| 30 |
elif load_method == 2:
|
| 31 |
yamoe = get_local_kernel(Path("result"), "yamoe")
|
| 32 |
elif load_method == 3:
|
| 33 |
-
yamoe = get_kernel("drbh/yamoe", revision="v0.
|
| 34 |
|
| 35 |
binned_experts_ref = yamoe.vendored.yamoe_ref.binned_experts_ref
|
| 36 |
GptOssExperts = yamoe.vendored.gpt_oss_mlp.GptOssExperts
|
|
|
|
| 30 |
elif load_method == 2:
|
| 31 |
yamoe = get_local_kernel(Path("result"), "yamoe")
|
| 32 |
elif load_method == 3:
|
| 33 |
+
yamoe = get_kernel("drbh/yamoe", revision="v0.2.0")
|
| 34 |
|
| 35 |
binned_experts_ref = yamoe.vendored.yamoe_ref.binned_experts_ref
|
| 36 |
GptOssExperts = yamoe.vendored.gpt_oss_mlp.GptOssExperts
|
gpt_oss_backward.py
CHANGED
|
@@ -19,7 +19,7 @@ if load_method == 1:
|
|
| 19 |
elif load_method == 2:
|
| 20 |
yamoe = get_local_kernel(Path("result"), "yamoe")
|
| 21 |
elif load_method == 3:
|
| 22 |
-
yamoe = get_kernel("drbh/yamoe", revision="v0.
|
| 23 |
|
| 24 |
torch.manual_seed(42)
|
| 25 |
|
|
|
|
| 19 |
elif load_method == 2:
|
| 20 |
yamoe = get_local_kernel(Path("result"), "yamoe")
|
| 21 |
elif load_method == 3:
|
| 22 |
+
yamoe = get_kernel("drbh/yamoe", revision="v0.2.0")
|
| 23 |
|
| 24 |
torch.manual_seed(42)
|
| 25 |
|
gpt_oss_match.py
CHANGED
|
@@ -19,7 +19,7 @@ if load_method == 1:
|
|
| 19 |
elif load_method == 2:
|
| 20 |
yamoe = get_local_kernel(Path("result"), "yamoe")
|
| 21 |
elif load_method == 3:
|
| 22 |
-
yamoe = get_kernel("drbh/yamoe", revision="v0.
|
| 23 |
|
| 24 |
torch.manual_seed(42)
|
| 25 |
|
|
|
|
| 19 |
elif load_method == 2:
|
| 20 |
yamoe = get_local_kernel(Path("result"), "yamoe")
|
| 21 |
elif load_method == 3:
|
| 22 |
+
yamoe = get_kernel("drbh/yamoe", revision="v0.2.0")
|
| 23 |
|
| 24 |
torch.manual_seed(42)
|
| 25 |
|
perf_plot.py
CHANGED
|
@@ -19,7 +19,7 @@ import numpy as np
|
|
| 19 |
# import yamoe
|
| 20 |
# import yamoe.reference as reference
|
| 21 |
|
| 22 |
-
yamoe = get_kernel("drbh/yamoe", revision="v0.
|
| 23 |
reference = yamoe.reference
|
| 24 |
|
| 25 |
# Setup
|
|
|
|
| 19 |
# import yamoe
|
| 20 |
# import yamoe.reference as reference
|
| 21 |
|
| 22 |
+
yamoe = get_kernel("drbh/yamoe", revision="v0.2.0")
|
| 23 |
reference = yamoe.reference
|
| 24 |
|
| 25 |
# Setup
|
readme_example.py
CHANGED
|
@@ -18,7 +18,7 @@ torch.cuda.manual_seed_all(42)
|
|
| 18 |
torch.backends.cudnn.deterministic = True
|
| 19 |
torch.backends.cudnn.benchmark = False
|
| 20 |
|
| 21 |
-
yamoe = get_kernel("drbh/yamoe", revision="v0.
|
| 22 |
|
| 23 |
# Configuration
|
| 24 |
batch_size, seq_len, hidden_dim = 16, 256, 2880
|
|
|
|
| 18 |
torch.backends.cudnn.deterministic = True
|
| 19 |
torch.backends.cudnn.benchmark = False
|
| 20 |
|
| 21 |
+
yamoe = get_kernel("drbh/yamoe", revision="v0.2.0")
|
| 22 |
|
| 23 |
# Configuration
|
| 24 |
batch_size, seq_len, hidden_dim = 16, 256, 2880
|