Re-enable the 80 char line width limit (#3305)

This commit is contained in:
Zhuohan Li
2024-03-10 19:49:14 -07:00
committed by GitHub
parent 4b59f00e91
commit 2f8844ba08
67 changed files with 557 additions and 528 deletions
+2 -1
View File
@@ -57,7 +57,8 @@ def test_fused_moe(
[torch.float32, torch.float16, torch.bfloat16])
@torch.inference_mode()
def test_mixtral_moe(dtype: torch.dtype):
"Make sure our Mixtral MoE implementation agrees with the one from huggingface."
"""Make sure our Mixtral MoE implementation agrees with the one from
huggingface."""
# Instantiate our and huggingface's MoE blocks
config = MixtralConfig()
+2 -1
View File
@@ -114,7 +114,8 @@ def test_contexted_kv_attention(
v_cache = v_cache.view(-1, block_size, num_kv_heads,
head_size).permute(0, 2, 3, 1).contiguous()
# Warm up the Triton kernel by calling it once before actually measuring generation time
# Warm up the Triton kernel by calling it once before actually measuring
# generation time
context_attention_fwd(query, k, v, output, k_cache, v_cache, block_table,
b_start_loc, b_seq_len, b_ctx_len, max_input_len)
torch.cuda.synchronize()