dylanyunlong/project_6_89d52222
1
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

[fix] ix_moe_bridge pybind名无ix_前缀: ix_paged_attention→paged_attention, ix_linear→linear

Browse Source
This commit is contained in:
root
2026-08-18 06:14:15 +00:00
parent 54a9b572af
commit b89bd3d72f
2 changed files with 4 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
ex_engine/python/xllm_ops.py
View File

@@ -131,7 +131,7 @@ def paged_attention(out, query, key_cache, value_cache,
block_size, max_context_len, alibi_slopes=None): block_size, max_context_len, alibi_slopes=None):
"""Paged attention decode. .so export: ix_paged_attention in ix_moe_bridge.so.""" """Paged attention decode. .so export: ix_paged_attention in ix_moe_bridge.so."""
bridge = _get("ix_moe_bridge") bridge = _get("ix_moe_bridge")
return bridge.ix_paged_attention( return bridge.paged_attention(
out, query, key_cache, value_cache, out, query, key_cache, value_cache,
scale, block_tables, context_lens, scale, block_tables, context_lens,
block_size, max_context_len, max_context_len, alibi_slopes block_size, max_context_len, max_context_len, alibi_slopes
@@ -166,7 +166,7 @@ def moe_compute_token_index(sorted_token_ids, expert_ids, num_tokens_post_padded
def ixformer_linear(input, weight, act_type=0, bias=None, out=None): def ixformer_linear(input, weight, act_type=0, bias=None, out=None):
"""GEMM via ixformer. .so export: ix_linear in ix_moe_bridge.so.""" """GEMM via ixformer. .so export: ix_linear in ix_moe_bridge.so."""
bridge = _get("ix_moe_bridge") bridge = _get("ix_moe_bridge")
return bridge.ix_linear(input, weight, bias) return bridge.linear(input, weight, bias)
# --- Fused QK-Norm + RoPE --- # --- Fused QK-Norm + RoPE ---
def fused_qknorm_rope(query, key, cos_sin_cache, positions, def fused_qknorm_rope(query, key, cos_sin_cache, positions,

4
qwen3_6_scripts/ex_engine/python/xllm_ops.py
View File

@@ -131,7 +131,7 @@ def paged_attention(out, query, key_cache, value_cache,
block_size, max_context_len, alibi_slopes=None): block_size, max_context_len, alibi_slopes=None):
"""Paged attention decode. .so export: ix_paged_attention in ix_moe_bridge.so.""" """Paged attention decode. .so export: ix_paged_attention in ix_moe_bridge.so."""
bridge = _get("ix_moe_bridge") bridge = _get("ix_moe_bridge")
return bridge.ix_paged_attention( return bridge.paged_attention(
out, query, key_cache, value_cache, out, query, key_cache, value_cache,
scale, block_tables, context_lens, scale, block_tables, context_lens,
block_size, max_context_len, max_context_len, alibi_slopes block_size, max_context_len, max_context_len, alibi_slopes
@@ -166,7 +166,7 @@ def moe_compute_token_index(sorted_token_ids, expert_ids, num_tokens_post_padded
def ixformer_linear(input, weight, act_type=0, bias=None, out=None): def ixformer_linear(input, weight, act_type=0, bias=None, out=None):
"""GEMM via ixformer. .so export: ix_linear in ix_moe_bridge.so.""" """GEMM via ixformer. .so export: ix_linear in ix_moe_bridge.so."""
bridge = _get("ix_moe_bridge") bridge = _get("ix_moe_bridge")
return bridge.ix_linear(input, weight, bias) return bridge.linear(input, weight, bias)
# --- Fused QK-Norm + RoPE --- # --- Fused QK-Norm + RoPE ---
def fused_qknorm_rope(query, key, cos_sin_cache, positions, def fused_qknorm_rope(query, key, cos_sin_cache, positions,