From 4b89576cbf3fa6466e4baa5a69ae0d59da56f249 Mon Sep 17 00:00:00 2001
From: the magician <82004885@qq.com>
Date: Wed, 15 Jul 2026 01:32:21 +0800
Subject: [PATCH] reduce docker build context and match working baseline

---
 .dockerignore               | 11 +++++++++++
 computility-run.yaml        |  1 +
 qwen3_6_scripts/xformers.py | 18 +++++++++---------
 3 files changed, 21 insertions(+), 9 deletions(-)
 create mode 100644 .dockerignore

diff --git a/.dockerignore b/.dockerignore
new file mode 100644
index 0000000..d2dfafc
--- /dev/null
+++ b/.dockerignore
@@ -0,0 +1,11 @@
+.git
+.gitignore
+.dockerignore
+worklogs
+**/__pycache__
+**/*.pyc
+**/*.pyo
+**/*.log
+**/*.pid
+nohup.out
+qwen3_6_scripts/nohup.out
diff --git a/computility-run.yaml b/computility-run.yaml
index a10c207..16569f3 100644
--- a/computility-run.yaml
+++ b/computility-run.yaml
@@ -50,3 +50,4 @@ env:
       value: '1'
     - name: LOOP1_NATIVE
       value: '1'
+
diff --git a/qwen3_6_scripts/xformers.py b/qwen3_6_scripts/xformers.py
index 265c446..06d9998 100644
--- a/qwen3_6_scripts/xformers.py
+++ b/qwen3_6_scripts/xformers.py
@@ -335,10 +335,10 @@ def _get_seq_len_block_table_args(
     on the type of attention operation.
 
     Decoder attn -> select entirely decoder self-attention-related fields
-    Encoder/decoder cross-attn -> select encoder sequence lengths &
+    Encoder/decoder cross-attn -> select encoder sequence lengths & 
                                   cross-attn block-tables fields
     Encoder attn -> select encoder sequence lengths fields & no block tables
-
+    
     Arguments:
 
     * attn_metadata: Attention metadata structure associated with attention op
@@ -384,11 +384,11 @@ class XFormersMetadataBuilder(CommonMetadataBuilder[XFormersMetadata]):
 class XFormersImpl(AttentionImpl[XFormersMetadata]):
     """
     If the input tensors contain prompt tokens, the layout is as follows:
-    |<--------------- num_prefill_tokens ----------------->|
+    |<--------------- num_prefill_tokens ----------------->|	
     |<--prefill_0-->|<--prefill_1-->|...|<--prefill_N-1--->|
 
-    Otherwise, the layout is as follows:
-    |<----------------- num_decode_tokens ------------------>|
+    Otherwise, the layout is as follows:	
+    |<----------------- num_decode_tokens ------------------>|	
     |<--decode_0-->|..........|<--decode_M-1-->|<--padding-->|
 
     Generation tokens can contain padding when cuda-graph is used.
@@ -446,7 +446,7 @@ class XFormersImpl(AttentionImpl[XFormersMetadata]):
         self.head_mapping = torch.repeat_interleave(
             torch.arange(self.num_kv_heads, dtype=torch.int32),
             self.num_queries_per_kv)
-
+        
     def forward(
         self,
         query: torch.Tensor,
@@ -474,10 +474,10 @@ class XFormersImpl(AttentionImpl[XFormersMetadata]):
               (1) key and value tensors were cached during prefill, and
               (2) cross-attention key and value tensors do not grow during
                   decode
-
+        
         A note on how the attn_type (attention type enum) argument impacts
         attention forward() behavior:
-
+    
             * DECODER: normal decoder-only behavior;
                 use decoder self-attention block table
             * ENCODER: no KV caching; pass encoder sequence
@@ -490,7 +490,7 @@ class XFormersImpl(AttentionImpl[XFormersMetadata]):
                 will match encoder sequence lengths, pass encoder sequence
                 attributes to kernel (encoder_seq_lens/encoder_seq_lens_tensor/
                 max_encoder_seq_len)
-
+    
         Args:
             query: shape = [num_tokens, num_heads * head_size]
             key: shape = [num_tokens, num_kv_heads * head_size]