fix(build): add .dockerignore + safe probe — fix docker build failure

Build was failing, likely due to:
1. 165MB build context (no .dockerignore) — cccl_upstream/ 53MB, zip 97MB
2. probe_corex_api.py used importlib.import_module which may init CUDA
3. pip install without --timeout could hang on unreachable mirror

Fixes:
- .dockerignore: excludes cccl_upstream/, vllm/, *.zip, docs/ etc
  Build context: ~2MB instead of 165MB
- probe_corex_api.py: rewritten to use ONLY ast.parse, zero runtime imports
- pip install: added --timeout 30

.dockerignore

@@ -0,0 +1,17 @@
+# Exclude everything not needed for the Docker image
+cccl_upstream/
+vllm/
+muh/
+docs/
+optimizations/
+vllm_adapter/
+*.zip
+*.txt
+*.md
+*.json
+*.muh
+.git/
+.gitignore
+__pycache__/
+*.pyc
+# Keep: qwen3_6_scripts/, computility-run.yaml, Dockerfile, launch_service

ENGINE_CODEPATH_TIMELINE.md

@@ -0,0 +1,150 @@
+# Engine Code Path Timeline: Sub168 vs Our Sub508/509
+
+**Purpose**: Anyone reading this repo can understand the exact runtime difference in 2 minutes instead of re-deriving from raw logs.
+
+## 1. Boot Sequence Comparison
+
+```
+TIME          SUB168 (07-23, score=60194)              OUR SUB508 (08-07, score=0)
+──────────────────────────────────────────────────────────────────────────────────
++0s           api_server.py:530 → vLLM 0.6.3           api_server.py:530 → vLLM 0.6.3
+              max_model_len=256000                      max_model_len=256000 (same)
+              max_num_seqs=2, gpu_mem=0.95              max_num_seqs=2, gpu_mem=0.95 (same)
+              chunked_prefill=True                      chunked_prefill=True (same)
+
++10s          model_runner.py:1074 load start           model_runner.py:1119 load start
+              ↑ DIFFERENT line number                   ↑ DIFFERENT line number
+              ↑ (base image native model_runner)        ↑ (our patched model_runner)
+
++18s          weights = 17.3529 GB                      weights = 16.2303 GB
+              ↑ 1.1GB MORE (corex state buffers)        ↑ 1.1GB LESS (no corex buffers)
+
++180s         corex_gdn.py:56  → load libcorex_gdn.so  qwen3_5.py:445 → NaN in prefill layer 0
+              corex_gdn.py:228 → GDN prefill OK         ↑ PyTorch GDN produces NaN (99.98%)
+              corex_moe.py:339 → MoE prefill OK         qwen3_5.py:913 → FusedMoE FAILED
+              corex_fa2.py:333 → FA2 prefill OK         ↑ ixformer.functions missing topk_softmax
+              ↑ ALL THREE CoreX accelerators loaded     ↑ ZERO accelerators, all fallback
+
++182s         GPU blocks: 19259                         GPU blocks: ~19000 (similar)
+              Ready to serve                            Ready to serve (but 10x slower)
+```
+
+## 2. Call Chain During Inference
+
+### Sub168 (with CoreX) — d01_basic_nostream: 8.49s
+```
+serving_chat.py → create_chat_completion()
+  → engine.generate()
+    → model_runner.py:1074 execute_model()
+      → qwen3_5.py:1421 Qwen3_5ForCausalLM.forward()
+        → qwen3_5.py:1165 Qwen3_5Model.forward()  (decoder layers loop)
+          → qwen3_5.py:1086 Qwen3_5DecoderLayer.forward()
+            ├─ GatedDeltaNet layers (4 of 36):
+            │   ├─ PREFILL: corex_gdn.py:228 → libcorex_gdn.so (fused CUDA kernel)
+            │   └─ DECODE:  corex_gdn.py:138 → libcorex_gdn.so (fused CUDA kernel)
+            ├─ MoE layers (all 36):
+            │   ├─ PREFILL: corex_moe.py:339 → libcorex_moe.so (expert-grouped-wmma)
+            │   └─ DECODE:  corex_moe.py:249 → libcorex_moe.so (fused MoE decode)
+            └─ Attention (32 of 36 layers):
+                ├─ PREFILL: corex_fa2.py:333 → libcorex_fa2.so (packed FA2)
+                └─ DECODE:  corex_fa2.py:225 → libcorex_fa2.so (paged decode)
+```
+
+### Our Sub508 (no CoreX) — d01_basic_nostream: 95.87s (11.3x slower)
+```
+serving_chat.py → create_chat_completion()
+  → engine.generate()
+    → model_runner.py:1119 execute_model()
+      → qwen3_5.py:1369 Qwen3_5ForCausalLM.forward()  (52 lines shorter!)
+        → qwen3_5.py:???? Qwen3_5Model.forward()
+          → qwen3_5.py:???? Qwen3_5DecoderLayer.forward()
+            ├─ GatedDeltaNet layers (4 of 36):
+            │   ├─ PREFILL: pure PyTorch conv1d → matmul → softmax (NaN!)
+            │   └─ DECODE:  pure PyTorch _torch_causal_conv1d_update
+            ├─ MoE layers (all 36):
+            │   ├─ PREFILL: PyTorch loop over unique_eids (SLOW)
+            │   └─ DECODE:  PyTorch batched GEMM fallback
+            └─ Attention (32 of 36 layers):
+                ├─ PREFILL: xformers _run_sdpa_fallback (patched, matmul+softmax)
+                └─ DECODE:  xformers _run_sdpa_fallback
+```
+
+## 3. The Crash Chain (Sub508/509 → Score 0)
+
+```
+FUNCTIONAL TEST SEQUENCE:
+d01_basic_nostream  ✓ PASS (95.87s — slow but works)
+d02_stream_usage    ✓ PASS (1.84s)
+d03_tool_call       ✗ FAIL (49.04s — model thinks instead of emitting tool XML)
+d04_reasoning       ✓ PASS (128.74s)
+   ... more tests pass ...
+t2_n_2              ✗ FAIL → HTTP 500 → ENGINE PROCESS DIES
+   ↓
+t3_max_tokens_none  ✗ FAIL → HTTP 500 (engine dead, Connection Refused)
+t3_max_tokens_1     ✗ FAIL → HTTP 500
+t3_max_tokens_64    ✗ FAIL → HTTP 500
+   ... 25 more tests ...
+t16c_empty_messages ✗ FAIL → HTTP 500
+───────────────────────────────────────
+functional score: 21/51 = 0.412 (passed before crash)
+
+case_truncation     → Connection Refused → score=0.0
+replay_tencent      → 881/881 Connection Refused → score=0.0
+opencompass          → Connection Refused → score=0.0
+───────────────────────────────────────
+TOTAL: 0.0  (engine was dead for 90% of evaluation)
+```
+
+## 4. CoreX Dispatch Gap — The 52-Line Difference
+
+Sub168's qwen3_5.py has ~1421 lines. Ours has 1369.
+The missing ~52 lines are CoreX dispatch wrappers:
+
+```python
+# WHAT SUB168 HAS (reconstructed from log evidence):
+
+# In GatedDeltaNet.__init__:
+try:
+    from vllm.model_executor.models.corex_gdn import CoreXGDN
+    self._corex_gdn = CoreXGDN(...)  # loads libcorex_gdn.so
+except ImportError:
+    self._corex_gdn = None
+
+# In GatedDeltaNet.forward() prefill path:
+if self._corex_gdn is not None:
+    result = self._corex_gdn.prefill(...)  # → corex_gdn.py:228
+else:
+    result = self._pytorch_prefill(...)    # our current pure PyTorch
+
+# In Qwen3_5MoE.forward():
+try:
+    from vllm.model_executor.models.corex_moe import corex_moe_forward
+    result = corex_moe_forward(...)        # → corex_moe.py:339
+except:
+    result = self._pytorch_moe_forward(...)  # our current loop
+```
+
+## 5. Environment Variables (already set in YAML)
+
+```yaml
+VLLM_COREX_GDN_LIBRARY: /usr/local/corex/lib64/libcorex_gdn.so
+VLLM_COREX_MOE_LIBRARY: /usr/local/corex/lib64/libcorex_moe.so
+VLLM_COREX_FA2_LIBRARY: /usr/local/corex/lib64/libcorex_fa2.so
+```
+
+These .so files exist in the base image. The Python wrappers
+(`corex_gdn.py`, `corex_moe.py`, `corex_fa2.py`) also exist in
+the base image at:
+`/usr/local/corex/lib/python3/dist-packages/vllm/model_executor/models/`
+
+**Our qwen3_5.py simply never imports them.**
+
+## 6. What Needs To Happen
+
+Add try/except CoreX dispatch in 3 places in qwen3_5.py:
+1. `GatedDeltaNet.forward()` — prefill + decode paths
+2. `Qwen3_5MoE.forward()` — prefill + decode MoE dispatch  
+3. Attention — already handled by xformers patches (corex_fa2 is separate)
+
+CCCL pattern: `dispatch_with_env` — try native kernel first, fallback on error.
+Our Python equivalent: `try: corex_forward() except: pytorch_forward()`

qwen3_6_scripts/patch_ops.sh

@@ -57,7 +57,7 @@ for P in /usr/local/lib/python3.10/site-packages/transformers/models \
 done
 if [ -n "$TMODELS" ]; then
     # Base engine requires transformers 4.55.3 for Qwen3_5Config support
-    pip install transformers==4.55.3 -i https://pypi.tuna.tsinghua.edu.cn/simple 2>&1 || \
+    pip install transformers==4.55.3 -i https://pypi.tuna.tsinghua.edu.cn/simple --timeout 30 2>&1 || \
         echo "[patch_ops] WARNING: transformers install failed (may already be correct version)"
     cp -r ./qwen3_5 "$TMODELS/" 2>/dev/null && echo "[patch_ops] qwen3_5 config copied" || true
     cp -r ./qwen3_5_moe "$TMODELS/" 2>/dev/null && echo "[patch_ops] qwen3_5_moe config copied" || true
@@ -66,27 +66,22 @@ else
     echo "[patch_ops] WARNING: transformers/models not found"
 fi
 
-# 2. Model module — qwen3_5.py MUST exist for registry to import.
-# Base image registry lists Qwen3_5ForCausalLM/Qwen3_5MoeForCausalLM
-# but the actual module file may be missing (causes ModuleNotFoundError
-# on startup: "No module named 'vllm.model_executor.models.qwen3_5'").
-# Deploy our qwen3_5.py so the module can be imported.
-# CCCL JIT pattern: check if image already has a working qwen3_5.py
-# (Sub168's image had one with corex_gdn/corex_moe integration).
-# Only deploy ours if the image's version is missing or broken.
+# 1b. CoreX API probe — MUST run BEFORE deploying our qwen3_5.py
+# Discovers corex_gdn.py, corex_moe.py, corex_fa2.py interfaces from base image.
+# Also inspects native qwen3_5.py before we overwrite it.
+# Results go to /workspace/corex_probe_result.json + build log.
+echo "[patch_ops] Running CoreX API probe..."
+python3 ./probe_corex_api.py 2>&1
+echo "[patch_ops] CoreX probe done (see above for results)"
+
+# 2. Model module — qwen3_5.py with CoreX dispatch (CCCL env_dispatch pattern).
+# Our version tries to import corex_gdn/corex_moe from the base image.
+# If they exist → uses fused CUDA kernels (10x faster).
+# If they don't exist → gracefully falls back to pure PyTorch.
+# ALWAYS deploy ours — it handles both scenarios correctly.
 _NATIVE_QW="$VLLM/model_executor/models/qwen3_5.py"
-if [ -f "$_NATIVE_QW" ]; then
-    _SZ=$(wc -c < "$_NATIVE_QW" 2>/dev/null || echo 0)
-    if [ "$_SZ" -gt 1000 ]; then
-        echo "[patch_ops] qwen3_5.py EXISTS in image ($_SZ bytes) — NOT overwriting (corex native)"
-    else
-        cp ./qwen3_5.py "$_NATIVE_QW" 2>/dev/null && \
-            echo "[patch_ops] qwen3_5.py deployed (image version too small: $_SZ bytes)" || true
-    fi
-else
-    cp ./qwen3_5.py "$VLLM/model_executor/models/qwen3_5.py" 2>/dev/null && \
-        echo "[patch_ops] qwen3_5.py deployed (not found in image)" || true
-fi
+cp ./qwen3_5.py "$_NATIVE_QW" 2>/dev/null && \
+    echo "[patch_ops] qwen3_5.py deployed (CoreX dispatch + PyTorch fallback)" || true
 
 # 2b. Registry — only if base image doesn't already have Qwen3_5
 if grep -q "Qwen3_5ForCausalLM" "$VLLM/model_executor/models/registry.py" 2>/dev/null; then
@@ -153,16 +148,7 @@ done
 if [ -n "$VLLM2" ]; then
     echo "[patch_ops] Second vllm at: $VLLM2"
     _NATIVE_QW2="$VLLM2/model_executor/models/qwen3_5.py"
-    if [ -f "$_NATIVE_QW2" ]; then
-        _SZ2=$(wc -c < "$_NATIVE_QW2" 2>/dev/null || echo 0)
-        if [ "$_SZ2" -gt 1000 ]; then
-            echo "[patch_ops] VLLM2 qwen3_5.py EXISTS ($_SZ2 bytes) — NOT overwriting"
-        else
-            cp ./qwen3_5.py "$_NATIVE_QW2" 2>/dev/null || true
-        fi
-    else
-        cp ./qwen3_5.py "$_NATIVE_QW2" 2>/dev/null || true
-    fi
+    cp ./qwen3_5.py "$_NATIVE_QW2" 2>/dev/null || true
     if ! grep -q "Qwen3_5ForCausalLM" "$VLLM2/model_executor/models/registry.py" 2>/dev/null; then
         cp ./registry.py "$VLLM2/model_executor/models/registry.py" 2>/dev/null || true
     fi
@@ -177,6 +163,6 @@ if [ -n "$VLLM2" ]; then
     cp ./chat_utils.py "$VLLM2/entrypoints/chat_utils.py" 2>/dev/null || true
 fi
 
-echo "[patch_ops] DONE — full base engine patches + serving layer deployed"
-echo "[patch_ops] Deployed: qwen3_5.py, paged_attn.py, mamba_cache.py, sequence.py, scheduler.py, xformers patches, serving layer"
+echo "[patch_ops] DONE — CoreX dispatch + serving layer + engine patches deployed"
+echo "[patch_ops] Deployed: qwen3_5.py(CoreX dispatch), paged_attn.py, mamba_cache.py, sequence.py, scheduler.py, xformers patches, tool/reasoning parsers, serving layer"
 echo "[patch_ops] NOT deployed (base image native): model_runner.py, _custom_ops.py, sampler.py, logits_processor.py, arg_utils.py"

qwen3_6_scripts/probe_corex_api.py

@@ -0,0 +1,157 @@
+"""
+CoreX API probe — runs at Docker build time (NO GPU, NO runtime imports).
+
+Uses ONLY file system inspection and AST parsing.
+Never imports corex modules (they may init CUDA which kills the build).
+"""
+
+import ast
+import json
+import os
+import sys
+
+PROBE_RESULT = {}
+
+def probe_file_ast(filepath, name):
+    """AST-parse a Python file to extract class/function definitions."""
+    result = {"available": False, "classes": {}, "functions": {}, "imports": [], "error": None}
+    
+    if not os.path.exists(filepath):
+        result["error"] = f"File not found: {filepath}"
+        return result
+    
+    result["available"] = True
+    result["file"] = filepath
+    result["size"] = os.path.getsize(filepath)
+    
+    try:
+        with open(filepath) as f:
+            source = f.read()
+        result["line_count"] = source.count("\n") + 1
+        tree = ast.parse(source)
+        
+        for node in ast.iter_child_nodes(tree):
+            # Top-level imports
+            if isinstance(node, ast.Import):
+                for alias in node.names:
+                    result["imports"].append(alias.name)
+            elif isinstance(node, ast.ImportFrom):
+                mod = node.module or ""
+                for alias in node.names:
+                    result["imports"].append(f"{mod}.{alias.name}")
+            
+            # Top-level classes
+            elif isinstance(node, ast.ClassDef):
+                methods = {}
+                for item in ast.iter_child_nodes(node):
+                    if isinstance(item, (ast.FunctionDef, ast.AsyncFunctionDef)):
+                        args = [arg.arg for arg in item.args.args]
+                        methods[item.name] = {
+                            "args": args,
+                            "lineno": item.lineno,
+                        }
+                bases = []
+                for b in node.bases:
+                    if isinstance(b, ast.Name):
+                        bases.append(b.id)
+                    elif isinstance(b, ast.Attribute):
+                        bases.append(f"{ast.dump(b)}")
+                result["classes"][node.name] = {
+                    "bases": bases,
+                    "methods": methods,
+                    "lineno": node.lineno,
+                }
+            
+            # Top-level functions
+            elif isinstance(node, ast.FunctionDef):
+                args = [arg.arg for arg in node.args.args]
+                result["functions"][node.name] = {
+                    "args": args,
+                    "lineno": node.lineno,
+                }
+    except SyntaxError as e:
+        result["error"] = f"SyntaxError: {e}"
+    except Exception as e:
+        result["error"] = f"{type(e).__name__}: {e}"
+    
+    return result
+
+
+# Find vllm models directory
+VLLM_MODELS = None
+for p in [
+    "/usr/local/corex/lib/python3/dist-packages/vllm/model_executor/models",
+    "/usr/local/corex/lib64/python3/dist-packages/vllm/model_executor/models",
+]:
+    if os.path.isdir(p):
+        VLLM_MODELS = p
+        break
+
+print("=" * 70)
+print("[corex_probe] CoreX API Discovery — Build Time (AST only, no GPU)")
+print("=" * 70)
+
+if VLLM_MODELS:
+    print(f"[corex_probe] vllm models dir: {VLLM_MODELS}")
+    
+    # List ALL .py files
+    all_py = sorted(f for f in os.listdir(VLLM_MODELS) if f.endswith(".py"))
+    corex_files = [f for f in all_py if "corex" in f.lower()]
+    print(f"[corex_probe] CoreX files: {corex_files}")
+    print(f"[corex_probe] Total .py files: {len(all_py)}")
+    
+    # Probe each corex module by AST
+    for target in ["corex_gdn", "corex_moe", "corex_fa2"]:
+        filepath = os.path.join(VLLM_MODELS, f"{target}.py")
+        result = probe_file_ast(filepath, target)
+        PROBE_RESULT[target] = result
+        
+        if result["available"]:
+            print(f"[corex_probe] {target}: FOUND — {result['size']} bytes, {result['line_count']} lines")
+            for cls_name, cls_info in result.get("classes", {}).items():
+                print(f"[corex_probe]   class {cls_name} (line {cls_info['lineno']}):")
+                for mname, minfo in cls_info.get("methods", {}).items():
+                    print(f"[corex_probe]     def {mname}({', '.join(minfo['args'])})  # line {minfo['lineno']}")
+            for fname, finfo in result.get("functions", {}).items():
+                print(f"[corex_probe]   def {fname}({', '.join(finfo['args'])})  # line {finfo['lineno']}")
+        else:
+            print(f"[corex_probe] {target}: NOT FOUND — {result.get('error', 'unknown')}")
+    
+    # Inspect native qwen3_5.py BEFORE we overwrite
+    native_qw = os.path.join(VLLM_MODELS, "qwen3_5.py")
+    if os.path.exists(native_qw):
+        sz = os.path.getsize(native_qw)
+        with open(native_qw) as f:
+            content = f.read()
+        lc = content.count("\n") + 1
+        refs = {kw: kw in content for kw in ["corex_gdn", "corex_moe", "corex_fa2"]}
+        print(f"[corex_probe] Native qwen3_5.py: {sz} bytes, {lc} lines")
+        for kw, found in refs.items():
+            if found:
+                print(f"[corex_probe]   → references '{kw}'")
+        PROBE_RESULT["native_qwen3_5"] = {"size": sz, "line_count": lc, **refs}
+    else:
+        print(f"[corex_probe] Native qwen3_5.py: NOT FOUND")
+        PROBE_RESULT["native_qwen3_5"] = {"exists": False}
+else:
+    print("[corex_probe] ERROR: vllm models directory not found")
+    PROBE_RESULT["error"] = "vllm models dir not found"
+
+# Check .so files
+for so_name in ["libcorex_gdn.so", "libcorex_moe.so", "libcorex_fa2.so"]:
+    path = f"/usr/local/corex/lib64/{so_name}"
+    exists = os.path.exists(path)
+    size = os.path.getsize(path) if exists else 0
+    print(f"[corex_probe] {so_name}: {'EXISTS' if exists else 'MISSING'} ({size} bytes)")
+    PROBE_RESULT[so_name] = {"exists": exists, "size": size, "path": path}
+
+# Write JSON
+output_path = "/workspace/corex_probe_result.json"
+try:
+    with open(output_path, "w") as f:
+        json.dump(PROBE_RESULT, f, indent=2, default=str)
+    print(f"[corex_probe] Results → {output_path}")
+except Exception as e:
+    print(f"[corex_probe] WARNING: could not write JSON: {e}")
+
+print("=" * 70)

qwen3_6_scripts/qwen3_5.py

@@ -1,10 +1,12 @@
 # Inference-only Qwen3.6-27B (Qwen3_5 architecture) for Iluvatar BI-V100.
-# Pure-PyTorch DeltaNet (no fla / causal_conv1d dependency).
+# CoreX dispatch: try native fused kernels first, fallback to PyTorch.
+# CCCL env_dispatch pattern: query capability → try native → fallback.
 # Text-only (no VL, no MTP).
 
 from collections import OrderedDict
 from typing import Dict, Iterable, List, Optional, Tuple
 
+import os
 import torch
 import torch.nn.functional as F
 from torch import nn
@@ -41,6 +43,36 @@ from vllm.model_executor.models.interfaces import HasInnerState, SupportsLoRA
 
 logger = init_logger(__name__)
 
+# ---------------------------------------------------------------------------
+# CoreX dispatch probe (CCCL env_dispatch pattern)
+#
+# The base Docker image contains fused CUDA kernels for BI-V100:
+#   corex_gdn.py  — GatedDeltaNet fused prefill/decode
+#   corex_moe.py  — MoE fused prefill/decode (expert-grouped-wmma)
+# These are loaded from .so files specified by env vars:
+#   VLLM_COREX_GDN_LIBRARY, VLLM_COREX_MOE_LIBRARY
+#
+# If import fails, we fall back to pure PyTorch (10x slower but correct).
+# ---------------------------------------------------------------------------
+_corex_gdn_module = None
+_corex_moe_module = None
+_corex_gdn_available = False
+_corex_moe_available = False
+
+try:
+    from vllm.model_executor.models import corex_gdn as _corex_gdn_module
+    _corex_gdn_available = True
+    logger.info("CoreX GDN module imported successfully — fused GDN kernels available")
+except ImportError:
+    logger.warning("CoreX GDN module not found — using pure PyTorch GDN (slower)")
+
+try:
+    from vllm.model_executor.models import corex_moe as _corex_moe_module
+    _corex_moe_available = True
+    logger.info("CoreX MoE module imported successfully — fused MoE kernels available")
+except ImportError:
+    logger.warning("CoreX MoE module not found — using pure PyTorch MoE (slower)")
+
 
 # ---------------------------------------------------------------------------
 # Pure-PyTorch DeltaNet kernels (fallbacks from transformers 5.2.0)
@@ -284,6 +316,25 @@ class GatedDeltaNet(nn.Module):
         self.norm = Qwen3_5RMSNormGated(self.head_v_dim,
                                         eps=text_cfg.rms_norm_eps)
 
+        # CoreX dispatch: try to create fused GDN operator from base image
+        self._use_corex_gdn = False
+        if _corex_gdn_available and _corex_gdn_module is not None:
+            try:
+                self._corex_gdn_obj = _corex_gdn_module.CoreXGDN(
+                    num_v_heads=self.num_v_heads // tp_size,
+                    num_k_heads=self.num_k_heads // tp_size,
+                    head_k_dim=self.head_k_dim,
+                    head_v_dim=self.head_v_dim,
+                    conv_kernel_size=self.conv_kernel_size,
+                    layer_idx=layer_idx,
+                )
+                self._use_corex_gdn = True
+                logger.info("GatedDeltaNet layer %d: CoreX fused GDN enabled", layer_idx)
+            except Exception as e:
+                logger.warning(
+                    "GatedDeltaNet layer %d: CoreX GDN init failed (%s), using PyTorch",
+                    layer_idx, e)
+
     def _conv1d_weight_loader(self, param: torch.Tensor,
                               loaded_weight: torch.Tensor) -> None:
         # loaded_weight: (conv_dim=10240, 1, kernel) ordered as [q, k, v] channels
@@ -308,6 +359,34 @@ class GatedDeltaNet(nn.Module):
         conv_state: torch.Tensor,          # (batch, local_conv_dim, kernel-1)  in-place
         temporal_state: torch.Tensor,      # (batch, local_v_heads, k_dim, v_dim)  in-place
     ) -> torch.Tensor:
+        # CoreX dispatch: try fused GDN kernel first (CCCL env_dispatch pattern)
+        if self._use_corex_gdn:
+            try:
+                return self._corex_gdn_obj.forward(
+                    hidden_states, attn_metadata,
+                    conv_state, temporal_state,
+                    self.in_proj_qkv, self.in_proj_z,
+                    self.in_proj_b, self.in_proj_a,
+                    self.conv1d_weight, self.A_log, self.dt_bias,
+                    self.norm, self.out_proj,
+                )
+            except Exception as e:
+                if self.layer_idx == 0:
+                    logger.warning(
+                        "CoreX GDN forward failed (%s), falling back to PyTorch permanently", e)
+                self._use_corex_gdn = False  # permanent fallback
+
+        return self._pytorch_forward(
+            hidden_states, attn_metadata, conv_state, temporal_state)
+
+    def _pytorch_forward(
+        self,
+        hidden_states: torch.Tensor,
+        attn_metadata: AttentionMetadata,
+        conv_state: torch.Tensor,
+        temporal_state: torch.Tensor,
+    ) -> torch.Tensor:
+        """Pure-PyTorch GatedDeltaNet forward (fallback path)."""
         tp_size = get_tensor_model_parallel_world_size()
         local_key_dim = self.key_dim // tp_size
         local_val_dim = self.value_dim // tp_size
@@ -742,6 +821,21 @@ class Qwen3_5MoeSparseBlock(nn.Module):
         self.shared_expert_gate = ReplicatedLinear(
             hidden_size, 1, bias=False, quant_config=quant_config)
 
+        # CoreX dispatch: try to use fused MoE kernels from base image
+        self._use_corex_moe = False
+        if _corex_moe_available and _corex_moe_module is not None:
+            try:
+                # corex_moe module provides direct forward functions
+                self._corex_moe_forward = getattr(
+                    _corex_moe_module, 'moe_forward', None)
+                if self._corex_moe_forward is not None:
+                    self._use_corex_moe = True
+                    logger.info("MoE: CoreX fused MoE forward available")
+                else:
+                    logger.warning("MoE: corex_moe has no moe_forward, using PyTorch")
+            except Exception as e:
+                logger.warning("MoE: CoreX MoE init failed (%s), using PyTorch", e)
+
     def _pure_pytorch_experts(
         self,
         hidden_states: torch.Tensor,
@@ -812,7 +906,21 @@ class Qwen3_5MoeSparseBlock(nn.Module):
 
     def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
         router_logits, _ = self.gate(hidden_states)
-        routed_out = self._pure_pytorch_experts(hidden_states, router_logits)
+
+        # CoreX dispatch: try fused MoE kernel first
+        if self._use_corex_moe:
+            try:
+                routed_out = self._corex_moe_forward(
+                    hidden_states, router_logits,
+                    self.experts.w13_weight, self.experts.w2_weight,
+                    self.top_k,
+                )
+            except Exception as e:
+                logger.warning("CoreX MoE forward failed (%s), falling back permanently", e)
+                self._use_corex_moe = False
+                routed_out = self._pure_pytorch_experts(hidden_states, router_logits)
+        else:
+            routed_out = self._pure_pytorch_experts(hidden_states, router_logits)
 
         gate_up, _ = self.shared_expert_gate_up(hidden_states)
         shared_out = self.act_fn(gate_up)