add some logs for david

dashboard/src/lib/components/HeaderNav.svelte

@@ -95,7 +95,7 @@
     {#if showHome}
       <button
         onclick={handleHome}
-        class="text-sm text-white/70 hover:text-exo-yellow transition-colors tracking-wider uppercase flex items-center gap-2 cursor-pointer"
+        class="text-sm text-exo-light-gray hover:text-exo-yellow transition-colors tracking-wider uppercase flex items-center gap-2 cursor-pointer"
         title="Back to topology view"
       >
         <svg
@@ -116,7 +116,7 @@
     {/if}
     <a
       href="/#/downloads"
-      class="text-sm text-white/70 hover:text-exo-yellow transition-colors tracking-wider uppercase flex items-center gap-2 cursor-pointer"
+      class="text-sm text-exo-light-gray hover:text-exo-yellow transition-colors tracking-wider uppercase flex items-center gap-2 cursor-pointer"
       title="View downloads overview"
     >
       {#if downloadProgress}

src/exo/main.py

@@ -261,13 +261,6 @@ def main():
     if args.offline:
         logger.info("Running in OFFLINE mode — no internet checks, local models only")
 
-    # Set trust_remote_code override env var for runner subprocesses
-    if args.trust_remote_code:
-        os.environ["EXO_TRUST_REMOTE_CODE"] = "1"
-        logger.warning(
-            "--trust-remote-code enabled: models may execute arbitrary code during loading"
-        )
-
     # Set FAST_SYNCH override env var for runner subprocesses
     if args.fast_synch is True:
         os.environ["EXO_FAST_SYNCH"] = "on"
@@ -292,7 +285,6 @@ class Args(CamelCaseModel):
     no_downloads: bool = False
     offline: bool = False
     fast_synch: bool | None = None  # None = auto, True = force on, False = force off
-    trust_remote_code: bool = False
 
     @classmethod
     def parse(cls) -> Self:
@@ -344,11 +336,6 @@ class Args(CamelCaseModel):
             action="store_true",
             help="Run in offline/air-gapped mode: skip internet checks, use only pre-staged local models",
         )
-        parser.add_argument(
-            "--trust-remote-code",
-            action="store_true",
-            help="Allow models to execute custom code during tokenizer loading (security-sensitive, CLI-only)",
-        )
         fast_synch_group = parser.add_mutually_exclusive_group()
         fast_synch_group.add_argument(
             "--fast-synch",

src/exo/worker/engines/mlx/auto_parallel.py

@@ -620,6 +620,34 @@ class DeepSeekShardingStrategy(TensorParallelShardingStrategy):
             if on_layer_loaded is not None:
                 on_layer_loaded(i, total)
 
+            def log_info(stuff: nn.Module, name: str):
+                logger.info(f"Info for {name}:")
+
+                weights = stuff.weight
+                logger.info(f"Weights: {weights.shape} {weights.dtype}")
+
+                if hasattr(stuff, "scales"):
+                    scales = stuff.scales
+                    logger.info(f"Scales: {scales.shape} {scales.dtype}")
+                else:
+                    logger.info("Scales: None")
+
+                if hasattr(stuff, "biases"):
+                    biases = stuff.biases
+                    logger.info(f"Biases: {biases.shape} {biases.dtype}")
+                else:
+                    logger.info("Biases: None")
+
+            if i == 9:
+                if getattr(layer.mlp, "shared_experts", None) is not None:
+                    log_info(
+                        layer.mlp.shared_experts.down_proj, "Shared experts Down Proj"
+                    )
+
+                log_info(layer.mlp.switch_mlp.gate_proj, "Switch MLP Gate Proj")
+
+                log_info(layer.self_attn.o_proj, "Self Attn O_proj")
+
         return model
 
 

src/exo/worker/engines/mlx/constants.py

@@ -13,6 +13,5 @@ KV_CACHE_BITS: int | None = None
 
 DEFAULT_TOP_LOGPROBS: int = 5
 
-# True for built-in models with known model cards; custom models added via API default to False
-# and can be overridden with the --trust-remote-code CLI flag.
+# TODO: We should really make this opt-in, but Kimi requires trust_remote_code=True
 TRUST_REMOTE_CODE: bool = True

src/exo/worker/engines/mlx/utils_mlx.py

@@ -291,14 +291,10 @@ def shard_and_load(
 
 def get_tokenizer(model_path: Path, shard_metadata: ShardMetadata) -> TokenizerWrapper:
     """Load tokenizer for a model shard. Delegates to load_tokenizer_for_model_id."""
-    trust_remote_code = (
-        shard_metadata.model_card.trust_remote_code
-        or os.environ.get("EXO_TRUST_REMOTE_CODE") == "1"
-    )
     return load_tokenizer_for_model_id(
         shard_metadata.model_card.model_id,
         model_path,
-        trust_remote_code=trust_remote_code,
+        trust_remote_code=shard_metadata.model_card.trust_remote_code,
     )