From bbd0ba0c259353fc5147a3255af1af34185b9ed2 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 14 May 2026 21:57:04 -0700
Subject: [PATCH] studio/mmproj: skip unwanted GGUF values via seek instead of
 read (#5431)

The previous _skip_gguf_value walked past discarded values with
f.read(n), which allocates and immediately drops a Python bytes
object. For weight GGUFs that carry tokenizer.ggml.tokens (~150K
unicode strings) this wasted ~10 MB of allocation per cold call.

Switch the discard path to f.seek(n, 1). The kernel never has to
copy the bytes into userspace and Python never allocates. Truncation
is now detected on the next read attempt rather than inline (an
out-of-range seek on a regular file is legal and the next read
returns short).

Measured on real downloaded GGUFs (Qwen3.5-4B IQ2_XXS 1.52 GB,
bartowski Qwen3.5-4B IQ2_M 1.70 GB, Qwen3.5-4B-MTP IQ2_M 1.94 GB):

  before:  142 ms cold per weight, ~11 MB read
  after:    90 ms cold per weight, ~4 MB read

Mmproj reads are unaffected (no tokenizer to skip). Cached re-reads
remain ~50 microseconds. All 161 in-tree backend tests + 85 isolated
sandbox tests pass.
---
 studio/backend/utils/models/gguf_metadata.py | 17 ++++++++++-------
 1 file changed, 10 insertions(+), 7 deletions(-)

diff --git a/studio/backend/utils/models/gguf_metadata.py b/studio/backend/utils/models/gguf_metadata.py
index 9aab88329d..5629bac58b 100644
--- a/studio/backend/utils/models/gguf_metadata.py
+++ b/studio/backend/utils/models/gguf_metadata.py
@@ -152,7 +152,9 @@ _FIXED_VTYPE_SIZES: Dict[int, int] = {
 
 
 def _skip_gguf_value(f, vtype: int) -> bool:
-    """Advance past one GGUF value. False on truncation or unknown type."""
+    """Advance past one GGUF value. ``f.seek(.., 1)`` past EOF is legal
+    on a regular file so truncation is detected on the next read; we
+    only return False for unknown types or sanity-bound overflow."""
     if vtype == 8:  # STRING
         slen_bytes = f.read(8)
         if len(slen_bytes) < 8:
@@ -160,7 +162,8 @@ def _skip_gguf_value(f, vtype: int) -> bool:
         slen = struct.unpack("<Q", slen_bytes)[0]
         if slen > 1 << 30:  # 1 GB sanity bound
             return False
-        return len(f.read(slen)) == slen
+        f.seek(slen, 1)
+        return True
     if vtype == 9:  # ARRAY
         head = f.read(12)
         if len(head) < 12:
@@ -176,18 +179,18 @@ def _skip_gguf_value(f, vtype: int) -> bool:
                 slen = struct.unpack("<Q", slen_bytes)[0]
                 if slen > 1 << 30:
                     return False
-                if len(f.read(slen)) != slen:
-                    return False
+                f.seek(slen, 1)
             return True
         sz = _FIXED_VTYPE_SIZES.get(atype)
         if sz is None:
             return False
-        total = sz * alen
-        return len(f.read(total)) == total
+        f.seek(sz * alen, 1)
+        return True
     sz = _FIXED_VTYPE_SIZES.get(vtype)
     if sz is None:
         return False
-    return len(f.read(sz)) == sz
+    f.seek(sz, 1)
+    return True
 
 
 def is_mmproj_by_metadata(meta: Optional[Dict[str, str]]) -> Optional[bool]: