utils: safetensors: dont slice data on torch level (#12266)

Torch has alignment enforcement when viewing with data type changes
but only relative to itself. Do all tensor constructions straight
off the memory-view individually so pytorch doesnt see an alignment
problem.

The is needed for handling misaligned safetensors weights, which are
reasonably common in third party models.

This limits usage of this safetensors loader to GPU compute only
as CPUs kernnel are very likely to bus error. But it works for
dynamic_vram, where we really dont want to take a deep copy and we
always use GPU copy_ which disentangles the misalignment.

comfy/utils.py

@@ -82,14 +82,12 @@ _TYPES = {
 def load_safetensors(ckpt):
     f = open(ckpt, "rb")
     mapping = mmap.mmap(f.fileno(), 0, access=mmap.ACCESS_READ)
+    mv = memoryview(mapping)
 
     header_size = struct.unpack("<Q", mapping[:8])[0]
     header = json.loads(mapping[8:8+header_size].decode("utf-8"))
 
-    with warnings.catch_warnings():
-        #We are working with read-only RAM by design
-        warnings.filterwarnings("ignore", message="The given buffer is not writable")
-        data_area = torch.frombuffer(mapping, dtype=torch.uint8)[8 + header_size:]
+    mv = mv[8 + header_size:]
 
     sd = {}
     for name, info in header.items():
@@ -97,7 +95,13 @@ def load_safetensors(ckpt):
             continue
 
         start, end = info["data_offsets"]
-        sd[name] = data_area[start:end].view(_TYPES[info["dtype"]]).view(info["shape"])
+        if start == end:
+            sd[name] = torch.empty(info["shape"], dtype =_TYPES[info["dtype"]])
+        else:
+            with warnings.catch_warnings():
+                #We are working with read-only RAM by design
+                warnings.filterwarnings("ignore", message="The given buffer is not writable")
+                sd[name] = torch.frombuffer(mv[start:end], dtype=_TYPES[info["dtype"]]).view(info["shape"])
 
     return sd, header.get("__metadata__", {}),