Upload folder using huggingface_hub

bitsandbytes_mps/bnb_quantized.mm

@@ -120,8 +120,8 @@ std::tuple<at::Tensor, at::Tensor> bnb_quantize_4bit(
     int64_t quant_type) {
   TORCH_CHECK(input.is_mps(), "Input must be on MPS device");
   TORCH_CHECK(
-      blocksize == 64 || blocksize == 128,
-      "Only blocksize 64 and 128 are supported");
+      blocksize == 64 || blocksize == 128 || blocksize == 256 || blocksize == 512,
+      "Only blocksize 64, 128, 256, and 512 are supported");
   TORCH_CHECK(
       quant_type == 1 || quant_type == 2,
       "quant_type must be 1 (FP4) or 2 (NF4)");
@@ -190,8 +190,8 @@ at::Tensor bnb_dequantize_4bit(
   TORCH_CHECK(packed.is_mps(), "packed must be on MPS device");
   TORCH_CHECK(absmax.is_mps(), "absmax must be on MPS device");
   TORCH_CHECK(
-      blocksize == 64 || blocksize == 128,
-      "Only blocksize 64 and 128 are supported");
+      blocksize == 64 || blocksize == 128 || blocksize == 256 || blocksize == 512,
+      "Only blocksize 64, 128, 256, and 512 are supported");
 
   int n = static_cast<int>(numel);
   int num_blocks =

build/torch210-metal-aarch64-darwin/_bitsandbytes_mps_9811962_dirty.abi3.so

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c9be20185521046ee042d66544cf94fa448c0e1c0455217ec81cef718d264ed9
+oid sha256:28f847d91d0a9ad71b2b191d8c2269e0f2254a1ddee68389133ff17d4f824ca3
 size 845176

build/torch29-metal-aarch64-darwin/_bitsandbytes_mps_9811962_dirty.abi3.so

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:be7a2bbf3cae711200855b297de2f3ba3d47379bf2ce52c61dd6cc3053075047
+oid sha256:6c9ee3bc582c08893dc599413baea1c7d56ffa4f5864f8f14f67b8ddfc5f22b7
 size 844504