Update int4pack related for gguf

pytorch · Dec 9, 2024 · 29d4ab7 · 29d4ab7
1 parent fff956c
commit 29d4ab7
Showing 1 changed file with 59 additions and 27 deletions.
diff --git a/torchchat/utils/gguf_loader.py b/torchchat/utils/gguf_loader.py
@@ -24,6 +24,9 @@
     pack_scales_and_zeros,
 )
 
+from torchao.dtypes.utils import is_device
+from torchao.utils import TORCH_VERSION_AT_LEAST_2_6
+
 
 logger: logging.Logger = logging.getLogger(__name__)
 
@@ -122,12 +125,20 @@ def linear_int4(input, weight_int4pack, scales_and_zeros, out_features, groupsiz
             input.dtype
         )  # cast back to input.dtype
     else:
-        c = torch.ops.aten._weight_int4pack_mm(
-            input,
-            weight_int4pack,
-            groupsize,
-            scales_and_zeros,
-        )
+        if TORCH_VERSION_AT_LEAST_2_6:
+            c = torch.ops.aten._weight_int4pack_mm_for_cpu(
+                input,
+                weight_int4pack,
+                groupsize,
+                scales_and_zeros,
+            )
+        else:
+            c = torch.ops.aten._weight_int4pack_mm(
+                input,
+                weight_int4pack,
+                groupsize,
+                scales_and_zeros,
+            )
     new_shape = origin_input_size[:-1] + (out_features,)
     c = c.reshape(new_shape)
     return c
@@ -178,16 +189,27 @@ def __init__(
         ), "must specify both weights and scales_and_zeros, or neither"
 
         if weight is None:
-            weight = torch.empty(
-                (
-                    out_features // 8,
-                    in_features // (inner_k_tiles * 16),
-                    32,
-                    inner_k_tiles // 2,
-                ),
-                dtype=torch.int32,
-                device=device,
-            )
+            if is_device(device, "cpu"):
+                weight = torch.empty(
+                    (
+                        out_features,
+                        in_features // 2,
+                    ),
+                    dtype=torch.uint8,
+                    device=device,
+                )
+            else:
+                weight = torch.empty(
+                    (
+                        out_features // 8,
+                        in_features // (inner_k_tiles * 16),
+                        32,
+                        inner_k_tiles // 2,
+                    ),
+                    dtype=torch.int32,
+                    device=device,
+                )
+
             scales_and_zeros = torch.empty(
                 (in_features // groupsize, out_features, 2),
                 dtype=get_precision(),
@@ -223,12 +245,17 @@ def _prepare_weight_and_scales_and_zeros(
         weight_int32, scales_and_zeros = group_quantize_tensor(
             weight_bf16, n_bit=4, groupsize=groupsize
         )
-        weight_uint8 = (weight_int32[::, ::2] << 4 | weight_int32[::, 1::2]).to(
-            torch.uint8
-        )
-        weight_int4pack = torch.ops.aten._convert_weight_to_int4pack(
-            weight_uint8, inner_k_tiles
-        )
+        if is_device(weight_int32.device.type, "cpu") and TORCH_VERSION_AT_LEAST_2_6:
+            weight_int4pack = torch.ops.aten._convert_weight_to_int4pack_for_cpu(
+                weight_int32, inner_k_tiles
+            )
+        else:
+            weight_uint8 = (weight_int32[::, ::2] << 4 | weight_int32[::, 1::2]).to(
+                torch.uint8
+            )
+            weight_int4pack = torch.ops.aten._convert_weight_to_int4pack(
+                weight_uint8, inner_k_tiles
+            )
         return weight_int4pack, scales_and_zeros
 
     @classmethod
@@ -608,10 +635,15 @@ def load_model_and_state_dict(
             if load_state_dict:
                 q, s, z = Q4_0.unpack(t)
                 scales_and_zeros = pack_scales_and_zeros(s, z)
-                q_uint8 = (q[::, ::2] << 4 | q[::, 1::2]).to(torch.uint8)
-                weight_int4pack = torch.ops.aten._convert_weight_to_int4pack(
-                    q_uint8, inner_k_tiles
-                )
+                if is_device(q.device.type, "cpu") and TORCH_VERSION_AT_LEAST_2_6:
+                    weight_int4pack = torch.ops.aten._convert_weight_to_int4pack_for_cpu(
+                        q, inner_k_tiles
+                    )
+                else:
+                    q_tmp = (q[::, ::2] << 4 | q[::, 1::2]).to(torch.uint8)
+                    weight_int4pack = torch.ops.aten._convert_weight_to_int4pack(
+                        q_tmp, inner_k_tiles
+                    )
                 state_dict[f"{fqn}.weight"] = weight_int4pack
                 state_dict[f"{fqn}.scales_and_zeros"] = scales_and_zeros
 
@@ -623,7 +655,7 @@ def load_model_and_state_dict(
                     in_features=in_features,
                     out_features=out_features,
                     bias=False,
-                    device="meta",
+                    device="cpu",
                     groupsize=Q4_0.groupsize,
                     inner_k_tiles=inner_k_tiles,
                 ),