fount-vlm-nell-02 0.3.11__py3-none-any.whl

mlx_vlm/models/paddleocr_vl/vision.py

@@ -0,0 +1,358 @@
+from typing import Optional
+
+import mlx.core as mx
+import mlx.nn as nn
+
+from ..interpolate import bilinear_interpolate
+from .config import VisionConfig
+
+
+def check_array_shape(arr):
+    out_channels, kH, KW, t = arr.shape
+
+    if t == 3:
+        return True
+
+    # Check if out_channels is the largest, and kH and KW are the same
+    if (out_channels >= kH) and (out_channels >= KW) and (kH == KW):
+        return True
+    else:
+        return False
+
+
+def rotate_half(x):
+    """Rotates half the hidden dims of the input."""
+    x1 = x[..., : x.shape[-1] // 2]
+    x2 = x[..., x.shape[-1] // 2 :]
+    return mx.concatenate([-x2, x1], axis=-1)
+
+
+def apply_rotary_pos_emb_vision(tensor, freqs) -> mx.array:
+    orig_dtype = tensor.dtype
+
+    cos = mx.cos(freqs)
+    sin = mx.sin(freqs)
+
+    cos = mx.expand_dims(cos, axis=1)  # Equivalent to unsqueeze(1)
+    cos = mx.tile(cos, (1, 1, 2))  # Equivalent to repeat(1, 1, 2)
+    cos = mx.expand_dims(cos, axis=0)  # Equivalent to [None, ...]
+
+    sin = mx.expand_dims(sin, axis=1)  # Equivalent to unsqueeze(1)
+    sin = mx.tile(sin, (1, 1, 2))  # Equivalent to repeat(1, 1, 2)
+    sin = mx.expand_dims(sin, axis=0)  # Equivalent to [None, ...]
+
+    output = (tensor * cos) + (rotate_half(tensor) * sin)
+    return output.astype(orig_dtype)
+
+
+class VisionRotaryEmbedding(nn.Module):
+    def __init__(self, dim: int, theta: float = 10000.0) -> None:
+        super().__init__()
+        self.dim = dim
+        self.theta = theta
+
+    def __call__(self, seqlen: int) -> mx.array:
+        inv_freq = 1.0 / (
+            self.theta ** (mx.arange(0, self.dim, 2, dtype=mx.float32) / self.dim)
+        )
+        seq = mx.arange(seqlen.tolist(), dtype=inv_freq.dtype)
+        freqs = mx.outer(seq, inv_freq)
+        return freqs
+
+
+class PaddleOCRVisionEmbeddings(nn.Module):
+    def __init__(
+        self,
+        patch_size: int = 14,
+        image_size: int = 384,
+        in_channels: int = 3,
+        embed_dim: int = 1152,
+    ) -> None:
+        super().__init__()
+        self.patch_size = patch_size
+        self.in_channels = in_channels
+        self.embed_dim = embed_dim
+
+        self.patch_embedding = nn.Conv2d(
+            in_channels=in_channels,
+            out_channels=embed_dim,
+            kernel_size=patch_size,
+            stride=patch_size,
+        )
+
+        num_patches = (image_size // patch_size) ** 2
+        self.position_embedding = nn.Embedding(num_patches, self.embed_dim)
+
+    def interpolate_pos_encoding(self, height: int, width: int) -> mx.array:
+        # Get the number of positions and embedding dimension
+        num_positions = self.position_embedding.weight.shape[0]
+
+        # Get all position embeddings (this will dequantize if quantized)
+        position_ids = mx.arange(num_positions)
+        patch_pos_embed = self.position_embedding(position_ids)
+        dim = patch_pos_embed.shape[-1]
+
+        # Reshape to 2D grid
+        sqrt_num_positions = int(num_positions**0.5)
+        patch_pos_embed = patch_pos_embed.reshape(
+            1, sqrt_num_positions, sqrt_num_positions, dim
+        )
+
+        # Interpolate to target size
+        patch_pos_embed = bilinear_interpolate(
+            patch_pos_embed[0],
+            height,
+            width,
+        ).astype(patch_pos_embed.dtype)
+        patch_pos_embed = patch_pos_embed.reshape(-1, dim)
+        return patch_pos_embed
+
+    def __call__(self, hidden_states: mx.array, grid_thw: mx.array) -> mx.array:
+        batch_size, squence_len, channel, patch_size, patch_size = hidden_states.shape
+        target_dtype = self.patch_embedding.weight.dtype
+        hidden_states = hidden_states.reshape(
+            batch_size * squence_len, channel, patch_size, patch_size
+        )
+        # For MLX-Conv2d
+        hidden_states = hidden_states.transpose(0, 2, 3, 1)
+        patch_embeds = self.patch_embedding(hidden_states).astype(target_dtype)
+        patch_embeds = patch_embeds.transpose(0, 3, 1, 2)
+        embeddings = patch_embeds.flatten(-2).squeeze(-1)
+        embeddings = embeddings.reshape(batch_size, squence_len, -1)
+
+        start = 0
+        embeddings = embeddings.squeeze(0)
+        tmp_embeddings = []
+        for image_grid in grid_thw:
+            t, h, w = image_grid.tolist()
+            end = start + t * h * w
+            image_embeddings = embeddings[start:end, :]
+            position_embedding = self.interpolate_pos_encoding(h, w)
+            image_embeddings = image_embeddings + position_embedding
+            tmp_embeddings.append(image_embeddings)
+            start = end
+        embeddings = mx.concatenate(tmp_embeddings, axis=0)
+
+        return embeddings
+
+
+class PaddleOCRProjector(nn.Module):
+    def __init__(self, dim, context_dim, spatial_merge_size) -> None:
+        super().__init__()
+
+        hidden_size = dim * (spatial_merge_size**2)
+        self.spatial_merge_size = spatial_merge_size
+        self.pre_norm = nn.LayerNorm(dim, eps=1e-6)
+        self.linear_1 = nn.Linear(hidden_size, hidden_size, bias=True)
+        self.act = nn.GELU()
+        self.linear_2 = nn.Linear(hidden_size, context_dim, bias=True)
+
+    def __call__(self, x: mx.array, grid_thw: mx.array) -> mx.array:
+        x_chunks = x.split(grid_thw.prod(axis=1).tolist(), axis=0)
+
+        processed_features = []
+        for x, image_grid in zip(x_chunks, grid_thw):
+            x = self.pre_norm(x)
+            t, h, w = image_grid.tolist()
+            d = x.shape[-1]
+            h_block = h // self.spatial_merge_size
+            w_block = w // self.spatial_merge_size
+
+            x = x.reshape(
+                t, h_block, self.spatial_merge_size, w_block, self.spatial_merge_size, d
+            )
+            x = x.transpose(0, 1, 3, 2, 4, 5)
+            x = x.reshape(
+                t * h_block * w_block,
+                self.spatial_merge_size * self.spatial_merge_size * d,
+            )
+
+            hidden_states = self.linear_1(x)
+            hidden_states = self.act(hidden_states)
+            hidden_states = self.linear_2(hidden_states)
+            processed_features.append(hidden_states)
+
+        return mx.concatenate(processed_features, axis=0)
+
+
+class Attention(nn.Module):
+    def __init__(self, dim: int, num_heads: int = 16) -> None:
+        super().__init__()
+        self.num_heads = num_heads
+        self.head_dim = head_dim = dim // num_heads
+        self.scale = head_dim**-0.5
+        self.qkv = nn.Linear(dim, dim * 3, bias=True)
+        self.out_proj = nn.Linear(dim, dim)
+
+    def __call__(
+        self, x: mx.array, cu_seqlens: mx.array, rotary_pos_emb: mx.array = None
+    ) -> mx.array:
+        seq_length = x.shape[0]
+        qkv = (
+            self.qkv(x).reshape(seq_length, 3, self.num_heads, -1).transpose(1, 0, 2, 3)
+        )
+        q, k, v = mx.split(qkv, 3)
+
+        q = apply_rotary_pos_emb_vision(mx.expand_dims(q, 0), rotary_pos_emb)[0]
+        k = apply_rotary_pos_emb_vision(mx.expand_dims(k, 0), rotary_pos_emb)[0]
+
+        attention_mask = mx.ones((1, seq_length, seq_length), dtype=x.dtype)
+
+        for i in range(1, len(cu_seqlens)):
+            start = int(cu_seqlens[i - 1])
+            end = int(cu_seqlens[i])
+            attention_mask[start:end, start:end] = 0
+
+        q = q.transpose(0, 2, 1, 3)
+        k = k.transpose(0, 2, 1, 3)
+        v = v.transpose(0, 2, 1, 3)
+
+        output = mx.fast.scaled_dot_product_attention(
+            q, k, v, scale=self.scale, mask=attention_mask
+        )
+        output = output.transpose(0, 2, 1, 3)
+        output = output.reshape(seq_length, -1)
+        return self.out_proj(output)
+
+
+class MLP(nn.Module):
+    def __init__(self, dim, hidden_dim):
+        super().__init__()
+        self.activation_fn = nn.GELU(approx="precise")
+        self.fc1 = nn.Linear(dim, hidden_dim)
+        self.fc2 = nn.Linear(hidden_dim, dim)
+
+    def __call__(self, x: mx.array) -> mx.array:
+        x = self.activation_fn(self.fc1(x))
+        x = self.fc2(x)
+        return x
+
+
+class PaddleOCRVisionEncoderLayer(nn.Module):
+    def __init__(self, config: VisionConfig) -> None:
+        super().__init__()
+        self.layer_norm1 = nn.LayerNorm(config.hidden_size, eps=1e-6)
+        self.layer_norm2 = nn.LayerNorm(config.hidden_size, eps=1e-6)
+
+        self.self_attn = Attention(
+            dim=config.hidden_size, num_heads=config.num_attention_heads
+        )
+        self.mlp = MLP(dim=config.hidden_size, hidden_dim=config.intermediate_size)
+
+    def __call__(self, hidden_states, cu_seqlens, rotary_pos_emb) -> mx.array:
+        hidden_states = hidden_states + self.self_attn(
+            self.layer_norm1(hidden_states),
+            cu_seqlens=cu_seqlens,
+            rotary_pos_emb=rotary_pos_emb,
+        )
+        hidden_states = hidden_states + self.mlp(self.layer_norm2(hidden_states))
+        return hidden_states
+
+
+class VisionModel(nn.Module):
+    def __init__(self, config: VisionConfig) -> None:
+        super().__init__()
+        self.config = config
+        self.model_type = config.model_type
+        if self.model_type != "paddleocr_vl":
+            raise ValueError(f"Unsupported model type: {self.model_type}")
+
+        self.embeddings = PaddleOCRVisionEmbeddings(
+            patch_size=config.patch_size,
+            image_size=config.image_size,
+            in_channels=config.num_channels,
+            embed_dim=config.hidden_size,
+        )
+
+        head_dim = config.hidden_size // config.num_attention_heads
+        self.rotary_pos_emb = VisionRotaryEmbedding(head_dim // 2)
+
+        self.layers = [
+            PaddleOCRVisionEncoderLayer(config) for _ in range(config.num_hidden_layers)
+        ]
+        self.post_layernorm = nn.LayerNorm(
+            config.hidden_size, eps=config.layer_norm_eps
+        )
+        self.projector = PaddleOCRProjector(
+            dim=config.hidden_size,
+            context_dim=1024,
+            spatial_merge_size=config.spatial_merge_size,
+        )
+
+    def rot_pos_emb(self, grid_thw):
+        pos_ids = []
+
+        split_hids = []
+        split_wids = []
+        for t, h, w in grid_thw:
+            image_pids = mx.arange(int(t * h * w)) % (h * w)
+            sample_hids = image_pids // w
+            sample_wids = image_pids % w
+            split_hids.append(sample_hids)
+            split_wids.append(sample_wids)
+
+        height_position_ids = mx.concatenate(split_hids, axis=0)
+        width_position_ids = mx.concatenate(split_wids, axis=0)
+
+        pos_ids = mx.stack([height_position_ids, width_position_ids], axis=-1)
+        max_grid_size = mx.max(grid_thw[:, 1:])
+        rotary_pos_emb_full = self.rotary_pos_emb(max_grid_size)
+        rotary_pos_emb_full = rotary_pos_emb_full[pos_ids]
+
+        return rotary_pos_emb_full.reshape(pos_ids.shape[0], -1)
+
+    def __call__(
+        self,
+        hidden_states: mx.array,
+        grid_thw: mx.array,
+        output_hidden_states: Optional[bool] = None,
+    ) -> mx.array:
+        hidden_states = self.embeddings(hidden_states, grid_thw)
+        rotary_pos_emb = self.rot_pos_emb(grid_thw)
+
+        # Assuming grid_thw has shape (batch_size, 3)
+        batch_size = grid_thw.shape[0]
+
+        # Calculate cu_seqlens for each item in the batch
+        cu_seqlens = []
+        for i in range(batch_size):
+            seq_len = grid_thw[i, 1] * grid_thw[i, 2]
+            cu_seqlens.append(mx.repeat(seq_len, grid_thw[i, 0]))
+
+        # Concatenate the cu_seqlens for all items in the batch
+        cu_seqlens = mx.concatenate(cu_seqlens)
+
+        cu_seqlens = mx.cumsum(cu_seqlens.astype(mx.int32), axis=0)
+        cu_seqlens = mx.pad(cu_seqlens, (1, 0), mode="constant", constant_values=0)
+
+        encoder_states = (hidden_states,) if output_hidden_states else None
+        for layer in self.layers:
+            hidden_states = layer(
+                hidden_states, cu_seqlens=cu_seqlens, rotary_pos_emb=rotary_pos_emb
+            )
+            if output_hidden_states:
+                encoder_states = encoder_states + (hidden_states,)
+
+        hidden_states = self.post_layernorm(hidden_states)
+        hidden_states = self.projector(hidden_states, grid_thw)
+        return hidden_states
+
+    def sanitize(self, weights):
+        sanitized_weights = {}
+        for k, v in weights.items():
+            if "position_ids" in k:
+                # Remove unused position_ids
+                continue
+            elif "patch_embedding.weight" in k:
+                # PyTorch conv2d weight tensors have shape:
+                #   [out_channels, in_channels, kH, KW]
+                # MLX conv2d expects the weight be of shape:
+                #   [out_channels, kH, KW, in_channels]
+                if check_array_shape(v):
+                    sanitized_weights[k] = v
+                else:
+                    sanitized_weights[k] = v.transpose(0, 2, 3, 1)
+            else:
+                sanitized_weights[k] = v
+
+        return sanitized_weights

mlx_vlm/models/paligemma/__init__.py

@@ -0,0 +1,4 @@
+from .config import ModelConfig, TextConfig, VisionConfig
+from .language import LanguageModel
+from .paligemma import Model
+from .vision import VisionModel

mlx_vlm/models/paligemma/config.py

@@ -0,0 +1,50 @@
+from dataclasses import dataclass, field
+from typing import List, Optional
+
+from ..base import BaseModelConfig
+
+
+@dataclass
+class ModelConfig(BaseModelConfig):
+    text_config: "TextConfig" = field(default_factory=lambda: TextConfig())
+    vision_config: "VisionConfig" = field(default_factory=lambda: VisionConfig())
+    model_type: str = "paligemma"
+    vocab_size: int = 257152
+    ignore_index: int = -100
+    image_token_index: int = 257152
+    hidden_size: int = 2048
+    pad_token_id: int = 0
+    eos_token_id: Optional[List[int]] = None
+
+
+@dataclass
+class TextConfig(BaseModelConfig):
+    model_type: str = "paligemma"
+    hidden_size: int = 2048
+    num_hidden_layers: int = 18
+    intermediate_size: int = 8192
+    num_attention_heads: int = 16
+    num_key_value_heads: int = 16
+    vocab_size: int = 256000
+    head_dim: int = 256
+    rms_norm_eps: float = 1e-6
+    rope_theta: float = 10000
+    rope_traditional: bool = False
+    attn_logit_softcapping: Optional[float] = None
+    final_logit_softcapping: Optional[float] = None
+    query_pre_attn_scalar: Optional[float] = None
+    max_position_embeddings: int = 4096
+
+
+@dataclass
+class VisionConfig(BaseModelConfig):
+    model_type: str = "siglip_vision_model"
+    num_hidden_layers: int = 27
+    hidden_size: int = 1152
+    intermediate_size: int = 4304
+    num_attention_heads: int = 16
+    patch_size: int = 14
+    projection_dim: int = 2048
+    image_size: int = 224
+    num_channels: int = 3
+    layer_norm_eps: float = 1e-6

mlx_vlm/models/paligemma/language.py

@@ -0,0 +1,253 @@
+from typing import Optional
+
+import mlx.core as mx
+import mlx.nn as nn
+
+from ..base import (
+    LanguageModelOutput,
+    create_attention_mask,
+    scaled_dot_product_attention,
+)
+from ..cache import KVCache
+from .config import TextConfig
+
+
+class RMSNorm(nn.Module):
+    def __init__(self, dims: int, eps: float = 1e-6):
+        super().__init__()
+        self.weight = mx.ones((dims,))
+        self.eps = eps
+
+    def __call__(self, x):
+        return mx.fast.rms_norm(x, 1.0 + self.weight, self.eps)
+
+
+class Attention(nn.Module):
+    def __init__(self, config: TextConfig):
+        super().__init__()
+
+        dim = config.hidden_size
+        self.n_heads = n_heads = config.num_attention_heads
+        self.n_kv_heads = n_kv_heads = config.num_key_value_heads
+        self.model_type = config.model_type
+        self.attn_logit_softcapping = config.attn_logit_softcapping
+        self.repeats = n_heads // n_kv_heads
+        self.head_dim = head_dim = (
+            config.hidden_size // n_heads
+            if self.model_type == "gemma"
+            else config.head_dim
+        )
+        self.scale = (
+            head_dim**-0.5
+            if self.model_type == "gemma"
+            else 1.0 / (config.query_pre_attn_scalar**0.5)
+        )
+
+        self.q_proj = nn.Linear(dim, n_heads * head_dim, bias=False)
+        self.k_proj = nn.Linear(dim, n_kv_heads * head_dim, bias=False)
+        self.v_proj = nn.Linear(dim, n_kv_heads * head_dim, bias=False)
+        self.o_proj = nn.Linear(n_heads * head_dim, dim, bias=False)
+
+        self.rope = nn.RoPE(
+            head_dim,
+            traditional=config.rope_traditional,
+            base=config.rope_theta,
+        )
+
+    def __call__(
+        self,
+        x: mx.array,
+        mask: Optional[mx.array] = None,
+        cache: Optional[KVCache] = None,
+    ) -> mx.array:
+        B, L, D = x.shape
+
+        queries, keys, values = self.q_proj(x), self.k_proj(x), self.v_proj(x)
+
+        # Prepare the queries, keys and values for the attention computation
+        queries = queries.reshape(B, L, self.n_heads, -1).transpose(0, 2, 1, 3)
+        keys = keys.reshape(B, L, self.n_kv_heads, -1).transpose(0, 2, 1, 3)
+        values = values.reshape(B, L, self.n_kv_heads, -1).transpose(0, 2, 1, 3)
+
+        if cache is not None:
+            queries = self.rope(queries, offset=cache.offset)
+            keys = self.rope(keys, offset=cache.offset)
+            keys, values = cache.update_and_fetch(keys, values)
+        else:
+            queries = self.rope(queries)
+            keys = self.rope(keys)
+
+        if self.model_type == "gemma":
+            output = scaled_dot_product_attention(
+                queries, keys, values, cache, scale=self.scale, mask=mask
+            )
+        else:
+            queries = queries * self.scale
+
+            if self.repeats > 1:
+                queries = queries.reshape(
+                    B, self.n_kv_heads, self.repeats, L, self.head_dim
+                )
+                keys = mx.expand_dims(keys, 2)
+                values = mx.expand_dims(values, 2)
+
+            scores = queries @ keys.swapaxes(-1, -2)
+            scores = mx.tanh(scores / self.attn_logit_softcapping)
+            scores *= self.attn_logit_softcapping
+
+            if mask is not None and isinstance(mask, mx.array):
+                scores = scores + mask
+            scores = mx.softmax(scores, precise=True, axis=-1)
+            output = scores @ values
+            if self.repeats > 1:
+                output = output.reshape(B, self.n_heads, L, self.head_dim)
+
+        output = output.transpose(0, 2, 1, 3).reshape(B, L, -1)
+        return self.o_proj(output)
+
+
+class MLP(nn.Module):
+    def __init__(self, dim, hidden_dim, model_type):
+        super().__init__()
+        self.gate_proj = nn.Linear(dim, hidden_dim, bias=False)
+        self.down_proj = nn.Linear(hidden_dim, dim, bias=False)
+        self.up_proj = nn.Linear(dim, hidden_dim, bias=False)
+        self.gelu = nn.GELU() if model_type == "gemma" else nn.GELU(approx="precise")
+
+    def __call__(self, x) -> mx.array:
+        return self.down_proj(self.gelu(self.gate_proj(x)) * self.up_proj(x))
+
+
+class TransformerBlock(nn.Module):
+    def __init__(self, config: TextConfig):
+        super().__init__()
+        self.model_type = config.model_type
+        self.num_attention_heads = config.num_attention_heads
+        self.hidden_size = config.hidden_size
+        self.self_attn = Attention(config)
+        self.mlp = MLP(config.hidden_size, config.intermediate_size, config.model_type)
+        self.input_layernorm = RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.post_attention_layernorm = RMSNorm(
+            config.hidden_size, eps=config.rms_norm_eps
+        )
+        self.config = config
+
+        if config.model_type == "gemma2":
+            self.pre_feedforward_layernorm = RMSNorm(
+                config.hidden_size, eps=config.rms_norm_eps
+            )
+            self.post_feedforward_layernorm = RMSNorm(
+                config.hidden_size, eps=config.rms_norm_eps
+            )
+
+    def __call__(
+        self,
+        x: mx.array,
+        mask: Optional[mx.array] = None,
+        cache: Optional[KVCache] = None,
+    ) -> mx.array:
+        # Self attention block
+        r = self.self_attn(self.input_layernorm(x), mask, cache)
+
+        if self.model_type == "gemma":
+            # Gemma: Post-attention residual connection then MLP
+            h = x + r
+            r = self.mlp(self.post_attention_layernorm(h))
+            out = h + r
+        else:
+            # Gemma2: Normalized residual connections with pre/post norms
+            h = x + self.post_attention_layernorm(r)
+            r = self.mlp(self.pre_feedforward_layernorm(h))
+            out = h + self.post_feedforward_layernorm(r)
+        return out
+
+
+class GemmaModel(nn.Module):
+    def __init__(self, config: TextConfig):
+        super().__init__()
+        self.config = config
+        self.vocab_size = config.vocab_size
+        self.num_hidden_layers = config.num_hidden_layers
+        assert self.vocab_size > 0
+        self.embed_tokens = nn.Embedding(config.vocab_size, config.hidden_size)
+        self.layers = [
+            TransformerBlock(config=config) for _ in range(config.num_hidden_layers)
+        ]
+        self.norm = RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+
+    def __call__(
+        self,
+        inputs: mx.array,
+        inputs_embeds: Optional[mx.array] = None,
+        mask: Optional[mx.array] = None,
+        cache=None,
+    ):
+        # for passing merged input embeddings
+        if inputs_embeds is None:
+            h = self.embed_tokens(inputs)
+        else:
+            h = inputs_embeds
+
+        h *= self.config.hidden_size**0.5
+
+        if cache is None:
+            cache = [None] * len(self.layers)
+
+        if mask is None or cache[0].offset > 0:
+            mask = create_attention_mask(h, cache, return_array=True)
+
+        for layer, c in zip(self.layers, cache):
+            h = layer(h, mask, c)
+
+        return self.norm(h)
+
+
+class LanguageModel(nn.Module):
+    def __init__(self, config: TextConfig):
+        super().__init__()
+        self.config = config
+        self.final_logit_softcapping = config.final_logit_softcapping
+        self.model_type = config.model_type
+        self.model = GemmaModel(config)
+
+        if self.model_type not in ["gemma", "gemma2"]:
+            raise ValueError(
+                f"Model type {self.model_type} not supported. Currently only 'gemma' is supported"
+            )
+
+    def __call__(
+        self,
+        inputs: mx.array,
+        inputs_embeds: Optional[mx.array] = None,
+        mask: Optional[mx.array] = None,
+        cache=None,
+        **kwargs,
+    ):
+        out = self.model(inputs, mask=mask, cache=cache, inputs_embeds=inputs_embeds)
+        out = self.model.embed_tokens.as_linear(out)
+
+        if self.model_type == "gemma2":
+            out = mx.tanh(out / self.final_logit_softcapping)
+            out = out * self.final_logit_softcapping
+        return LanguageModelOutput(logits=out)
+
+    def sanitize(self, weights):
+        return {
+            k: v for k, v in weights.items() if "self_attn.rotary_emb.inv_freq" not in k
+        }
+
+    @property
+    def layers(self):
+        return self.model.layers
+
+    @property
+    def head_dim(self):
+        return (
+            self.config.hidden_size // self.config.num_attention_heads
+            if self.model_type == "gemma"
+            else self.config.head_dim
+        )
+
+    @property
+    def n_kv_heads(self):
+        return self.config.num_key_value_heads