generation works but outputs gibberish

llms/mlx_lm/models/cache.py

@@ -338,30 +338,3 @@ class MambaCache(_BaseCache):
     @state.setter
     def state(self, v):
         self.cache = v
-
-
-class Mamba2Cache:
-    def __init__(self, num_layers):
-        self.conv_states = [None] * num_layers
-        self.ssm_states = [None] * num_layers
-        self.seqlen_offset = 0
-
-    def __getitem__(self, idx):
-        return (self.conv_states[idx], self.ssm_states[idx])
-
-    def __setitem__(self, idx, value):
-        self.conv_states[idx], self.ssm_states[idx] = value
-
-    @property
-    def state(self):
-        return {
-            'conv_states': self.conv_states,
-            'ssm_states': self.ssm_states,
-            'seqlen_offset': self.seqlen_offset
-        }
-
-    @state.setter
-    def state(self, v):
-        self.conv_states = v['conv_states']
-        self.ssm_states = v['ssm_states']
-        self.seqlen_offset = v['seqlen_offset']

llms/mlx_lm/models/mamba2.py

@@ -1,16 +1,11 @@
-# Copyright © 2024 Apple Inc.
-
 import math
 from dataclasses import dataclass, field
-from typing import Tuple, Union, Optional
+from typing import Tuple, Union
-
-import mlx.nn as nn
 import mlx.core as mx
+import mlx.nn as nn
 
 from .base import BaseModelArgs
-from .cache import Mamba2Cache
+from .cache import MambaCache
-
-# python -m mlx_lm.generate --model rokyang/mamba2-130m-hf  --prompt "hello how are you."
 
 @dataclass
 class ModelArgs(BaseModelArgs):
@@ -48,6 +43,7 @@ class ModelArgs(BaseModelArgs):
         if self.time_step_rank == "auto":
             self.time_step_rank = math.ceil(self.hidden_size / 16)
 
+
 class MambaRMSNormGated(nn.Module):
     def __init__(self, hidden_size, eps=1e-6):
         super().__init__()
@@ -97,175 +93,108 @@ class DepthWiseConv1d(nn.Module):
         return y, x[:, -K + 1 :, :]
     
 
-class Mamba2Mixer(nn.Module):
+class Mamba2Block(nn.Module):
-    def __init__(self, args, layer_idx):
+    def __init__(self, args: ModelArgs):
         super().__init__()
-        self.layer_idx = layer_idx
+        self.args = args
-        self.hidden_size = args.hidden_size
         self.intermediate_size = args.intermediate_size
-        self.num_heads = args.num_heads
+        self.time_step_rank = args.time_step_rank
-        self.head_dim = args.head_dim
-        self.ssm_state_size = args.state_size
-        self.n_groups = args.n_groups
         self.conv_kernel_size = args.conv_kernel
-        self.use_conv_bias = args.use_conv_bias
+        self.hidden_size = args.hidden_size
-        self.use_bias = args.use_bias
+        self.state_size = args.state_size
-        self.time_step_min = args.time_step_min
+        self.num_heads = args.num_heads
-        self.time_step_max = args.time_step_max
+        self.head_dim = args.hidden_size // args.num_heads
-        self.chunk_size = args.chunk_size
+        self.n_groups = args.n_groups
-        self.conv_dim = self.intermediate_size + 2 * self.n_groups * self.ssm_state_size
 
-        projection_size = self.intermediate_size + self.conv_dim + self.num_heads
+        self.conv_dim = args.intermediate_size + 2 * args.n_groups * args.state_size
+        self.conv1d = DepthWiseConv1d(
+            in_channels=self.conv_dim,
+            out_channels=self.conv_dim,
+            kernel_size=args.conv_kernel,
+            bias=args.use_conv_bias,
+            groups=self.conv_dim,
+            padding=args.conv_kernel - 1
+        )
+
+        projection_size = args.intermediate_size + self.conv_dim + args.num_heads
         self.in_proj = nn.Linear(
-            self.hidden_size,
+            args.hidden_size,
             projection_size,
             bias=args.use_bias
         )
-        self.conv1d = nn.Conv1d(
+
-            self.conv_dim,
-            self.conv_dim,
-            self.conv_kernel_size,
-            groups=self.conv_dim,
-            bias=self.use_conv_bias
-        )
         self.act = nn.SiLU()
-        self.norm = MambaRMSNormGated(self.intermediate_size, eps=args.layer_norm_epsilon)
+
-        self.out_proj = nn.Linear(
+        self.A_log = mx.zeros(args.num_heads)
-            self.intermediate_size,
+        self.D = mx.ones((args.num_heads,))
-            self.hidden_size,
+        self.dt_bias = mx.zeros(args.num_heads)
-            bias=self.use_bias
+
+        self.out_proj = nn.Linear(args.intermediate_size, args.hidden_size, bias=args.use_bias)
+        self.norm = MambaRMSNormGated(args.intermediate_size, eps=args.layer_norm_epsilon)
+
+    def ssm_step(self, x, state, dt_proj):
+        A = -mx.exp(self.A_log)
+        D = self.D
+        delta = nn.softplus(dt_proj + self.dt_bias)
+        
+        B, C = mx.split(x, indices_or_sections=[self.state_size * self.n_groups], axis=-1)
+        
+        batch_size = B.shape[0]
+        B = B.reshape(batch_size, self.n_groups, self.state_size)
+        C = C.reshape(batch_size, -1, self.state_size)
+        
+        delta = delta.reshape(batch_size, self.num_heads, 1)
+        A = A.reshape(1, self.num_heads, 1)
+        
+        if state is None:
+            new_state = delta * B
+        else:
+            new_state = delta * (B + state * mx.exp(delta * A))
+        
+        y = mx.sum(new_state[:, :, None, :] * C[:, None, :, :], axis=(-1, -2))
+        y = y + D * x[:, :self.num_heads]
+        return y, new_state
+
+    def __call__(self, x, cache):
+        B, T, D = x.shape
+        if cache is None:
+            cache = [None, None]
+
+        outputs = []
+        for t in range(T):
+            xt = x[:, t, :]
+            xz = self.in_proj(xt)
+            
+            x_t, z_t, dt_proj = mx.split(
+                xz,
+                indices_or_sections=[self.conv_dim, self.conv_dim + self.intermediate_size],
+                axis=-1
             )
 
-        self.A_log = mx.zeros(self.num_heads)
+            # Use the new DepthWiseConv1d with caching
-        self.D = mx.ones(self.num_heads)
+            conv_out, cache[0] = self.conv1d(mx.expand_dims(x_t, 1), cache[0])
-        self.dt_bias = mx.zeros(self.num_heads)
+            x_t = conv_out.squeeze(1)
+            x_t = nn.silu(x_t)
+            y_t, cache[1] = self.ssm_step(x_t, cache[1], dt_proj)
+            z_t = nn.silu(z_t)
             
-    def __call__(self, input_states, cache):
+            # Element-wise multiplication
-        batch_size, seq_len, _ = input_states.shape
+            output_t = y_t[:, :, None] * z_t[:, None, :]
-        dtype = input_states.dtype
             
-        projected_states = self.in_proj(input_states)
+            # Sum across the second dimension to match the intermediate_size
+            output_t = output_t.sum(axis=1)
             
-        # Calculate the sizes of each split
+            output_t = self.out_proj(output_t)
-        total_size = projected_states.shape[-1]
+            outputs.append(output_t)
-        remaining_size = total_size - self.intermediate_size - self.conv_dim - self.num_heads
-        d_mlp = remaining_size // 2
-        sizes = [
-            d_mlp,
-            d_mlp,
-            self.intermediate_size,
-            self.conv_dim,
-            self.num_heads
-        ]
         
-        # Perform the split operation
+        output = mx.stack(outputs, axis=1)
-        split_result = mx.split(projected_states, sizes, axis=-1)
+        return output
-        
-        # Print debug information
-        print(f"Number of split parts: {len(split_result)}")
-        print(f"Shapes of split parts: {[part.shape for part in split_result]}")
-        
-        # Flexibly handle the split result
-        _, _, _, gate, hidden_states, dt = split_result
-
-        if cache is not None:
-            conv_state = cache.conv_states[self.layer_idx]
-            if conv_state is None:
-                # Initialize conv_state if it's None
-                conv_state = mx.zeros((batch_size, 1, self.conv_kernel_size, hidden_states.shape[-1]))
-            
-            conv_state = mx.roll(conv_state, -1, -2)  # Roll along the kernel dimension
-            
-            # Reshape hidden_states to match conv_state dimensions
-            hidden_states_reshaped = hidden_states[:, None, None, :]
-            
-            conv_state = mx.concat([conv_state[:, :, :-1, :], hidden_states_reshaped], axis=-2)
-            cache.conv_states[self.layer_idx] = conv_state
-            
-            # Adjust the convolution operation
-            hidden_states = mx.sum(conv_state * self.conv1d.weight[:, :, None, :], axis=(-2, -1))
-            
-            if self.use_conv_bias:
-                hidden_states += self.conv1d.bias
-            hidden_states = self.act(hidden_states)[:, None, :]
-        else:
-            hidden_states = hidden_states.transpose(0, 2, 1)
-            hidden_states = self.act(self.conv1d(hidden_states)).transpose(0, 2, 1)
-
-        hidden_states, B, C = mx.split(hidden_states, [self.intermediate_size, self.n_groups * self.ssm_state_size, self.n_groups * self.ssm_state_size], axis=-1)
-
-        A = -mx.exp(self.A_log.astype(mx.float32))
-        dt = nn.softplus(dt + self.dt_bias)
-        dt = mx.clip(dt, self.time_step_min, self.time_step_max)
-
-        hidden_states = hidden_states.reshape(batch_size, seq_len, -1, self.head_dim).astype(mx.float32)
-        B = B.reshape(batch_size, seq_len, -1, self.ssm_state_size).astype(mx.float32)
-        C = C.reshape(batch_size, seq_len, -1, self.ssm_state_size).astype(mx.float32)
-
-        B = mx.repeat(B, repeats=self.num_heads // self.n_groups, axis=2)
-        C = mx.repeat(C, repeats=self.num_heads // self.n_groups, axis=2)
-
-        if cache is not None and cache.seqlen_offset > 0:
-            ssm_state = cache.ssm_states[self.layer_idx]
-            dA = mx.exp(dt[:, None, :, None] * A[None, :, None, None])
-            dB = dt[:, None, :, None] * B
-            dBx = dB * hidden_states[:, :, :, None]
-            ssm_state = ssm_state * dA + dBx
-            cache.ssm_states[self.layer_idx] = ssm_state
-
-            y = mx.sum(ssm_state * C[:, None, :, :], axis=-1)
-            D = self.D[None, :, None].expand(self.D.shape[0], self.head_dim)
-            y = y + hidden_states * D
-
-            y = y.reshape(batch_size, -1)[:, None, :]
-        else:
-            # Implement chunked computation here (simplified version)
-            pad_size = self.chunk_size - (seq_len % self.chunk_size)
-            hidden_states_padded = mx.pad(hidden_states, [(0, 0), (0, pad_size), (0, 0), (0, 0)])
-            B_padded = mx.pad(B, [(0, 0), (0, pad_size), (0, 0), (0, 0)])
-            C_padded = mx.pad(C, [(0, 0), (0, pad_size), (0, 0), (0, 0)])
-
-            chunks = seq_len // self.chunk_size + (1 if pad_size > 0 else 0)
-            y_list = []
-            ssm_state = mx.zeros((batch_size, self.num_heads, self.head_dim, self.ssm_state_size))
-
-            for i in range(chunks):
-                chunk_start = i * self.chunk_size
-                chunk_end = (i + 1) * self.chunk_size
-                chunk_h = hidden_states_padded[:, chunk_start:chunk_end]
-                chunk_B = B_padded[:, chunk_start:chunk_end]
-                chunk_C = C_padded[:, chunk_start:chunk_end]
-
-                chunk_dt = dt[:, chunk_start:chunk_end]
-                dA = mx.exp(chunk_dt[:, :, None, None] * A[None, None, :, None])
-                dB = chunk_dt[:, :, None, None] * chunk_B
-                dBx = dB * chunk_h[:, :, :, None]
-
-                chunk_y = mx.zeros_like(chunk_h)
-                for j in range(self.chunk_size):
-                    ssm_state = ssm_state * dA[:, j] + dBx[:, j]
-                    chunk_y[:, j] = mx.sum(ssm_state * chunk_C[:, j], axis=-1)
-
-                y_list.append(chunk_y)
-
-            y = mx.concat(y_list, axis=1)
-            if pad_size > 0:
-                y = y[:, :seq_len]
-
-            D = self.D[None, :, None].expand(self.D.shape[0], self.head_dim)
-            y = y + hidden_states * D
-            y = y.reshape(batch_size, seq_len, -1)
-
-        y = self.norm(y, gate)
-        contextualized_states = self.out_proj(y.astype(dtype))
-
-        return contextualized_states
 
 
-class Mamba2Block(nn.Module):
+class ResidualBlock(nn.Module):
-    def __init__(self, args: ModelArgs, layer_idx: int):
+    def __init__(self, args: ModelArgs):
         super().__init__()
-        self.mixer = Mamba2Mixer(args, layer_idx)
+        self.mixer = Mamba2Block(args)
         self.norm = nn.RMSNorm(args.hidden_size)
 
     def __call__(self, x: mx.array, cache):
@@ -277,24 +206,16 @@ class Mamba2(nn.Module):
         super().__init__()
         self.args = args
         self.embeddings = nn.Embedding(args.vocab_size, args.hidden_size)
-        self.layers = [Mamba2Block(args, idx) for idx in range(args.num_hidden_layers)]
+        self.layers = [ResidualBlock(args) for _ in range(args.num_hidden_layers)]
         self.norm_f = nn.RMSNorm(args.hidden_size, eps=args.layer_norm_epsilon)
 
-    def __call__(
+    def __call__(self, x: mx.array, cache):
-        self,
+        x = self.embeddings(x)
-        inputs: mx.array,
-        cache=None
-    ):
-        hidden_states = self.embeddings(inputs)
-        
         if cache is None:
-            cache = Mamba2Cache(len(self.layers))
+            cache = [None] * len(self.layers)
-
+        for layer, c in zip(self.layers, cache):
-        for i, layer in enumerate(self.layers):
+            x = layer(x, c)
-            hidden_states = layer(hidden_states, cache[i])
+        return self.norm_f(x)
-
-        hidden_states = self.norm_f(hidden_states)
-        return hidden_states
 
 
 class Model(nn.Module):
@@ -302,7 +223,10 @@ class Model(nn.Module):
         super().__init__()
         self.args = args
         self.model_type = args.model_type
+        
         self.backbone = Mamba2(args)
+        # self.norm_f = nn.RMSNorm(args.hidden_size, eps=args.layer_norm_epsilon)
+
         if not args.tie_word_embeddings:
             self.lm_head = nn.Linear(args.hidden_size, args.vocab_size, bias=False)
 
@@ -316,9 +240,6 @@ class Model(nn.Module):
         else:
             logits = self.lm_head(x)
 
-        print(logits)
-        print(logits.shape)
-
         return logits
     
     def sanitize(self, weights):
@@ -328,7 +249,7 @@ class Model(nn.Module):
         return weights
 
     def make_cache(self):
-        return [Mamba2Cache(self.args.num_hidden_layers) for _ in range(len(self.layers))]
+        return [MambaCache() for _ in range(len(self.layers))]
 
     @property
     def layers(self):