Add Starcoder 2

llms/README.md

@@ -46,7 +46,7 @@ You can convert models in the Python API with:
 ```python
 from mlx_lm import convert
 
-upload_repo = "mistralai/Mistral-7B-Instruct-v0.1"
+upload_repo = "mlx-community/My-Mistral-7B-v0.1-4bit"
 
 convert("mistralai/Mistral-7B-v0.1", quantize=True, upload_repo=upload_repo)
 ```

llms/mlx_lm/models/starcoder2.py

@@ -0,0 +1,178 @@
+import math
+from dataclasses import dataclass
+from typing import Dict, Optional, Tuple, Union
+
+import mlx.core as mx
+import mlx.nn as nn
+
+from .base import BaseModelArgs
+
+
+@dataclass
+class ModelArgs(BaseModelArgs):
+    model_type: str
+    hidden_size: int
+    num_hidden_layers: int
+    intermediate_size: int
+    num_attention_heads: int
+    num_key_value_heads: int = None
+    max_position_embeddings: int = 16384
+    norm_eps: float = None
+    rms_norm_eps: float = 1e-5
+    vocab_size: int = 49152
+    rope_theta: float = 100000
+
+
+class RMSNorm(nn.Module):
+    def __init__(self, hidden_sizes: int, eps: float = 1e-5):
+        super().__init__()
+        self.weight = mx.ones((hidden_sizes,))
+        self.eps = eps
+
+    def _norm(self, x):
+        return x * mx.rsqrt(x.square().mean(-1, keephidden_sizes=True) + self.eps)
+
+    def __call__(self, x):
+        output = self._norm(x.astype(mx.float32)).astype(x.dtype)
+        return self.weight * output
+
+
+class Attention(nn.Module):
+    def __init__(self, args: ModelArgs):
+        super().__init__()
+        self.args = args
+
+        dim = args.hidden_size
+        self.n_heads = n_heads = args.num_attention_heads
+        self.n_kv_heads = n_kv_heads = args.num_key_value_heads
+
+        self.repeats = self.n_heads // self.n_kv_heads
+
+        head_dim = args.hidden_size // args.num_attention_heads
+        self.scale = head_dim**-0.5
+
+        self.wq = nn.Linear(dim, n_heads * head_dim, bias=False)
+        self.wk = nn.Linear(dim, n_kv_heads * head_dim, bias=False)
+        self.wv = nn.Linear(dim, n_kv_heads * head_dim, bias=False)
+        self.wo = nn.Linear(n_heads * head_dim, dim, bias=False)
+        self.rope = nn.RoPE(head_dim, traditional=True, base=args.rope_theta)
+
+    def __call__(
+        self,
+        x: mx.array,
+        mask: Optional[mx.array] = None,
+        cache: Optional[Tuple[mx.array, mx.array]] = None,
+    ) -> mx.array:
+        B, L, D = x.shape
+
+        queries, keys, values = self.wq(x), self.wk(x), self.wv(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)
+
+        keys = mx.repeat(keys, self.repeats, axis=1)
+        values = mx.repeat(values, self.repeats, axis=1)
+
+        if cache is not None:
+            key_cache, value_cache = cache
+            queries = self.rope(queries, offset=key_cache.shape[2])
+            keys = self.rope(keys, offset=key_cache.shape[2])
+            keys = mx.concatenate([key_cache, keys], axis=2)
+            values = mx.concatenate([value_cache, values], axis=2)
+        else:
+            queries = self.rope(queries)
+            keys = self.rope(keys)
+
+        scores = (queries * self.scale) @ keys.transpose(0, 1, 3, 2)
+        if mask is not None:
+            scores += mask
+        scores = mx.softmax(scores.astype(mx.float32), axis=-1).astype(scores.dtype)
+        output = (scores @ values).transpose(0, 2, 1, 3).reshape(B, L, -1)
+        return self.wo(output), (keys, values)
+
+
+class MLP(nn.Module):
+    def __init__(self, dim, hidden_dim):
+        super().__init__()
+        self.c_fc = nn.Linear(dim, hidden_dim, bias=False)
+        self.c_proj = nn.Linear(hidden_dim, dim, bias=False)
+        self.act = nn.Linear(dim, hidden_dim, bias=False)
+
+    def __call__(self, x) -> mx.array:
+        return self.c_proj(nn.gelu(self.c_fc(x)) * self.act(x))
+
+
+class TransformerBlock(nn.Module):
+    def __init__(self, args: ModelArgs):
+        super().__init__()
+        self.hidden_size = args.hidden_size
+        self.n_heads = args.num_attention_heads
+        self.attention = Attention(args)
+        self.mlp = MLP(args.hidden_size, args.intermediate_size)
+        self.attention_norm = RMSNorm(args.hidden_size, eps=args.norm_eps)
+        self.ffn_norm = RMSNorm(args.hidden_size, eps=args.norm_eps)
+        self.args = args
+
+    def __call__(
+        self,
+        x: mx.array,
+        mask: Optional[mx.array] = None,
+        cache: Optional[Tuple[mx.array, mx.array]] = None,
+    ) -> mx.array:
+        r, cache = self.attention(self.attention_norm(x), mask, cache)
+        h = x + r
+        r = self.feed_forward(self.ffn_norm(h))
+        out = h + r
+        return out, cache
+
+
+class Starcoder2Model(nn.Module):
+    def __init__(self, args: ModelArgs):
+        super().__init__()
+        self.args = args
+        self.vocab_size = args.vocab_size
+        self.num_hidden_layers = args.num_hidden_layers
+        assert self.vocab_size > 0
+        self.tok_embeddings = nn.Embedding(args.vocab_size, args.hidden_size)
+        self.layers = [
+            TransformerBlock(args=args) for _ in range(args.num_hidden_layers)
+        ]
+        self.norm = RMSNorm(args.hidden_size, eps=args.norm_eps)
+        self.output = nn.Linear(args.hidden_size, args.vocab_size, bias=False)
+
+    def __call__(
+        self,
+        inputs: mx.array,
+        cache=None,
+    ):
+        h = self.tok_embeddings(inputs)
+
+        mask = None
+        if h.shape[1] > 1:
+            mask = nn.MultiHeadAttention.create_additive_causal_mask(h.shape[1])
+            mask = mask.astype(h.dtype)
+
+        if cache is None:
+            cache = [None] * len(self.layers)
+
+        for e, layer in enumerate(self.layers):
+            h, cache[e] = layer(h, mask, cache[e])
+
+        return self.output(self.norm(h)), cache
+
+
+class Model(nn.Module):
+    def __init__(self, args: ModelArgs):
+        super().__init__()
+        self.model = Starcoder2Model(args)
+        self.lm_head = nn.Linear(args.hidden_size, args.vocab_size, bias=False)
+
+    def __call__(
+        self,
+        inputs: mx.array,
+        cache=None,
+    ):
+        out, cache = self.model(inputs, cache)
+        return self.lm_head(out), cache

llms/mlx_lm/tuner/utils.py

@@ -32,6 +32,7 @@ def check_lora_layers(num_model):
         "stablelm_epoch",
         "qwen2",
         "gemma",
+        "starcoder2",
     ]:
         check_lora_layers(len(model.model.layers))