From 8801beb66f61d16114d4014fec32a266778e4481 Mon Sep 17 00:00:00 2001
From: Awni Hannun <awni@apple.com>
Date: Mon, 2 Dec 2024 11:42:58 -0800
Subject: [PATCH] Add olmo2 (#1128)

* add olmo2

* add olmo2
---
 llms/mlx_lm/models/olmo2.py | 312 ++++++++++++++++++++++++++++++++++++
 llms/mlx_lm/tuner/utils.py  |   1 +
 llms/tests/test_models.py   |  20 +++
 3 files changed, 333 insertions(+)
 create mode 100644 llms/mlx_lm/models/olmo2.py

diff --git a/llms/mlx_lm/models/olmo2.py b/llms/mlx_lm/models/olmo2.py
new file mode 100644
index 00000000..a28fdcc1
--- /dev/null
+++ b/llms/mlx_lm/models/olmo2.py
@@ -0,0 +1,312 @@
+# Copyright © 2023-2024 Apple Inc.
+
+from dataclasses import dataclass
+from typing import Any, Dict, Optional, Union
+
+import mlx.core as mx
+import mlx.nn as nn
+
+from .base import BaseModelArgs, create_attention_mask, scaled_dot_product_attention
+
+
+@dataclass
+class ModelArgs(BaseModelArgs):
+    model_type: str
+    hidden_size: int
+    num_hidden_layers: int
+    intermediate_size: int
+    num_attention_heads: int
+    rms_norm_eps: float
+    vocab_size: int
+    head_dim: Optional[int] = None
+    max_position_embeddings: Optional[int] = None
+    num_key_value_heads: Optional[int] = None
+    attention_bias: bool = False
+    mlp_bias: bool = False
+    rope_theta: float = 10000
+    rope_traditional: bool = False
+    rope_scaling: Optional[Dict[str, Union[float, str]]] = None
+    tie_word_embeddings: bool = True
+
+    def __post_init__(self):
+        if self.num_key_value_heads is None:
+            self.num_key_value_heads = self.num_attention_heads
+
+        if self.rope_scaling:
+            if not "factor" in self.rope_scaling:
+                raise ValueError(f"rope_scaling must contain 'factor'")
+            rope_type = self.rope_scaling.get("type") or self.rope_scaling.get(
+                "rope_type"
+            )
+            if rope_type is None:
+                raise ValueError(
+                    f"rope_scaling must contain either 'type' or 'rope_type'"
+                )
+            if rope_type not in ["linear", "dynamic", "llama3"]:
+                raise ValueError(
+                    "rope_scaling 'type' currently only supports 'linear', 'dynamic' or 'llama3'"
+                )
+
+
+class DynamicNTKScalingRoPE(nn.Module):
+    """Implements the rotary positional encoding with Dynamic NTK scaling and Llama 3 RoPE."""
+
+    def __init__(
+        self,
+        dims: int,
+        max_position_embeddings: int = 2048,
+        traditional: bool = False,
+        base: float = 10000,
+        scale: float = 1.0,
+        rope_type: str = "default",
+        rope_scaling: dict = None,
+    ):
+        super().__init__()
+        self.dims = dims
+        self.max_position_embeddings = max_position_embeddings
+        self.traditional = traditional
+        self.scale = scale
+        self.rope_type = rope_type
+        self.rope_scaling = rope_scaling
+        self.base = base
+        self.compute_freqs()
+
+    def compute_freqs(self):
+        if self.rope_type != "llama3":
+            self._freqs = None
+            return
+        factor = self.rope_scaling["factor"]
+        low_freq_factor = self.rope_scaling.get("low_freq_factor", 1.0)
+        high_freq_factor = self.rope_scaling.get("high_freq_factor", 4.0)
+        old_context_len = self.rope_scaling.get(
+            "original_max_position_embeddings",
+            8192,
+        )
+
+        low_freq_wavelen = old_context_len / low_freq_factor
+        high_freq_wavelen = old_context_len / high_freq_factor
+
+        freqs = self.base ** (mx.arange(0, self.dims, 2) / self.dims)
+        wavelens = 2 * mx.pi * freqs
+
+        freqs = mx.where(wavelens > low_freq_wavelen, freqs * factor, freqs)
+        is_medium_freq = (wavelens > high_freq_wavelen) & (wavelens < low_freq_wavelen)
+        smooth_factors = (old_context_len / wavelens - low_freq_factor) / (
+            high_freq_factor - low_freq_factor
+        )
+        smooth_freqs = freqs / ((1 - smooth_factors) / factor + smooth_factors)
+        self._freqs = mx.where(is_medium_freq, smooth_freqs, freqs)
+        self.base = None
+
+    def extra_repr(self):
+        return (
+            f"{self.dims}, traditional={self.traditional}, "
+            f"max_position_embeddings={self.max_position_embeddings}, "
+            f"scaling_factor={self.scale}, rope_type={self.rope_type}"
+        )
+
+    def __call__(self, x, offset: int = 0):
+        return mx.fast.rope(
+            x,
+            self.dims,
+            traditional=self.traditional,
+            base=self.base,
+            scale=self.scale,
+            offset=offset,
+            freqs=self._freqs,
+        )
+
+
+def initialize_rope(args: ModelArgs):
+    head_dim = args.head_dim or args.hidden_size // args.num_attention_heads
+
+    rope_scaling = args.rope_scaling
+    rope_type = "default"
+    rope_scale = 1.0
+
+    if rope_scaling is not None:
+        rope_type = (
+            rope_scaling.get("type") or rope_scaling.get("rope_type") or "default"
+        )
+        if rope_type == "linear":
+            rope_scale = 1 / rope_scaling["factor"]
+        elif rope_type == "llama3":
+            rope_scale = 1.0  # The scaling is handled internally for llama3
+
+    return DynamicNTKScalingRoPE(
+        dims=head_dim,
+        max_position_embeddings=args.max_position_embeddings,
+        traditional=args.rope_traditional,
+        base=args.rope_theta,
+        scale=rope_scale,
+        rope_type=rope_type,
+        rope_scaling=rope_scaling,
+    )
+
+
+class Attention(nn.Module):
+    def __init__(self, args: ModelArgs):
+        super().__init__()
+
+        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.head_dim = head_dim = args.head_dim or args.hidden_size // n_heads
+
+        self.scale = head_dim**-0.5
+        if hasattr(args, "attention_bias"):
+            attention_bias = args.attention_bias
+        else:
+            attention_bias = False
+
+        self.q_proj = nn.Linear(dim, n_heads * head_dim, bias=attention_bias)
+        self.k_proj = nn.Linear(dim, n_kv_heads * head_dim, bias=attention_bias)
+        self.v_proj = nn.Linear(dim, n_kv_heads * head_dim, bias=attention_bias)
+        self.o_proj = nn.Linear(n_heads * head_dim, dim, bias=attention_bias)
+
+        self.rope = initialize_rope(args)
+        self.q_norm = nn.RMSNorm(n_heads * head_dim, args.rms_norm_eps)
+        self.k_norm = nn.RMSNorm(n_kv_heads * head_dim, args.rms_norm_eps)
+
+    def __call__(
+        self,
+        x: mx.array,
+        mask: Optional[mx.array] = None,
+        cache: Optional[Any] = None,
+    ) -> mx.array:
+        B, L, D = x.shape
+
+        queries, keys, values = self.q_proj(x), self.k_proj(x), self.v_proj(x)
+        queries = self.q_norm(queries)
+        keys = self.k_norm(keys)
+
+        # 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)
+
+        output = scaled_dot_product_attention(
+            queries, keys, values, cache=cache, scale=self.scale, mask=mask
+        )
+
+        output = output.transpose(0, 2, 1, 3).reshape(B, L, -1)
+        return self.o_proj(output)
+
+
+class MLP(nn.Module):
+    def __init__(self, args: ModelArgs):
+        super().__init__()
+
+        dim = args.hidden_size
+        hidden_dim = args.intermediate_size
+        if hasattr(args, "mlp_bias"):
+            mlp_bias = args.mlp_bias
+        else:
+            mlp_bias = False
+
+        self.gate_proj = nn.Linear(dim, hidden_dim, bias=mlp_bias)
+        self.down_proj = nn.Linear(hidden_dim, dim, bias=mlp_bias)
+        self.up_proj = nn.Linear(dim, hidden_dim, bias=mlp_bias)
+
+    def __call__(self, x) -> mx.array:
+        return self.down_proj(nn.silu(self.gate_proj(x)) * self.up_proj(x))
+
+
+class TransformerBlock(nn.Module):
+    def __init__(self, args: ModelArgs):
+        super().__init__()
+        self.num_attention_heads = args.num_attention_heads
+        self.hidden_size = args.hidden_size
+        self.self_attn = Attention(args)
+        self.mlp = MLP(args)
+        self.post_attention_layernorm = nn.RMSNorm(
+            args.hidden_size, eps=args.rms_norm_eps
+        )
+        self.post_feedforward_layernorm = nn.RMSNorm(
+            args.hidden_size, eps=args.rms_norm_eps
+        )
+        self.args = args
+
+    def __call__(
+        self,
+        x: mx.array,
+        mask: Optional[mx.array] = None,
+        cache: Optional[Any] = None,
+    ) -> mx.array:
+        r = self.post_attention_layernorm(self.self_attn(x, mask, cache))
+        h = x + r
+        r = self.post_feedforward_layernorm(self.mlp(h))
+        out = h + r
+        return out
+
+
+class LlamaModel(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.embed_tokens = nn.Embedding(args.vocab_size, args.hidden_size)
+        self.layers = [
+            TransformerBlock(args=args) for _ in range(args.num_hidden_layers)
+        ]
+        self.norm = nn.RMSNorm(args.hidden_size, eps=args.rms_norm_eps)
+
+    def __call__(
+        self,
+        inputs: mx.array,
+        cache=None,
+    ):
+        h = self.embed_tokens(inputs)
+
+        mask = create_attention_mask(h, cache)
+
+        if cache is None:
+            cache = [None] * len(self.layers)
+
+        for layer, c in zip(self.layers, cache):
+            h = layer(h, mask, cache=c)
+
+        return self.norm(h)
+
+
+class Model(nn.Module):
+    def __init__(self, args: ModelArgs):
+        super().__init__()
+        self.args = args
+        self.model_type = args.model_type
+        self.model = LlamaModel(args)
+        if not args.tie_word_embeddings:
+            self.lm_head = nn.Linear(args.hidden_size, args.vocab_size, bias=False)
+
+    def __call__(
+        self,
+        inputs: mx.array,
+        cache=None,
+    ):
+        out = self.model(inputs, cache)
+        if self.args.tie_word_embeddings:
+            out = self.model.embed_tokens.as_linear(out)
+        else:
+            out = self.lm_head(out)
+        return out
+
+    def sanitize(self, weights):
+        # Remove unused precomputed rotary freqs
+        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
diff --git a/llms/mlx_lm/tuner/utils.py b/llms/mlx_lm/tuner/utils.py
index 7c78ee91..835cb482 100644
--- a/llms/mlx_lm/tuner/utils.py
+++ b/llms/mlx_lm/tuner/utils.py
@@ -98,6 +98,7 @@ def linear_to_lora_layers(
         "cohere",
         "minicpm",
         "deepseek",
+        "olmo2",
     ]:
         keys = set(["self_attn.q_proj", "self_attn.v_proj"])
         if model.model_type in ["mixtral", "phimoe"]:
diff --git a/llms/tests/test_models.py b/llms/tests/test_models.py
index 93b881b9..edb594d7 100644
--- a/llms/tests/test_models.py
+++ b/llms/tests/test_models.py
@@ -792,6 +792,26 @@ class TestModels(unittest.TestCase):
             model, args.model_type, args.vocab_size, args.num_hidden_layers
         )
 
+    def test_olmo2(self):
+        from mlx_lm.models import olmo2
+
+        args = olmo2.ModelArgs(
+            model_type="olmo2",
+            hidden_size=128,
+            attention_bias=False,
+            intermediate_size=256,
+            num_attention_heads=4,
+            num_hidden_layers=4,
+            num_key_value_heads=2,
+            rms_norm_eps=1e-4,
+            rope_theta=1000,
+            vocab_size=1000,
+        )
+        model = olmo2.Model(args)
+        self.model_test_runner(
+            model, args.model_type, args.vocab_size, args.num_hidden_layers
+        )
+
 
 if __name__ == "__main__":
     unittest.main()