Feat: Support Granite 3.0 MoE in SGLang (#7959)

docs/supported_models/generative_models.md

@@ -45,3 +45,5 @@ in the GitHub search bar.
 | **SmolLM** (135M–1.7B)            | `HuggingFaceTB/SmolLM-1.7B`                      | Hugging Face’s ultra-small LLM series (135M–1.7B params) offering surprisingly strong results, enabling advanced AI on mobile/edge devices. |
 | **GLM-4** (Multilingual 9B)        | `ZhipuAI/glm-4-9b-chat`                          | Zhipu’s GLM-4 series (up to 9B parameters) – open multilingual models with support for 1M-token context and even a 5.6B multimodal variant (Phi-4V). |
 | **MiMo** (7B series)               | `XiaomiMiMo/MiMo-7B-RL`                         | Xiaomi's reasoning-optimized model series, leverages Multiple-Token Prediction for faster inference. |
+| **Granite 3.0, 3.1** (IBM)               | `ibm-granite/granite-3.1-8b-instruct`                          | IBM's open dense foundation models optimized for reasoning, code, and business AI use cases. Integrated with Red Hat and watsonx systems. |
+| **Granite 3.0 MoE** (IBM)               | `ibm-granite/granite-3.0-3b-a800m-instruct`                          | IBM’s Mixture-of-Experts models offering strong performance with cost-efficiency. MoE expert routing designed for enterprise deployment at scale. |

python/sglang/srt/models/granitemoe.py

+"""Inference-only GraniteMoe model."""
+
+from typing import Iterable, Optional
+
+import torch
+from torch import nn
+from transformers import GraniteConfig
+
+from sglang.srt.distributed import get_tensor_model_parallel_world_size
+from sglang.srt.layers.layernorm import RMSNorm
+from sglang.srt.layers.linear import (
+    QKVParallelLinear,
+    ReplicatedLinear,
+    RowParallelLinear,
+)
+from sglang.srt.layers.logits_processor import LogitsProcessor, LogitsProcessorOutput
+from sglang.srt.layers.moe.fused_moe_triton import FusedMoE
+from sglang.srt.layers.pooler import Pooler, PoolingType
+from sglang.srt.layers.quantization.base_config import QuantizationConfig
+from sglang.srt.layers.radix_attention import RadixAttention
+from sglang.srt.layers.rotary_embedding import get_rope
+from sglang.srt.layers.vocab_parallel_embedding import (
+    ParallelLMHead,
+    VocabParallelEmbedding,
+)
+from sglang.srt.model_executor.forward_batch_info import ForwardBatch
+from sglang.srt.models import mixtral
+from sglang.srt.utils import add_prefix
+
+
+class GraniteMoeMoE(nn.Module):
+    """A tensor-parallel MoE implementation for GraniteMoe that shards each
+    expert across all ranks.
+    Each expert's weights are sharded across all ranks and a fused MoE
+    kernel is used for the forward pass, and finally we reduce the outputs
+    across ranks.
+    """
+
+    def __init__(
+        self,
+        num_experts: int,
+        top_k: int,
+        hidden_size: int,
+        intermediate_size: int,
+        params_dtype: Optional[torch.dtype] = None,
+        quant_config: Optional[QuantizationConfig] = None,
+        tp_size: Optional[int] = None,
+        prefix: str = "",
+    ):
+        super().__init__()
+        self.hidden_size = hidden_size
+
+        # Gate always runs at half / full precision for now.
+        self.gate = ReplicatedLinear(
+            hidden_size,
+            num_experts,
+            bias=False,
+            params_dtype=params_dtype,
+            quant_config=None,
+            prefix=f"{prefix}.gate",
+        )
+
+        self.experts = FusedMoE(
+            num_experts=num_experts,
+            top_k=top_k,
+            hidden_size=hidden_size,
+            intermediate_size=intermediate_size,
+            params_dtype=params_dtype,
+            reduce_results=True,
+            renormalize=True,
+            quant_config=quant_config,
+            tp_size=tp_size,
+            prefix=f"{prefix}.experts",
+        )
+
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        # NOTE: hidden_states can have either 1D or 2D shape.
+        orig_shape = hidden_states.shape
+        hidden_states = hidden_states.view(-1, self.hidden_size)
+        router_logits, _ = self.gate(hidden_states)
+        final_hidden_states = self.experts(hidden_states, router_logits)
+        return final_hidden_states.view(orig_shape)
+
+
+class GraniteMoeAttention(nn.Module):
+
+    def __init__(
+        self,
+        hidden_size: int,
+        num_heads: int,
+        num_kv_heads: int,
+        max_position: int = 4096 * 32,
+        layer_id: int = 0,
+        rope_theta: float = 10000,
+        quant_config: Optional[QuantizationConfig] = None,
+        attention_multiplier: Optional[float] = None,
+        prefix: str = "",
+    ) -> None:
+        super().__init__()
+        self.hidden_size = hidden_size
+        tp_size = get_tensor_model_parallel_world_size()
+        self.total_num_heads = num_heads
+        assert self.total_num_heads % tp_size == 0
+        self.num_heads = self.total_num_heads // tp_size
+        self.total_num_kv_heads = num_kv_heads
+        if self.total_num_kv_heads >= tp_size:
+            # Number of KV heads is greater than TP size, so we partition
+            # the KV heads across multiple tensor parallel GPUs.
+            assert self.total_num_kv_heads % tp_size == 0
+        else:
+            # Number of KV heads is less than TP size, so we replicate
+            # the KV heads across multiple tensor parallel GPUs.
+            assert tp_size % self.total_num_kv_heads == 0
+        self.num_kv_heads = max(1, self.total_num_kv_heads // tp_size)
+        self.head_dim = hidden_size // self.total_num_heads
+        self.q_size = self.num_heads * self.head_dim
+        self.kv_size = self.num_kv_heads * self.head_dim
+        self.scaling = (
+            attention_multiplier
+            if attention_multiplier is not None
+            else self.head_dim**-1
+        )
+        self.rope_theta = rope_theta
+
+        self.qkv_proj = QKVParallelLinear(
+            hidden_size,
+            self.head_dim,
+            self.total_num_heads,
+            self.total_num_kv_heads,
+            bias=False,
+            quant_config=quant_config,
+            prefix=f"{prefix}.qkv_proj",
+        )
+        self.o_proj = RowParallelLinear(
+            self.total_num_heads * self.head_dim,
+            hidden_size,
+            bias=False,
+            quant_config=quant_config,
+            prefix=f"{prefix}.o_proj",
+        )
+        self.rotary_emb = get_rope(
+            self.head_dim,
+            rotary_dim=self.head_dim,
+            max_position=max_position,
+            base=int(self.rope_theta),
+            is_neox_style=True,
+        )
+        self.attn = RadixAttention(
+            self.num_heads,
+            self.head_dim,
+            self.scaling,
+            num_kv_heads=self.num_kv_heads,
+            layer_id=layer_id,
+            quant_config=quant_config,
+            prefix=f"{prefix}.attn",
+        )
+
+    def forward(
+        self,
+        positions: torch.Tensor,
+        hidden_states: torch.Tensor,
+        forward_batch: ForwardBatch,
+    ) -> torch.Tensor:
+        qkv, _ = self.qkv_proj(hidden_states)
+        q, k, v = qkv.split([self.q_size, self.kv_size, self.kv_size], dim=-1)
+        q, k = self.rotary_emb(positions, q, k)
+        attn_output = self.attn(q, k, v, forward_batch)
+        output, _ = self.o_proj(attn_output)
+        return output
+
+
+class GraniteMoeDecoderLayer(nn.Module):
+
+    def __init__(
+        self,
+        config: GraniteConfig,
+        layer_id: int = 0,
+        quant_config: Optional[QuantizationConfig] = None,
+        prefix: str = "",
+    ) -> None:
+        super().__init__()
+        self.hidden_size = config.hidden_size
+        rope_theta = getattr(config, "rope_theta", 10000)
+        self.self_attn = GraniteMoeAttention(
+            hidden_size=self.hidden_size,
+            num_heads=config.num_attention_heads,
+            max_position=config.max_position_embeddings,
+            num_kv_heads=config.num_key_value_heads,
+            rope_theta=rope_theta,
+            layer_id=layer_id,
+            quant_config=quant_config,
+            prefix=f"{prefix}.self_attn",
+            attention_multiplier=config.attention_multiplier,
+        )
+        self.block_sparse_moe = GraniteMoeMoE(
+            num_experts=config.num_local_experts,
+            top_k=config.num_experts_per_tok,
+            hidden_size=config.hidden_size,
+            intermediate_size=config.intermediate_size,
+            quant_config=quant_config,
+            prefix=f"{prefix}.block_sparse_moe",
+        )
+
+        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.residual_multiplier = config.residual_multiplier
+
+    def forward(
+        self,
+        positions: torch.Tensor,
+        hidden_states: torch.Tensor,
+        forward_batch: ForwardBatch,
+    ) -> torch.Tensor:
+        residual = hidden_states
+        hidden_states = self.input_layernorm(hidden_states)
+        # Self Attention
+        hidden_states = self.self_attn(
+            positions=positions,
+            hidden_states=hidden_states,
+            forward_batch=forward_batch,
+        )
+        hidden_states = residual + hidden_states * self.residual_multiplier
+        residual = hidden_states
+        hidden_states = self.post_attention_layernorm(hidden_states)
+        hidden_states = self.block_sparse_moe(hidden_states)
+        hidden_states = residual + hidden_states * self.residual_multiplier
+
+        return hidden_states
+
+
+class GraniteMoeModel(nn.Module):
+
+    def __init__(
+        self,
+        config: GraniteConfig,
+        quant_config: Optional[QuantizationConfig] = None,
+        prefix: str = "",
+    ):
+        super().__init__()
+        self.embed_tokens = VocabParallelEmbedding(
+            config.vocab_size,
+            config.hidden_size,
+            org_num_embeddings=config.vocab_size,
+        )
+        self.embedding_multiplier = config.embedding_multiplier
+
+        self.layers = nn.ModuleList(
+            [
+                GraniteMoeDecoderLayer(
+                    config,
+                    i,
+                    quant_config=quant_config,
+                    prefix=add_prefix(f"layers.{i}", prefix),
+                )
+                for i in range(config.num_hidden_layers)
+            ]
+        )
+        self.norm = RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+
+    def get_input_embeddings(self, input_ids: torch.Tensor) -> torch.Tensor:
+        return self.embed_tokens(input_ids)
+
+    def forward(
+        self,
+        input_ids: torch.Tensor,
+        positions: torch.Tensor,
+        forward_batch: ForwardBatch,
+        inputs_embeds: Optional[torch.Tensor] = None,
+    ) -> torch.Tensor:
+        if inputs_embeds is not None:
+            hidden_states = inputs_embeds
+        else:
+            hidden_states = self.get_input_embeddings(input_ids)
+        hidden_states *= self.embedding_multiplier
+
+        for i in range(len(self.layers)):
+            layer = self.layers[i]
+            hidden_states = layer(
+                positions,
+                hidden_states,
+                forward_batch,
+            )
+        hidden_states = self.norm(hidden_states)
+        return hidden_states
+
+
+class GraniteMoeForCausalLM(nn.Module):
+
+    def __init__(
+        self,
+        config: GraniteConfig,
+        quant_config: Optional[QuantizationConfig] = None,
+        prefix: str = "",
+    ):
+        super().__init__()
+        self.config = config
+        self.quant_config = quant_config
+
+        self.model = GraniteMoeModel(
+            config, quant_config=quant_config, prefix=add_prefix("model", prefix)
+        )
+        self.lm_head = ParallelLMHead(
+            config.vocab_size,
+            config.hidden_size,
+            quant_config=quant_config,
+            prefix=add_prefix("lm_head", prefix),
+        )
+        if config.tie_word_embeddings:
+            self.lm_head.weight = self.model.embed_tokens.weight
+        # Granite logit scaling factors are applied via division, but
+        # LogitsProcessor expects a multiplicative factor.
+        if hasattr(config, "logits_scaling"):
+            logit_scale = 1.0 / config.logits_scaling
+        else:
+            logit_scale = None
+        self.logits_processor = LogitsProcessor(config, logit_scale=logit_scale)
+        self.pooler = Pooler(pooling_type=PoolingType.LAST, normalize=True)
+
+    @torch.no_grad()
+    def forward(
+        self,
+        input_ids: torch.Tensor,
+        positions: torch.Tensor,
+        forward_batch: ForwardBatch,
+        input_embeds: torch.Tensor = None,
+        get_embedding: bool = False,
+    ) -> LogitsProcessorOutput:
+        hidden_states = self.model(input_ids, positions, forward_batch, input_embeds)
+        if not get_embedding:
+            logits_processor_output: LogitsProcessorOutput = self.logits_processor(
+                input_ids, hidden_states, self.lm_head, forward_batch
+            )
+            return logits_processor_output
+        else:
+            return self.pooler(hidden_states, forward_batch)
+
+    def load_weights(self, weights: Iterable[tuple[str, torch.Tensor]]) -> set[str]:
+        new_weights = {}
+        for n, p in weights:
+            if n.endswith(".block_sparse_moe.input_linear.weight"):
+                for e in range(p.size(0)):
+                    w1_name = n.replace(
+                        ".block_sparse_moe.input_linear.weight",
+                        f".block_sparse_moe.experts.{e}.w1.weight",
+                    )
+                    w3_name = n.replace(
+                        ".block_sparse_moe.input_linear.weight",
+                        f".block_sparse_moe.experts.{e}.w3.weight",
+                    )
+                    w1_param, w3_param = p[e].chunk(2, dim=0)
+                    assert w1_name not in new_weights
+                    assert w3_name not in new_weights
+                    new_weights[w1_name] = w1_param
+                    new_weights[w3_name] = w3_param
+            elif n.endswith(".block_sparse_moe.output_linear.weight"):
+                for e in range(p.size(0)):
+                    w2_name = n.replace(
+                        ".block_sparse_moe.output_linear.weight",
+                        f".block_sparse_moe.experts.{e}.w2.weight",
+                    )
+                    w2_param = p[e]
+                    assert w2_name not in new_weights
+                    new_weights[w2_name] = w2_param
+            elif n.endswith(".block_sparse_moe.router.layer.weight"):
+                gate_name = n.replace(
+                    ".block_sparse_moe.router.layer.weight",
+                    ".block_sparse_moe.gate.weight",
+                )
+                assert gate_name not in new_weights
+                new_weights[gate_name] = p
+            else:
+                new_weights[n] = p
+        mixtral.MixtralForCausalLM.load_weights(self, new_weights.items())
+
+
+EntryClass = [GraniteMoeForCausalLM]