From 7ebc28f5d6571068ff2f9eb44dc6487ded4a8189 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E8=B5=B5=E6=99=A8=E9=98=B3?= Date: Sun, 26 Oct 2025 13:58:54 -0700 Subject: [PATCH] [WIP] support MiniMax M2 model (#12129) Signed-off-by: Xinyuan Tong Signed-off-by: xuebi Co-authored-by: Xinyuan Tong Co-authored-by: Xinyuan Tong <115166877+JustinTong0323@users.noreply.github.com> Co-authored-by: Roger Young <42564206+rogeryoungh@users.noreply.github.com> Co-authored-by: xuebi --- docs/supported_models/generative_models.md | 1 + .../srt/function_call/function_call_parser.py | 2 + python/sglang/srt/function_call/minimax_m2.py | 367 +++++++ python/sglang/srt/models/minimax_m2.py | 922 ++++++++++++++++++ python/sglang/srt/parser/reasoning_parser.py | 29 +- 5 files changed, 1320 insertions(+), 1 deletion(-) create mode 100644 python/sglang/srt/function_call/minimax_m2.py create mode 100644 python/sglang/srt/models/minimax_m2.py diff --git a/docs/supported_models/generative_models.md b/docs/supported_models/generative_models.md index fdb18b845..f08cb847b 100644 --- a/docs/supported_models/generative_models.md +++ b/docs/supported_models/generative_models.md @@ -35,6 +35,7 @@ in the GitHub search bar. | **MiniCPM** (v3, 4B) | `openbmb/MiniCPM3-4B` | OpenBMB’s series of compact LLMs for edge devices; MiniCPM 3 (4B) achieves GPT-3.5-level results in text tasks. | | **OLMo** (2, 3) | `allenai/OLMo-2-1124-7B-Instruct` | Allen AI’s series of Open Language Models designed to enable the science of language models. | | **OLMoE** (Open MoE) | `allenai/OLMoE-1B-7B-0924` | Allen AI’s open Mixture-of-Experts model (7B total, 1B active parameters) delivering state-of-the-art results with sparse expert activation. | +| **MiniMax-M2** | `minimax/MiniMax-M2` | MiniMax’s SOTA LLM for coding & agentic workflows. | | **StableLM** (3B, 7B) | `stabilityai/stablelm-tuned-alpha-7b` | StabilityAI’s early open-source LLM (3B & 7B) for general text generation; a demonstration model with basic instruction-following ability. | | **Command-R** (Cohere) | `CohereForAI/c4ai-command-r-v01` | Cohere’s open conversational LLM (Command series) optimized for long context, retrieval-augmented generation, and tool use. | | **DBRX** (Databricks) | `databricks/dbrx-instruct` | Databricks’ 132B-parameter MoE model (36B active) trained on 12T tokens; competes with GPT-3.5 quality as a fully open foundation model. | diff --git a/python/sglang/srt/function_call/function_call_parser.py b/python/sglang/srt/function_call/function_call_parser.py index d382a56f3..a11845c18 100644 --- a/python/sglang/srt/function_call/function_call_parser.py +++ b/python/sglang/srt/function_call/function_call_parser.py @@ -16,6 +16,7 @@ from sglang.srt.function_call.glm4_moe_detector import Glm4MoeDetector from sglang.srt.function_call.gpt_oss_detector import GptOssDetector from sglang.srt.function_call.kimik2_detector import KimiK2Detector from sglang.srt.function_call.llama32_detector import Llama32Detector +from sglang.srt.function_call.minimax_m2 import MinimaxM2Detector from sglang.srt.function_call.mistral_detector import MistralDetector from sglang.srt.function_call.pythonic_detector import PythonicDetector from sglang.srt.function_call.qwen3_coder_detector import Qwen3CoderDetector @@ -49,6 +50,7 @@ class FunctionCallParser: "qwen25": Qwen25Detector, "qwen3_coder": Qwen3CoderDetector, "step3": Step3Detector, + "minimax-m2": MinimaxM2Detector, } def __init__(self, tools: List[Tool], tool_call_parser: str): diff --git a/python/sglang/srt/function_call/minimax_m2.py b/python/sglang/srt/function_call/minimax_m2.py new file mode 100644 index 000000000..8ce740bb2 --- /dev/null +++ b/python/sglang/srt/function_call/minimax_m2.py @@ -0,0 +1,367 @@ +import ast +import html +import json +import logging +import re +from typing import Any, Dict, List, Tuple + +from sglang.srt.entrypoints.openai.protocol import Tool +from sglang.srt.function_call.base_format_detector import BaseFormatDetector +from sglang.srt.function_call.core_types import ( + StreamingParseResult, + ToolCallItem, + _GetInfoFunc, +) +from sglang.srt.function_call.ebnf_composer import EBNFComposer + +logger = logging.getLogger(__name__) + + +def _safe_val(raw: str) -> Any: + raw = html.unescape(raw.strip()) + try: + return json.loads(raw) + except Exception: + try: + return ast.literal_eval(raw) + except Exception: + return raw + + +class MinimaxM2Detector(BaseFormatDetector): + """ + Detector for MiniMax M2 models. + Assumes function call format: + + + value1 + value2 + + + """ + + def __init__(self): + super().__init__() + self.tool_call_start_token: str = "" + self.tool_call_end_token: str = "" + self.tool_call_prefix: str = '" + self.tool_call_regex = re.compile( + r"(.*?)|(.*?)$", + re.DOTALL, + ) + self.tool_call_function_regex = re.compile( + r"|| bool: + return self.tool_call_start_token in text + + def detect_and_parse(self, text: str, tools: List[Tool]) -> StreamingParseResult: + normal, calls = self._extract(text, tools) + return StreamingParseResult(normal_text=normal, calls=calls) + + def parse_streaming_increment( + self, new_text: str, tools: List[Tool] + ) -> StreamingParseResult: + self._buf += new_text + normal = "" + calls: List[ToolCallItem] = [] + + # Build tool indices for validation + if not hasattr(self, "_tool_indices"): + self._tool_indices = self._get_tool_indices(tools) + + while True: + # If we're not in a tool call and don't see a start token, return normal text + if not self._in_tool_call and self.tool_call_start_token not in self._buf: + normal += self._buf + self._buf = "" + break + + # Look for tool call start + if not self._in_tool_call: + s = self._buf.find(self.tool_call_start_token) + if s == -1: + normal += self._buf + self._buf = "" + break + + normal += self._buf[:s] + self._buf = self._buf[s:] + + self._in_tool_call = True + self._function_name_sent = False + self._current_function_name = "" + self._current_parameters = {} + self._streamed_parameters = {} + + # Remove the start token + self._buf = self._buf[len(self.tool_call_start_token) :] + continue + + # We're in a tool call, try to parse function name if not sent yet + if not self._function_name_sent: + # Look for function name pattern: + function_match = re.search(r"]+)\">", self._buf) + if function_match: + function_name = function_match.group(1).strip() + + # Validate function name + if function_name in self._tool_indices: + self._current_function_name = function_name + self._function_name_sent = True + + # Initialize tool call tracking + if self.current_tool_id == -1: + self.current_tool_id = 0 + + # Ensure tracking arrays are large enough + while len(self.prev_tool_call_arr) <= self.current_tool_id: + self.prev_tool_call_arr.append({}) + while len(self.streamed_args_for_tool) <= self.current_tool_id: + self.streamed_args_for_tool.append("") + + # Store tool call info + self.prev_tool_call_arr[self.current_tool_id] = { + "name": function_name, + "arguments": {}, + } + + # Send tool name with empty parameters + calls.append( + ToolCallItem( + tool_index=self.current_tool_id, + name=function_name, + parameters="", + ) + ) + + # Remove the processed function declaration + self._buf = self._buf[function_match.end() :] + continue + else: + # Invalid function name, reset state + logger.warning(f"Invalid function name: {function_name}") + self._reset_streaming_state() + normal += self._buf + self._buf = "" + break + else: + # Function name not complete yet, wait for more text + break + + # Parse parameters incrementally + if self._function_name_sent: + # Process parameters and get any calls to emit + parameter_calls = self._parse_and_stream_parameters(self._buf) + calls.extend(parameter_calls) + + # Check if tool call is complete + if self.tool_call_function_end_token in self._buf: + end_pos = self._buf.find(self.tool_call_function_end_token) + + # Add closing brace to complete the JSON object + current_streamed = self.streamed_args_for_tool[self.current_tool_id] + if current_streamed: + # Count opening and closing braces to check if JSON is complete + open_braces = current_streamed.count("{") + close_braces = current_streamed.count("}") + if open_braces > close_braces: + calls.append( + ToolCallItem( + tool_index=self.current_tool_id, + name=None, + parameters="}", + ) + ) + self.streamed_args_for_tool[self.current_tool_id] = ( + current_streamed + "}" + ) + + # Complete the tool call + self._buf = self._buf[ + end_pos + len(self.tool_call_function_end_token) : + ] + self._reset_streaming_state(True) + self.current_tool_id += 1 + continue + else: + # Tool call not complete yet, wait for more text + break + + return StreamingParseResult(normal_text=normal, calls=calls) + + def _parse_and_stream_parameters(self, text_to_parse: str) -> List[ToolCallItem]: + """ + Parse complete parameter blocks from text and return any tool call items to emit. + + This method: + 1. Finds all complete blocks + 2. Parses them into a dictionary + 3. Compares with current parameters and generates diff if needed + 4. Updates internal state + + Args: + text_to_parse: The text to search for parameter blocks + + Returns: + List of ToolCallItem objects to emit (may be empty) + """ + calls: List[ToolCallItem] = [] + + # Find all complete parameter patterns + param_matches = list( + re.finditer( + r"]+)\">(.*?)", + text_to_parse, + re.DOTALL, + ) + ) + + # Build new parameters dictionary + new_params = {} + for match in param_matches: + param_name = match.group(1).strip() + param_value = match.group(2) + new_params[param_name] = _safe_val(param_value) + + # Calculate parameter diff to stream with proper incremental JSON building + if new_params != self._current_parameters: + previous_args_json = self.streamed_args_for_tool[self.current_tool_id] + + # Build incremental JSON properly + if not self._current_parameters: + # First parameter(s) - start JSON object but don't close it yet + items = [] + for key, value in new_params.items(): + items.append( + f"{json.dumps(key, ensure_ascii=False)}: {json.dumps(value, ensure_ascii=False)}" + ) + json_fragment = "{" + ", ".join(items) + + calls.append( + ToolCallItem( + tool_index=self.current_tool_id, + name=None, + parameters=json_fragment, + ) + ) + self.streamed_args_for_tool[self.current_tool_id] = json_fragment + + else: + # Additional parameters - add them incrementally + new_keys = set(new_params.keys()) - set(self._current_parameters.keys()) + if new_keys: + # Build the continuation part (no closing brace yet) + continuation_parts = [] + for key in new_keys: + value = new_params[key] + continuation_parts.append( + f"{json.dumps(key, ensure_ascii=False)}: {json.dumps(value, ensure_ascii=False)}" + ) + + json_fragment = ", " + ", ".join(continuation_parts) + + calls.append( + ToolCallItem( + tool_index=self.current_tool_id, + name=None, + parameters=json_fragment, + ) + ) + self.streamed_args_for_tool[self.current_tool_id] = ( + previous_args_json + json_fragment + ) + + # Update current state + self._current_parameters = new_params + self.prev_tool_call_arr[self.current_tool_id]["arguments"] = new_params + + return calls + + def _reset_streaming_state(self, still_in_tool_call: bool = False): + """Reset streaming state for the next tool call""" + self._in_tool_call = still_in_tool_call + self._function_name_sent = False + self._current_function_name = "" + self._current_parameters = {} + self._streamed_parameters = {} + self.current_tool_name_sent = False + + def _extract(self, text: str, tools: List[Tool]) -> Tuple[str, List[ToolCallItem]]: + normal_parts: List[str] = [] + calls: List[ToolCallItem] = [] + cursor = 0 + while True: + s = text.find(self.tool_call_start_token, cursor) + if s == -1: + normal_parts.append(text[cursor:]) + break + normal_parts.append(text[cursor:s]) + e = text.find(self.tool_call_end_token, s) + if e == -1: + normal_parts.append(text[s:]) + break + block = text[s : e + len(self.tool_call_end_token)] + cursor = e + len(self.tool_call_end_token) + calls.extend(self._parse_block(block, tools)) + return "".join(normal_parts), calls + + def _parse_block(self, block: str, tools: List[Tool]) -> List[ToolCallItem]: + res: List[ToolCallItem] = [] + for m in self.tool_call_function_regex.findall(block): + txt = m[0] if m[0] else m[1] + if '">' not in txt: + continue + idx = txt.index('">') + fname = txt[:idx].strip() + body = txt[idx + 2 :] + params: Dict[str, Any] = {} + for pm in self.tool_call_parameter_regex.findall(body): + ptxt = pm[0] if pm[0] else pm[1] + if '">' not in ptxt: + continue + pidx = ptxt.index('">') + pname = ptxt[:pidx].strip() + pval = ptxt[pidx + 2 :].lstrip("\n").rstrip("\n") + params[pname] = _safe_val(pval) + raw = {"name": fname, "arguments": params} + try: + # TODO: fix idx in function call, the index for a function + # call will always be -1 in parse_base_json + res.extend(self.parse_base_json(raw, tools)) + except Exception: + logger.warning("invalid tool call for %s dropped", fname) + return res + + def supports_structural_tag(self) -> bool: + return False + + def structure_info(self) -> _GetInfoFunc: + raise NotImplementedError + + def build_ebnf(self, tools: List[Tool]): + return EBNFComposer.build_ebnf( + tools, + individual_call_start_token=self.tool_call_start_token.replace("\n", "\\n"), + individual_call_end_token=self.tool_call_end_token.replace("\n", "\\n"), + tool_call_separator="\\n", + function_format="xml", + call_rule_fmt='"\\n" {arguments_rule} "\\n"', + key_value_rule_fmt='"\\n" {valrule} "\\n"', + key_value_separator='"\\n"', + ) diff --git a/python/sglang/srt/models/minimax_m2.py b/python/sglang/srt/models/minimax_m2.py new file mode 100644 index 000000000..c5c5074fb --- /dev/null +++ b/python/sglang/srt/models/minimax_m2.py @@ -0,0 +1,922 @@ +# Copyright 2023-2024 SGLang Team +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# ============================================================================== + +# Adapted from DeepSeek and Mixtral implementation +"""Inference-only MiniMax M2 model compatible with HuggingFace weights.""" + +import logging +from typing import Iterable, Optional, Set, Tuple, Union + +import torch +from torch import nn +from transformers import PretrainedConfig + +from sglang.srt.distributed import ( + get_moe_expert_parallel_world_size, + get_pp_group, + get_tensor_model_parallel_rank, + get_tensor_model_parallel_world_size, + tensor_model_parallel_all_reduce, +) +from sglang.srt.eplb.expert_distribution import get_global_expert_distribution_recorder +from sglang.srt.eplb.expert_location_dispatch import ExpertLocationDispatchInfo +from sglang.srt.layers.activation import SiluAndMul +from sglang.srt.layers.communicator import ( + LayerCommunicator, + LayerScatterModes, + ScatterMode, +) +from sglang.srt.layers.layernorm import RMSNorm +from sglang.srt.layers.linear import ( + MergedColumnParallelLinear, + QKVParallelLinear, + ReplicatedLinear, + RowParallelLinear, +) +from sglang.srt.layers.logits_processor import LogitsProcessor +from sglang.srt.layers.moe.ep_moe.layer import get_moe_impl_class +from sglang.srt.layers.moe.fused_moe_triton.layer import FusedMoE +from sglang.srt.layers.moe.topk import TopK +from sglang.srt.layers.moe.utils import get_moe_a2a_backend +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.utils import PPMissingLayer +from sglang.srt.layers.vocab_parallel_embedding import ( + ParallelLMHead, + VocabParallelEmbedding, +) +from sglang.srt.model_executor.forward_batch_info import ForwardBatch, PPProxyTensors +from sglang.srt.model_loader.weight_utils import ( + default_weight_loader, + maybe_remap_kv_scale_name, +) +from sglang.srt.server_args import get_global_server_args +from sglang.srt.two_batch_overlap import model_forward_maybe_tbo +from sglang.srt.utils import ( + BumpAllocator, + add_prefix, + get_compiler_backend, + is_non_idle_and_non_empty, + make_layers, +) + +logger = logging.getLogger(__name__) + + +class MiniMaxM2RMSNormTP(nn.Module): + """RMSNorm with Tensor Parallel support for QK normalization.""" + + def __init__(self, hidden_size: int, eps: float = 1e-6) -> None: + super().__init__() + self.tp_world = get_tensor_model_parallel_world_size() + self.tp_rank = get_tensor_model_parallel_rank() + + # Weight parameter is sharded across TP ranks + self.weight = nn.Parameter(torch.ones(int(hidden_size / self.tp_world))) + self.weight.weight_loader = self.weight_loader + self.variance_epsilon = eps + + @staticmethod + def weight_loader( + param: nn.Parameter, + loaded_weight: torch.Tensor, + ) -> None: + """Custom weight loader that handles TP sharding.""" + tp_world = get_tensor_model_parallel_world_size() + tp_rank = get_tensor_model_parallel_rank() + + shard_size = loaded_weight.shape[0] // tp_world + shard = slice(tp_rank * shard_size, (tp_rank + 1) * shard_size) + param.data.copy_(loaded_weight[shard]) + + @torch.compile(dynamic=True, backend=get_compiler_backend()) + def forward( + self, + x: torch.Tensor, + residual: Optional[torch.Tensor] = None, + ) -> Union[torch.Tensor, Tuple[torch.Tensor, torch.Tensor]]: + """Forward pass with TP-aware variance computation.""" + assert residual is None, "RMSNormTP does not support residual connection." + + orig_dtype = x.dtype + x = x.to(torch.float32) + + # Compute variance across the full dimension (not just local shard) + variance = x.pow(2).mean(dim=-1, keepdim=True, dtype=torch.float32) + + if self.tp_world > 1: + # All-reduce variance across TP ranks to get global variance + variance = tensor_model_parallel_all_reduce(variance) / self.tp_world + + # Normalize and apply local weight shard + x = x * torch.rsqrt(variance + self.variance_epsilon) + x = x.to(orig_dtype) * self.weight + + return x + + +class MiniMaxM2MLP(nn.Module): + def __init__( + self, + hidden_size: int, + intermediate_size: int, + quant_config: Optional[QuantizationConfig] = None, + prefix: str = "mlp", + ) -> None: + super().__init__() + + self.gate_up_proj = MergedColumnParallelLinear( + hidden_size, + [intermediate_size] * 2, + bias=False, + quant_config=quant_config, + prefix=add_prefix("gate_up_proj", prefix), + ) + self.down_proj = RowParallelLinear( + intermediate_size, + hidden_size, + bias=False, + quant_config=quant_config, + prefix=add_prefix("down_proj", prefix), + ) + self.act_fn = SiluAndMul() + return + + def forward(self, x: torch.Tensor) -> torch.Tensor: + gate_up, _ = self.gate_up_proj(x) + x = self.act_fn(gate_up) + x, _ = self.down_proj(x) + return x + + +class MiniMaxM2MoE(nn.Module): + """MiniMax MoE implementation using DeepEP for Expert Parallel support.""" + + def __init__( + self, + config: PretrainedConfig, + layer_id: int, + quant_config: Optional[QuantizationConfig] = None, + prefix: str = "", + ): + super().__init__() + self.tp_size = get_tensor_model_parallel_world_size() + if self.tp_size > config.num_local_experts: + raise ValueError( + f"Tensor parallel size {self.tp_size} is greater than " + f"the number of experts {config.num_local_experts}." + ) + self.use_routing_bias = getattr(config, "use_routing_bias", False) + if self.use_routing_bias: + self.e_score_correction_bias = nn.Parameter( + torch.empty(config.num_local_experts, dtype=torch.float32) + ) + self.e_score_correction_bias.weight_loader = ( + MiniMaxM2MoE.ebias_weight_loader + ) + else: + self.e_score_correction_bias = None + + self.experts = get_moe_impl_class(quant_config)( + num_experts=config.num_local_experts + + get_global_server_args().ep_num_redundant_experts, + top_k=config.num_experts_per_tok, + hidden_size=config.hidden_size, + intermediate_size=config.intermediate_size, + layer_id=layer_id, + quant_config=quant_config, + prefix=add_prefix("experts", prefix), + ) + self.topk = TopK( + top_k=config.num_experts_per_tok, + renormalize=True, + scoring_func=config.scoring_func, + use_grouped_topk=True, # TODO: Use "grouped top-k" flag only for hardcoded sigmoid scoring + num_expert_group=1, + topk_group=1, + correction_bias=self.e_score_correction_bias, + routed_scaling_factor=1.0, + ) + + self.gate = ReplicatedLinear( + config.hidden_size, + config.num_local_experts, + bias=False, + params_dtype=torch.float32, + quant_config=None, + prefix=add_prefix("gate", prefix), + ) + + self.layer_id = layer_id + + if get_moe_a2a_backend().is_deepep(): + self.ep_size = get_moe_expert_parallel_world_size() + self.top_k = config.num_experts_per_tok + + @staticmethod + def ebias_weight_loader(param: nn.Parameter, loaded_weight: torch.Tensor) -> None: + assert param.size() == loaded_weight.size() + param.data.copy_(loaded_weight.to(torch.float32)) + + def forward( + self, hidden_states: torch.Tensor, forward_batch: ForwardBatch + ) -> torch.Tensor: + if get_moe_a2a_backend().is_deepep(): + return self.forward_deepep(hidden_states, forward_batch) + else: + return self.forward_normal(hidden_states) + + def forward_normal(self, hidden_states: torch.Tensor) -> torch.Tensor: + num_tokens, hidden_dim = hidden_states.shape + hidden_states = hidden_states.view(-1, hidden_dim) + + # router_logits: (num_tokens, n_experts) + router_logits, _ = self.gate(hidden_states.to(torch.float32)) + topk_output = self.topk(hidden_states, router_logits) + + final_hidden_states = self.experts(hidden_states, topk_output) + if self.tp_size > 1: + final_hidden_states = tensor_model_parallel_all_reduce(final_hidden_states) + + return final_hidden_states.view(num_tokens, hidden_dim) + + def forward_deepep( + self, hidden_states: torch.Tensor, forward_batch: ForwardBatch + ) -> torch.Tensor: + if hidden_states.shape[0] > 0: + # router_logits: (num_tokens, n_experts) + router_logits, _ = self.gate(hidden_states.to(torch.float32)) + topk_weights, topk_idx, _ = self.topk( + hidden_states, + router_logits, + num_token_non_padded=forward_batch.num_token_non_padded, + expert_location_dispatch_info=ExpertLocationDispatchInfo.init_new( + layer_id=self.layer_id, + ), + ) + else: + topk_weights, topk_idx, _ = self.topk.empty_topk_output( + hidden_states.shape[0], self.top_k + ) + final_hidden_states = self.experts( + hidden_states=hidden_states, + topk_idx=topk_idx, + topk_weights=topk_weights, + forward_batch=forward_batch, + ) + + return final_hidden_states + + # TBO Operations for MiniMax MoE + def op_gate(self, state): + """Gate operation for TBO - compute router logits""" + if is_non_idle_and_non_empty( + state.forward_batch.forward_mode, state.hidden_states_mlp_input + ): # router_logits: (num_tokens, num_experts) + state.router_logits, _ = self.gate(state.hidden_states_mlp_input) + else: + state.router_logits = None + + def op_select_experts(self, state): + """Expert selection operation for TBO""" + router_logits = state.pop("router_logits") + hidden_states = state.hidden_states_mlp_input + + if router_logits is not None: + with get_global_expert_distribution_recorder().with_current_layer( + self.layer_id + ): + state.topk_weights_local, state.topk_idx_local, _ = self.topk( + hidden_states=hidden_states, + router_logits=router_logits, + num_token_non_padded=state.forward_batch.num_token_non_padded, + expert_location_dispatch_info=ExpertLocationDispatchInfo.init_new( + layer_id=self.layer_id, + ), + ) + else: + state.topk_idx_local = torch.full( + (0, self.top_k), -1, dtype=torch.int, device=hidden_states.device + ) + state.topk_weights_local = torch.empty( + (0, self.top_k), dtype=torch.float32, device=hidden_states.device + ) + + def op_dispatch_a(self, state): + """Dispatch A operation for TBO - start async dispatch""" + if self.ep_size > 1: + self.experts.deepep_dispatcher.dispatch_a( + hidden_states=state.pop("hidden_states_mlp_input"), + topk_idx=state.pop("topk_idx_local"), + topk_weights=state.pop("topk_weights_local"), + forward_batch=state.forward_batch, + tbo_subbatch_index=state.get("tbo_subbatch_index"), + ) + + def op_dispatch_b(self, state): + """Dispatch B operation for TBO - complete async dispatch""" + if self.ep_size > 1: + with get_global_expert_distribution_recorder().with_current_layer( + self.layer_id + ): + state.dispatch_output = self.experts.deepep_dispatcher.dispatch_b( + tbo_subbatch_index=state.get("tbo_subbatch_index"), + ) + + def op_experts(self, state): + """Expert computation for TBO""" + state.hidden_states_experts_output = self.experts.moe_impl( + dispatch_output=state.dispatch_output, + ) + + def op_combine_a(self, state): + """Combine A operation for TBO - start async combine""" + if self.ep_size > 1: + self.experts.deepep_dispatcher.combine_a( + hidden_states=state.pop("hidden_states_experts_output"), + topk_idx=state.dispatch_output.topk_idx, + topk_weights=state.dispatch_output.topk_weights, + forward_batch=state.forward_batch, + tbo_subbatch_index=state.get("tbo_subbatch_index"), + ) + state.pop("dispatch_output") + + def op_combine_b(self, state): + """Combine B operation for TBO - complete async combine""" + if self.ep_size > 1: + state.hidden_states_after_combine = ( + self.experts.deepep_dispatcher.combine_b( + tbo_subbatch_index=state.get("tbo_subbatch_index"), + ) + ) + + def op_output(self, state): + """Output operation for TBO - final MLP output""" + final_hidden_states = state.pop("hidden_states_after_combine") + # MiniMax doesn't have shared experts like DeepSeek, so no need to add them + state.hidden_states_mlp_output = final_hidden_states + + +class MiniMaxM2Attention(nn.Module): + """MiniMax Attention implementation with QK normalization and partial RoPE.""" + + def __init__( + self, + config: PretrainedConfig, + layer_id: int = 0, + quant_config: Optional[QuantizationConfig] = None, + prefix: str = "", + ) -> None: + super().__init__() + self.hidden_size = config.hidden_size + tp_size = get_tensor_model_parallel_world_size() + + # Get dimensions from config + self.total_num_heads = config.num_attention_heads + assert self.total_num_heads % tp_size == 0 + self.num_heads = self.total_num_heads // tp_size + self.total_num_kv_heads = config.num_key_value_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) + + # Use head_dim from config if available, otherwise calculate + self.head_dim = getattr( + config, "head_dim", self.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 = self.head_dim**-0.5 + + # RoPE settings - support partial RoPE + self.rope_theta = getattr(config, "rope_theta", 10000) + self.max_position_embeddings = getattr(config, "max_position_embeddings", 8192) + self.rotary_dim = getattr( + config, "rotary_dim", self.head_dim + ) # MiniMax uses rotary_dim=64 + + # QK Normalization settings + self.use_qk_norm = getattr(config, "use_qk_norm", False) + self.qk_norm_type = getattr(config, "qk_norm_type", "per_layer") + + self.qkv_proj = QKVParallelLinear( + self.hidden_size, + self.head_dim, + self.total_num_heads, + self.total_num_kv_heads, + bias=False, + quant_config=quant_config, + prefix=add_prefix("qkv_proj", prefix), + ) + + self.o_proj = RowParallelLinear( + self.total_num_heads * self.head_dim, + self.hidden_size, + bias=False, + reduce_results=False, + quant_config=quant_config, + prefix=add_prefix("o_proj", prefix), + ) + + # Setup RoPE with partial rotary dimension + rope_scaling = getattr(config, "rope_scaling", None) + self.rotary_emb = get_rope( + self.head_dim, + rotary_dim=self.rotary_dim, # Use partial rotary dimension + max_position=self.max_position_embeddings, + base=self.rope_theta, + rope_scaling=rope_scaling, + ) + + # QK Normalization layers + if self.use_qk_norm: + if self.qk_norm_type == "per_layer": + # Use RMSNormTP for proper tensor parallel support + # Use total dimensions (before TP sharding) for correct normalization + self.q_norm = MiniMaxM2RMSNormTP( + self.total_num_heads * self.head_dim, eps=config.rms_norm_eps + ) + self.k_norm = MiniMaxM2RMSNormTP( + self.total_num_kv_heads * self.head_dim, eps=config.rms_norm_eps + ) + else: + raise ValueError(f"Unsupported qk_norm_type: {self.qk_norm_type}") + + 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=add_prefix("attn", prefix), + ) + + def forward_prepare( + self, + positions: torch.Tensor, + hidden_states: torch.Tensor, + forward_batch: ForwardBatch, + ): + qkv, _ = self.qkv_proj(hidden_states) + q, k, v = qkv.split([self.q_size, self.kv_size, self.kv_size], dim=-1) + if self.use_qk_norm: + q = self.q_norm(q.contiguous()) + k = self.k_norm(k.contiguous()) + else: + q, k = q.contiguous(), k.contiguous() + q, k = self.rotary_emb(positions, q, k) + inner_state = q, k, v, forward_batch + return None, forward_batch, inner_state + + def forward_core(self, intermediate_state): + _, _, inner_state = intermediate_state + attn_output = self.attn(*inner_state) + output, _ = self.o_proj(attn_output) + return output + + def forward( + self, + positions: torch.Tensor, + hidden_states: torch.Tensor, + forward_batch: ForwardBatch, + ) -> torch.Tensor: + s = self.forward_prepare( + positions=positions, + hidden_states=hidden_states, + forward_batch=forward_batch, + ) + return self.forward_core(s) + + def op_prepare(self, state): + state.attn_intermediate_state = self.forward_prepare( + positions=state.positions, + hidden_states=state.pop("hidden_states_after_comm_pre_attn"), + forward_batch=state.forward_batch, + ) + + def op_core(self, state): + state.hidden_states_after_attn = self.forward_core( + state.pop("attn_intermediate_state") + ) + + +class MiniMaxM2DecoderLayer(nn.Module): + """MiniMax Decoder Layer implementation with MoE support.""" + + def __init__( + self, + config: PretrainedConfig, + layer_id: int, + quant_config: Optional[QuantizationConfig] = None, + prefix: str = "", + ) -> None: + super().__init__() + self.hidden_size = config.hidden_size + self.layer_id = layer_id + + # TBO support: All MiniMax layers are sparse (MoE) + self.is_layer_sparse = True + + self.self_attn = MiniMaxM2Attention( + config=config, + layer_id=layer_id, + quant_config=quant_config, + prefix=add_prefix("self_attn", prefix), + ) + + self.block_sparse_moe = MiniMaxM2MoE( + config=config, + layer_id=layer_id, + quant_config=quant_config, + prefix=add_prefix("mlp", prefix), + ) + + self.input_layernorm = RMSNorm( + config.hidden_size, eps=getattr(config, "rms_norm_eps", 1e-6) + ) + self.post_attention_layernorm = RMSNorm( + config.hidden_size, eps=getattr(config, "rms_norm_eps", 1e-6) + ) + + is_previous_layer_sparse = True + self.layer_scatter_modes = LayerScatterModes.init_new( + layer_id=layer_id, + num_layers=config.num_hidden_layers, + is_layer_sparse=self.is_layer_sparse, + is_previous_layer_sparse=is_previous_layer_sparse, + ) + + self.layer_communicator = LayerCommunicator( + layer_scatter_modes=self.layer_scatter_modes, + input_layernorm=self.input_layernorm, + post_attention_layernorm=self.post_attention_layernorm, + allow_reduce_scatter=True, + ) + + def forward( + self, + positions: torch.Tensor, + hidden_states: torch.Tensor, + forward_batch: ForwardBatch, + residual: Optional[torch.Tensor], + ) -> torch.Tensor: + # Self Attention + hidden_states, residual = self.layer_communicator.prepare_attn( + hidden_states, residual, forward_batch + ) + + hidden_states = self.self_attn( + positions=positions, + hidden_states=hidden_states, + forward_batch=forward_batch, + ) + + # Fully Connected (MLP or MoE) + + hidden_states, residual = self.layer_communicator.prepare_mlp( + hidden_states, residual, forward_batch + ) + + hidden_states = self.block_sparse_moe(hidden_states, forward_batch) + + hidden_states, residual = self.layer_communicator.postprocess_layer( + hidden_states, residual, forward_batch + ) + + return hidden_states, residual + + # TBO Operations for MiniMax Decoder Layer + def op_comm_prepare_attn( + self, + state, + positions: torch.Tensor, + hidden_states: torch.Tensor, + forward_batch: ForwardBatch, + residual: Optional[torch.Tensor], + zero_allocator: BumpAllocator, + tbo_subbatch_index: Optional[int] = None, + ): + """Communication prepare for attention - TBO operation""" + state.hidden_states_after_comm_pre_attn, state.residual_after_input_ln = ( + self.layer_communicator.prepare_attn(hidden_states, residual, forward_batch) + ) + state.update( + dict( + forward_batch=forward_batch, + positions=positions, + zero_allocator=zero_allocator, + tbo_subbatch_index=tbo_subbatch_index, + ) + ) + + def op_comm_prepare_mlp(self, state): + """Communication prepare for MLP - TBO operation""" + state.hidden_states_mlp_input, state.residual_after_comm_pre_mlp = ( + self.layer_communicator.prepare_mlp( + state.pop("hidden_states_after_attn"), + state.pop("residual_after_input_ln"), + state.forward_batch, + ) + ) + + def op_mlp(self, state): + hidden_states = state.pop("hidden_states_mlp_input") + state.hidden_states_mlp_output = self.block_sparse_moe( + hidden_states, state.forward_batch + ) + + def op_comm_postprocess_layer(self, state): + """Communication postprocess for layer - TBO operation""" + hidden_states, residual = self.layer_communicator.postprocess_layer( + state.pop("hidden_states_mlp_output"), + state.pop("residual_after_comm_pre_mlp"), + state.forward_batch, + ) + + output = dict( + positions=state.positions, + hidden_states=hidden_states, + residual=residual, + forward_batch=state.forward_batch, + zero_allocator=state.zero_allocator, + tbo_subbatch_index=state.tbo_subbatch_index, + ) + return output + + +class MiniMaxM2Model(nn.Module): + """MiniMax Model implementation.""" + + fall_back_to_pt_during_load = False + + def __init__( + self, + config: PretrainedConfig, + quant_config: Optional[QuantizationConfig] = None, + prefix: str = "", + ) -> None: + super().__init__() + + self.padding_idx = getattr(config, "pad_token_id", 0) + self.vocab_size = config.vocab_size + self.pp_group = get_pp_group() + + self.embed_tokens = VocabParallelEmbedding( + config.vocab_size, + config.hidden_size, + ) + + def layer_fn(idx, prefix: str) -> nn.Module: + return MiniMaxM2DecoderLayer( + config=config, + layer_id=idx, + quant_config=quant_config, + prefix=prefix, + ) + + self.layers, self.start_layer, self.end_layer = make_layers( + config.num_hidden_layers, + layer_fn, + pp_rank=self.pp_group.rank_in_group, + pp_size=self.pp_group.world_size, + prefix=add_prefix("layers", prefix), + ) + if self.pp_group.is_last_rank: + self.norm = RMSNorm(config.hidden_size, eps=config.rms_norm_eps) + else: + self.norm = PPMissingLayer(return_tuple=True) + + 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, + input_embeds: torch.Tensor = None, + pp_proxy_tensors: Optional[PPProxyTensors] = None, + ) -> Union[torch.Tensor, PPProxyTensors]: + if self.pp_group.is_first_rank: + if input_embeds is None: + hidden_states = self.get_input_embeddings(input_ids) + else: + hidden_states = input_embeds + residual = None + else: + assert pp_proxy_tensors is not None + hidden_states = pp_proxy_tensors["hidden_states"] + residual = pp_proxy_tensors["residual"] + + if forward_batch.can_run_tbo: + hidden_states, residual = model_forward_maybe_tbo( + layers=self.layers, + enable_tbo=True, + input_data_scatter_mode=ScatterMode.model_input_output(), + positions=positions, + forward_batch=forward_batch, + hidden_states=hidden_states, + residual=residual, + ) + else: + for i in range(self.start_layer, self.end_layer): + with get_global_expert_distribution_recorder().with_current_layer(i): + layer = self.layers[i] + hidden_states, residual = layer( + positions=positions, + forward_batch=forward_batch, + hidden_states=hidden_states, + residual=residual, + ) + + if not self.pp_group.is_last_rank: + return PPProxyTensors( + {"hidden_states": hidden_states, "residual": residual} + ) + + if residual is not None: + hidden_states, _ = self.norm(hidden_states, residual) + else: + hidden_states = self.norm(hidden_states) + + return hidden_states + + +class MiniMaxM2ForCausalLM(nn.Module): + """MiniMax M2 model for causal language modeling.""" + + def __init__( + self, + config: PretrainedConfig, + quant_config: Optional[QuantizationConfig] = None, + prefix: str = "", + ) -> None: + super().__init__() + + self.config = config + self.quant_config = quant_config + + self.model = MiniMaxM2Model( + config, quant_config, prefix=add_prefix("model", prefix) + ) + + if get_pp_group().is_last_rank: + self.lm_head = ParallelLMHead( + config.vocab_size, + config.hidden_size, + quant_config=None, + prefix=add_prefix("lm_head", prefix), + ) + else: + self.lm_head = PPMissingLayer() + + self.logits_processor = LogitsProcessor(config) + + def get_input_embeddings(self, input_ids: torch.Tensor) -> torch.Tensor: + return self.model.get_input_embeddings(input_ids) + + @torch.no_grad() + def forward( + self, + input_ids: torch.Tensor, + positions: torch.Tensor, + forward_batch: ForwardBatch, + input_embeds: torch.Tensor = None, + ) -> torch.Tensor: + # _print_tensor_info(input_ids, "input_ids") + hidden_states = self.model(input_ids, positions, forward_batch, input_embeds) + return self.logits_processor( + input_ids, hidden_states, self.lm_head, forward_batch + ) + + def load_weights(self, weights: Iterable[Tuple[str, torch.Tensor]]): + """Load model weights with proper mapping for MiniMax architecture.""" + + stacked_params_mapping = [ + # (param_name, shard_name, shard_id) + ("qkv_proj", "q_proj", "q"), + ("qkv_proj", "k_proj", "k"), + ("qkv_proj", "v_proj", "v"), + ("gate_up_proj", "gate_proj", 0), + ("gate_up_proj", "up_proj", 1), + ] + + # Params for weights, fp8 weight scales, fp8 activation scales + # (param_name, weight_name, expert_id, shard_id) + expert_params_mapping = FusedMoE.make_expert_params_mapping( + ckpt_gate_proj_name="w1", + ckpt_down_proj_name="w2", + ckpt_up_proj_name="w3", + num_experts=self.config.num_local_experts, + ) + + params_dict = dict(self.named_parameters()) + loaded_params: Set[str] = set() + for name, loaded_weight in weights: + if "rotary_emb.inv_freq" in name: + continue + + spec_layer = get_spec_layer_idx_from_weight_name(self.config, name) + if spec_layer is not None: + continue # skip spec decode layers for main model + + for param_name, weight_name, shard_id in stacked_params_mapping: + # Skip non-stacked layers and experts (experts handled below). + if weight_name not in name: + continue + # We have mlp.experts[0].gate_proj in the checkpoint. + # Since we handle the experts below in expert_params_mapping, + # we need to skip here BEFORE we update the name, otherwise + # name will be updated to mlp.experts[0].gate_up_proj, which + # will then be updated below in expert_params_mapping + # for mlp.experts[0].gate_gate_up_proj, which breaks load. + if ("mlp.experts." in name) and name not in params_dict: + continue + name = name.replace(weight_name, param_name) + # Skip loading extra bias for GPTQ models. + if name.endswith(".bias") and name not in params_dict: + continue + + param = params_dict[name] + weight_loader = param.weight_loader + weight_loader(param, loaded_weight, shard_id) + break + else: + for mapping in expert_params_mapping: + param_name, weight_name, expert_id, shard_id = mapping + if weight_name not in name: + continue + name = name.replace(weight_name, param_name) + + param = params_dict[name] + weight_loader = param.weight_loader + weight_loader( + param, + loaded_weight, + name, + shard_id=shard_id, + expert_id=expert_id, + ) + break + else: + # Skip loading extra bias for GPTQ models. + if name.endswith(".bias") and name not in params_dict: + continue + + # Remapping the name of FP8 kv-scale. + name = maybe_remap_kv_scale_name(name, params_dict) + if name is None: + continue + + param = params_dict[name] + weight_loader = getattr( + param, "weight_loader", default_weight_loader + ) + weight_loader(param, loaded_weight) + loaded_params.add(name) + return loaded_params + + @classmethod + def get_model_config_for_expert_location(cls, config): + from sglang.srt.eplb.expert_location import ModelConfigForExpertLocation + + return ModelConfigForExpertLocation( + num_layers=config.num_hidden_layers, + num_logical_experts=config.num_local_experts, + num_groups=None, + ) + + +def get_spec_layer_idx_from_weight_name( + config: PretrainedConfig, weight_name: str +) -> Optional[int]: + if hasattr(config, "num_mtp_modules") and (config.num_mtp_modules > 0): + layer_idx = config.num_hidden_layers + for i in range(config.num_mtp_modules): + if weight_name.startswith(f"model.layers.{layer_idx + i}."): + return layer_idx + i + return None + + +# Entry class for model registration +EntryClass = MiniMaxM2ForCausalLM diff --git a/python/sglang/srt/parser/reasoning_parser.py b/python/sglang/srt/parser/reasoning_parser.py index 0c01ede9c..2aefeedb1 100644 --- a/python/sglang/srt/parser/reasoning_parser.py +++ b/python/sglang/srt/parser/reasoning_parser.py @@ -249,6 +249,31 @@ class GptOssDetector(BaseReasoningFormatDetector): ) +class MiniMaxAppendThinkDetector(BaseReasoningFormatDetector): + """ + Append `` token to the beginning of the text. + """ + + def __init__(self, stream_reasoning: bool = True, force_reasoning: bool = False): + # scheduler.py need `reasoning_parser.detector.think_end_token` + super().__init__( + "", + "", + force_reasoning=force_reasoning, + stream_reasoning=stream_reasoning, + ) + self.is_first_chunk = False + + def parse_streaming_increment(self, new_text: str) -> StreamingParseResult: + if not self.is_first_chunk: + self.is_first_chunk = True + new_text = self.think_start_token + new_text + return StreamingParseResult(normal_text=new_text) + + def detect_and_parse(self, text: str) -> StreamingParseResult: + return StreamingParseResult(normal_text=self.think_start_token + text) + + class ReasoningParser: """ Parser that handles both streaming and non-streaming scenarios for extracting @@ -268,6 +293,8 @@ class ReasoningParser: "kimi": KimiDetector, "qwen3": Qwen3Detector, "qwen3-thinking": Qwen3Detector, + "minimax": Qwen3Detector, + "minimax-append-think": MiniMaxAppendThinkDetector, "step3": DeepSeekR1Detector, } @@ -285,7 +312,7 @@ class ReasoningParser: raise ValueError(f"Unsupported model type: {model_type}") # Special cases where we override force_reasoning - if model_type.lower() in {"qwen3-thinking", "gpt-oss"}: + if model_type.lower() in {"qwen3-thinking", "gpt-oss", "minimax"}: force_reasoning = True # Only pass force_reasoning if explicitly set, let detectors use their defaults