BF16 from MiniMaxAI/MiniMax-M2.7 (FP8 → BF16 blockwise dequant, exact)

.gitattributes

@@ -33,3 +33,7 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+figures/agent_harness.png filter=lfs diff=lfs merge=lfs -text
+figures/agent_teams.gif filter=lfs diff=lfs merge=lfs -text
+figures/banner.png filter=lfs diff=lfs merge=lfs -text
+figures/mle_bench.png filter=lfs diff=lfs merge=lfs -text

LICENSE

@@ -0,0 +1,18 @@
+NON-COMMERCIAL LICENSE
+Non-commercial use permitted based on MIT-style terms; commercial use requires prior written authorization.
+Copyright (c) 2026 MiniMax
+Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software for non-commercial purposes, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or provide copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
+1. The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
+2. If the Software (or any derivative works thereof) is used for any Commercial Use, you shall prominently display "Built with MiniMax M2.7" on a related website, user interface, blogpost, about page or product documentation.
+3. Any Commercial Use of the Software or any derivative work thereof is prohibited without obtaining a separate, prior written authorization from MiniMax.  To request such authorization, please contact api@minimax.io with the subject line "M2.7 licensing".
+4. "Commercial Use" means any use of the Software or any derivative work thereof that is primarily intended for commercial advantage or monetary compensation, which includes, without limitation: (i) offering products or services to third parties for a fee, which utilize, incorporate, or rely on the Software or its derivatives, (ii) the commercial use of APIs provided by or for the Software or its derivatives, including to support or enable commercial products, services, or operations, whether in a cloud-based, hosted, or other similar environment, and (iii) the deployment or provision of the Software or its derivatives that have been subjected to post-training, fine-tuning, instruction-tuning, or any other form of modification, for any commercial purpose.
+5. Permitted Free Uses. The following uses are expressly permitted free of charge: (a) personal use, including self-hosted deployment for coding, development of applications, agents, tools, integrations, research, experimentation, or other personal purposes; (b) use by non-profit organizations, academic institutions, and researchers for non-commercial research or educational purposes; (c) modification of the Software solely for the uses described in (a) or (b) above.
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+Appendix: Prohibited Uses
+You agree you will not use, or allow others to use, the Software or any derivatives of the Software to:
+1. Generate or disseminate content prohibited by applicable laws or regulations.
+2. Assist with, engage in or otherwise support any military purpose.
+3. Exploit, harm, or attempt to exploit or harm minors.
+4. Generate or disseminate false or misleading information with the intent to cause harm.
+5. Promote discrimination, hate speech, or harmful behavior against individuals or groups based on race or ethnic origin, religion, disability, age, nationality and national origin, veteran status, sexual orientation, gender or gender identity, caste, immigration status, or any other characteristic that is associated with systemic discrimination or marginalization.

README.md

@@ -0,0 +1,95 @@
+---
+license: other
+license_name: minimax-license
+license_link: https://huggingface.co/MiniMaxAI/MiniMax-M2.7/blob/main/LICENSE
+base_model: MiniMaxAI/MiniMax-M2.7
+base_model_relation: quantized
+library_name: transformers
+pipeline_tag: text-generation
+language:
+  - en
+  - zh
+  - ru
+tags:
+  - minimax
+  - minimax-m2
+  - moe
+  - mixture-of-experts
+  - bf16
+  - dequantized
+---
+
+# MiniMax-M2.7 — BF16 (dequantized from FP8)
+
+Plain `bfloat16` weights of [MiniMaxAI/MiniMax-M2.7](https://huggingface.co/MiniMaxAI/MiniMax-M2.7),
+reconstructed from the upstream block-FP8 (E4M3, 128×128 blocks) checkpoint
+via shard-by-shard blockwise dequantization. **No calibration, no rounding loss
+beyond the original FP8→BF16 cast** — every block is materialized exactly:
+
+```
+bf16_block = (fp8_block.float() * scale_fp32).bfloat16()
+```
+
+## Why this exists
+
+`MiniMaxAI/MiniMax-M2.7` ships natively in FP8. On Ampere and earlier
+(e.g. RTX A5000) FP8 tensor cores don't exist and inference engines have
+to emulate FP8 through FP16 — paying double bandwidth without the speed
+benefit. For further offline quantization (AWQ, GPTQ, RTN INT8, …) you
+need plain BF16 weights anyway: `transformers + torch_dtype=bfloat16`
+won't materialize the attention projections under the FP8 quant config,
+which trips up `llmcompressor`'s GPTQ tracer.
+
+This repo is the missing intermediate: **upstream MiniMax-M2.7 weights in
+plain BF16 safetensors**, ready to be fed into any standard quantization
+pipeline.
+
+## Contents
+
+- 47 shards `model-NNNNN-of-00047.safetensors`
+- rebuilt `model.safetensors.index.json` (no `*.weight_scale_inv` entries)
+- `config.json` with the upstream `quantization_config` stripped
+- tokenizer + custom modeling `.py` files copied verbatim from the FP8 source
+
+Total ≈ **458 GB**.
+
+## Provenance
+
+Produced on a single 48 GB GPU pod (~30 minutes wall time) using a
+~150-line script — see
+[`dequant_fp8_blockwise.py`](https://github.com/operationrange/zonatelecom-agent/blob/main/scripts/quant/dequant_fp8_blockwise.py).
+
+Process per shard:
+
+1. open `model-XXXXX-of-00130.safetensors` from the FP8 source
+2. for each `*.weight` (FP8 e4m3fn): look up `*.weight_scale_inv` (FP32, 128×128)
+3. broadcast scale to weight shape, multiply, cast to BF16
+4. drop the scale tensor
+5. write `model-NNNNN-of-00047.safetensors` (5 GB shards)
+
+Other tensors (embeddings, layer norms, MoE routers/gates that were already
+unquantized in the upstream config's `modules_to_not_convert`) are passed
+through with a BF16 cast.
+
+## Quick load
+
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+m = AutoModelForCausalLM.from_pretrained(
+    "operationrange/MiniMax-M2.7-BF16",
+    torch_dtype="bfloat16",
+    device_map="auto",
+    trust_remote_code=True,
+)
+tok = AutoTokenizer.from_pretrained("operationrange/MiniMax-M2.7-BF16", trust_remote_code=True)
+```
+
+Inference at full BF16 needs ≥ ~470 GB combined GPU+CPU memory, so this
+checkpoint is mostly intended as a starting point for further compression
+(AWQ-INT4, GPTQ-INT8, etc.) rather than direct serving.
+
+## License
+
+Inherits the [MiniMax-M2 license](https://huggingface.co/MiniMaxAI/MiniMax-M2.7/blob/main/LICENSE) from the
+upstream model. No weights were modified — only the storage format.

chat_template.jinja

@@ -0,0 +1,159 @@
+{# ----------‑‑‑ special token variables ‑‑‑---------- #}
+{%- set toolcall_begin_token   = '<minimax:tool_call>'         -%}
+{%- set toolcall_end_token     = '</minimax:tool_call>'        -%}
+{#- Tool Rendering Functions ============================================== -#}
+{%- macro render_tool_namespace(namespace_name, tool_list) -%}
+{%- for tool in tool_list -%}
+<tool>{{ tool.function | tojson(ensure_ascii=False) }}</tool>
+{% endfor -%}
+{%- endmacro -%}
+{%- macro visible_text(content) -%}
+    {%- if content is string -%}
+        {{ content }}
+    {%- elif content is iterable and content is not mapping -%}
+        {%- for item in content -%}
+            {%- if item is mapping and item.type == 'text' -%}
+                {{- item.text }}
+            {%- elif item is string -%}
+                {{- item }}
+            {%- endif -%}
+        {%- endfor -%}
+    {%- else -%}
+        {{- content }}
+    {%- endif -%}
+{%- endmacro -%}
+{#- System Message Construction ============================================ -#}
+{%- macro build_system_message(system_message) -%}
+    {%- if system_message and system_message.content -%}
+        {{- visible_text(system_message.content) }}
+    {%- else -%}
+        {%- if model_identity is not defined -%}
+            {%- set model_identity = "You are a helpful assistant. Your name is MiniMax-M2.7 and is built by MiniMax." -%}
+        {%- endif -%}
+        {{- model_identity }}
+    {%- endif -%}
+    
+    {#- Handle current_date -#}
+    {%- if system_message and system_message.current_date -%}
+        {{- '\n' ~ 'Current date: ' + system_message.current_date }}
+    {%- endif -%}
+    {#- Handle current_location -#}
+    {%- if system_message and system_message.current_location -%}
+        {{- '\n' ~ 'Current location: ' + system_message.current_location }}
+    {%- endif -%}
+{%- endmacro -%}
+{#- Main Template Logic ================================================= -#}
+{#- Extract system message (only first message if it's system) -#}
+{%- set system_message = none -%}
+{%- set conversation_messages = messages -%}
+{%- if messages and messages[0].role == "system" -%}
+    {%- set system_message = messages[0] -%}
+    {%- set conversation_messages = messages[1:] -%}
+{%- endif -%}
+{#- Get the last user message turn, for interleved thinking -#}
+{%- set ns = namespace(last_user_index=-1) %}
+{% for m in conversation_messages %}
+    {%- if m.role == 'user' %}
+        {% set ns.last_user_index = loop.index0 -%}
+    {%- endif %}
+{%- endfor %}
+{#- Render system message -#}
+{{- ']~!b[' ~ ']~b]system' ~ '\n' }}
+{{- build_system_message(system_message) }}
+{#- Render tools if available -#}
+{%- if tools -%}
+    {{- '\n\n' ~ '# Tools' ~ '\n' ~ 'You may call one or more tools to assist with the user query.\nHere are the tools available in JSONSchema format:' ~ '\n' }}
+    {{- '\n' ~ '<tools>' ~ '\n' }}
+    {{- render_tool_namespace("functions", tools) }}
+    {{- '</tools>' ~ '\n\n' }}
+{{- 'When making tool calls, use XML format to invoke tools and pass parameters:' ~ '\n' }}
+{{- '\n' ~ toolcall_begin_token }}
+<invoke name="tool-name-1">
+<parameter name="param-key-1">param-value-1</parameter>
+<parameter name="param-key-2">param-value-2</parameter>
+...
+</invoke>
+{{- '\n' ~ toolcall_end_token }}
+{%- endif -%}
+{{- '[e~[\n' }}
+
+{#- Render messages -#}
+{%- set last_tool_call = namespace(name=none) -%}
+{%- for message in conversation_messages -%}
+    {%- if message.role == 'assistant' -%}
+        {#- Only render reasoning_content if no user message follows -#}
+        {{- ']~b]ai' ~ '\n' }}
+
+        {%- set reasoning_content = '' %}
+        {%- set content = visible_text(message.content) %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].strip('\n').split('<think>')[-1].strip('\n') %}
+                {%- set content = content.split('</think>')[-1].strip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if reasoning_content and loop.index0 > ns.last_user_index -%}
+            {{- '<think>' ~ '\n' ~ reasoning_content ~ '\n' ~ '</think>' ~ '\n\n' }}
+        {%- endif -%}
+        {%- if content -%}
+            {{- content }}
+        {%- endif -%}
+        {%- if message.tool_calls -%}
+            {{- '\n' ~ toolcall_begin_token ~ '\n' }}
+
+            {%- for tool_call in message.tool_calls -%}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<invoke name="' + tool_call.name + '">' }}
+                {% set _args = tool_call.arguments %}
+                {%- for k, v in _args.items() %}
+                {{- '<parameter name="' + k + '">' }}
+                {{- v | tojson(ensure_ascii=False) if v is not string else v }}
+                {{- '</parameter>' }}
+                {% endfor %}
+                {{- '</invoke>' ~ '\n' }}
+            {%- endfor -%}
+            
+            {{- toolcall_end_token}}
+            {%- set last_tool_call.name = message.tool_calls[-1].name -%}
+        {%- else -%}
+            {%- set last_tool_call.name = none -%}
+        {%- endif -%}
+        {{- '[e~[' ~ '\n' }}
+        
+    {%- elif message.role == 'tool' -%}
+    {%- if last_tool_call.name is none -%}
+        {{- raise_exception("Message has tool role, but there was no previous assistant message with a tool call!") }}
+    {%- endif -%}
+    {%- if loop.first or (conversation_messages[loop.index0 - 1].role != 'tool') -%}
+        {{- ']~b]tool' }}
+    {%- endif -%}
+    {%- if message.content is string -%}
+        {{- '\n<response>' }}
+        {{- message.content }}
+        {{- '</response>' }}
+    {%- else -%}
+        {%- for tr in message.content -%}
+            {{- '\n<response>' }}
+            {{- tr.output if tr.output is defined else (tr.text if tr.type == 'text' and tr.text is defined else tr) }}
+            {{- '\n</response>' }}
+        {%- endfor -%}
+    {%- endif -%}
+    {%- if loop.last or (conversation_messages[loop.index0 + 1].role != 'tool') -%}
+        {{- '[e~[\n' -}}
+    {%- endif -%}
+        
+    {%- elif message.role == 'user' -%}
+        {{- ']~b]user' ~ '\n' }}
+        {{- visible_text(message.content) }}
+        {{- '[e~[' ~ '\n' }}
+    {%- endif -%}
+{%- endfor -%}
+
+{#- Generation prompt -#}
+{%- if add_generation_prompt -%}
+{{- ']~b]ai' ~ '\n' ~ '<think>' ~ '\n' }}
+{%- endif -%}

config.json

@@ -0,0 +1,100 @@
+{
+  "architectures": [
+    "MiniMaxM2ForCausalLM"
+  ],
+  "attn_type_list": [
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1
+  ],
+  "auto_map": {
+    "AutoConfig": "configuration_minimax_m2.MiniMaxM2Config",
+    "AutoModelForCausalLM": "modeling_minimax_m2.MiniMaxM2ForCausalLM"
+  },
+  "dtype": "bfloat16",
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 3072,
+  "intermediate_size": 1536,
+  "max_position_embeddings": 204800,
+  "model_type": "minimax_m2",
+  "mtp_transformer_layers": 1,
+  "num_attention_heads": 48,
+  "num_experts_per_tok": 8,
+  "num_hidden_layers": 62,
+  "num_key_value_heads": 8,
+  "num_local_experts": 256,
+  "num_mtp_modules": 3,
+  "qk_norm_type": "per_layer",
+  "rms_norm_eps": 1e-06,
+  "rope_theta": 5000000,
+  "rotary_dim": 64,
+  "scoring_func": "sigmoid",
+  "shared_intermediate_size": 0,
+  "tie_word_embeddings": false,
+  "transformers_version": "4.46.1",
+  "use_cache": true,
+  "use_mtp": true,
+  "use_qk_norm": true,
+  "use_routing_bias": true,
+  "vocab_size": 200064
+}

configuration_minimax_m2.py

@@ -0,0 +1,200 @@
+#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
+#           This file was automatically generated from src/transformers/models/minimax_m2/modular_minimax_m2.py.
+#               Do NOT edit this file manually as any edits will be overwritten by the generation of
+#             the file from the modular. If any change should be done, please apply the change to the
+#                          modular_minimax_m2.py file directly. One of our CI enforces this.
+#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
+# coding=utf-8
+# Copyright 2025 the HuggingFace Team. All rights reserved.
+#
+# 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.
+
+
+from transformers.configuration_utils import PretrainedConfig
+
+
+class MiniMaxM2Config(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of a [`MiniMaxM2Model`]. It is used to instantiate an
+    MiniMaxM2 model according to the specified arguments, defining the model architecture. Instantiating a configuration
+    with the defaults will yield a similar configuration to that of the MiniMaxM2-7B-v0.1 or MiniMaxM2-7B-Instruct-v0.1.
+
+    [minimax_m2ai/MiniMaxM2-8x7B](https://huggingface.co/minimax_m2ai/MiniMaxM2-8x7B)
+    [minimax_m2ai/MiniMaxM2-7B-Instruct-v0.1](https://huggingface.co/minimax_m2ai/MiniMaxM2-7B-Instruct-v0.1)
+
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+
+
+    Args:
+        vocab_size (`int`, *optional*, defaults to 32000):
+            Vocabulary size of the MiniMaxM2 model. Defines the number of different tokens that can be represented by the
+            `inputs_ids` passed when calling [`MiniMaxM2Model`]
+        hidden_size (`int`, *optional*, defaults to 4096):
+            Dimension of the hidden representations.
+        intermediate_size (`int`, *optional*, defaults to 14336):
+            Dimension of the MLP representations.
+        num_hidden_layers (`int`, *optional*, defaults to 32):
+            Number of hidden layers in the Transformer encoder.
+        num_attention_heads (`int`, *optional*, defaults to 32):
+            Number of attention heads for each attention layer in the Transformer encoder.
+        num_key_value_heads (`int`, *optional*, defaults to 8):
+            This is the number of key_value heads that should be used to implement Grouped Query Attention. If
+            `num_key_value_heads=num_attention_heads`, the model will use Multi Head Attention (MHA), if
+            `num_key_value_heads=1` the model will use Multi Query Attention (MQA) otherwise GQA is used. When
+            converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be constructed
+            by meanpooling all the original heads within that group. For more details, check out [this
+            paper](https://huggingface.co/papers/2305.13245). If it is not specified, will default to `8`.
+        head_dim (`int`, *optional*, defaults to `hidden_size // num_attention_heads`):
+            The attention head dimension.
+        hidden_act (`str` or `function`, *optional*, defaults to `"silu"`):
+            The non-linear activation function (function or string) in the decoder.
+        max_position_embeddings (`int`, *optional*, defaults to `4096*32`):
+            The maximum sequence length that this model might ever be used with. MiniMaxM2's sliding window attention
+            allows sequence of up to 4096*32 tokens.
+        initializer_range (`float`, *optional*, defaults to 0.02):
+            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
+        rms_norm_eps (`float`, *optional*, defaults to 1e-05):
+            The epsilon used by the rms normalization layers.
+        use_cache (`bool`, *optional*, defaults to `True`):
+            Whether or not the model should return the last key/values attentions (not used by all models). Only
+            relevant if `config.is_decoder=True`.
+        pad_token_id (`int`, *optional*):
+            The id of the padding token.
+        bos_token_id (`int`, *optional*, defaults to 1):
+            The id of the "beginning-of-sequence" token.
+        eos_token_id (`int`, *optional*, defaults to 2):
+            The id of the "end-of-sequence" token.
+        tie_word_embeddings (`bool`, *optional*, defaults to `False`):
+            Whether the model's input and output word embeddings should be tied.
+        rope_theta (`float`, *optional*, defaults to 1000000.0):
+            The base period of the RoPE embeddings.
+        sliding_window (`int`, *optional*):
+            Sliding window attention window size. If not specified, will default to `4096`.
+        attention_dropout (`float`, *optional*, defaults to 0.0):
+            The dropout ratio for the attention probabilities.
+        num_experts_per_tok (`int`, *optional*, defaults to 2):
+            The number of experts to route per-token, can be also interpreted as the `top-k` routing
+            parameter
+        num_local_experts (`int`, *optional*, defaults to 8):
+            Number of experts per Sparse MLP layer.
+        output_router_logits (`bool`, *optional*, defaults to `False`):
+            Whether or not the router logits should be returned by the model. Enabling this will also
+            allow the model to output the auxiliary loss. See [here]() for more details
+        router_aux_loss_coef (`float`, *optional*, defaults to 0.001):
+            The aux loss factor for the total loss.
+        router_jitter_noise (`float`, *optional*, defaults to 0.0):
+            Amount of noise to add to the router.
+
+    ```python
+    >>> from transformers import MiniMaxM2Model, MiniMaxM2Config
+
+    >>> # Initializing a MiniMaxM2 7B style configuration
+    >>> configuration = MiniMaxM2Config()
+
+    >>> # Initializing a model from the MiniMaxM2 7B style configuration
+    >>> model = MiniMaxM2Model(configuration)
+
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```"""
+
+    model_type = "minimax_m2"
+    keys_to_ignore_at_inference = ["past_key_values"]
+    base_model_tp_plan = {
+        "layers.*.self_attn.q_proj": "colwise",
+        "layers.*.self_attn.k_proj": "colwise",
+        "layers.*.self_attn.v_proj": "colwise",
+        "layers.*.self_attn.o_proj": "rowwise",
+        "layers.*.block_sparse_moe.gate": "colwise_rep",  # we need to replicate here to correctly route experts
+        "layers.*.block_sparse_moe.experts.*.w1": "colwise",
+        "layers.*.block_sparse_moe.experts.*.w2": "rowwise",
+        "layers.*.block_sparse_moe.experts.*.w3": "colwise",
+    }
+    base_model_pp_plan = {
+        "embed_tokens": (["input_ids"], ["inputs_embeds"]),
+        "layers": (["hidden_states", "attention_mask"], ["hidden_states"]),
+        "norm": (["hidden_states"], ["hidden_states"]),
+    }
+
+    def __init__(
+        self,
+        vocab_size=32000,
+        hidden_size=4096,
+        intermediate_size=14336,
+        num_hidden_layers=32,
+        num_attention_heads=32,
+        num_key_value_heads=8,
+        head_dim=None,
+        hidden_act="silu",
+        max_position_embeddings=4096 * 32,
+        initializer_range=0.02,
+        rms_norm_eps=1e-5,
+        use_cache=True,
+        pad_token_id=None,
+        bos_token_id=1,
+        eos_token_id=2,
+        tie_word_embeddings=False,
+        rope_theta=1e6,
+        sliding_window=None,
+        attention_dropout=0.0,
+        num_experts_per_tok=2,
+        num_local_experts=8,
+        output_router_logits=False,
+        router_aux_loss_coef=0.001,
+        router_jitter_noise=0.0,
+        **kwargs,
+    ):
+        self.vocab_size = vocab_size
+        self.max_position_embeddings = max_position_embeddings
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.sliding_window = sliding_window
+
+        # for backward compatibility
+        if num_key_value_heads is None:
+            num_key_value_heads = num_attention_heads
+
+        self.num_key_value_heads = num_key_value_heads
+        self.hidden_act = hidden_act
+        self.initializer_range = initializer_range
+        self.rms_norm_eps = rms_norm_eps
+        self.use_cache = use_cache
+        self.rope_theta = rope_theta
+        self.attention_dropout = attention_dropout
+        self.head_dim = head_dim
+
+        self.num_experts_per_tok = num_experts_per_tok
+        self.num_local_experts = num_local_experts
+        self.output_router_logits = output_router_logits
+        self.router_aux_loss_coef = router_aux_loss_coef
+        self.router_jitter_noise = router_jitter_noise
+
+        self.use_qk_norm = kwargs.pop("use_qk_norm", False)
+        self.rotary_dim = kwargs.pop("rotary_dim", self.head_dim)
+        self.partial_rotary_factor = kwargs.pop("partial_rotary_factor", 1)
+        if self.head_dim is not None:
+            self.partial_rotary_factor = self.rotary_dim / self.head_dim
+
+        super().__init__(
+            pad_token_id=pad_token_id,
+            bos_token_id=bos_token_id,
+            eos_token_id=eos_token_id,
+            tie_word_embeddings=tie_word_embeddings,
+            **kwargs,
+        )
+
+
+__all__ = ["MiniMaxM2Config"]

generation_config.json

@@ -0,0 +1,9 @@
+{
+  "bos_token_id": 200019,
+  "do_sample": true,
+  "eos_token_id": 200020,
+  "temperature": 1.0,
+  "top_p": 0.95,
+  "top_k": 40,
+  "transformers_version": "4.46.1"
+}

model-00000-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a6bdb3acc9a40f4048287376148f2653091c0320ff4b8676be493661bab51331
+size 6150780488

model-00001-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7419c4c4cfc49bc42ccea34b18e6e10868ab809566b23db27d00c9a90972f62e
+size 2415952592

model-00002-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9921692ddfbe03c5fd2a64cbddb3056278152b8b04f2027806be55565d0f0100
+size 4921586920

model-00003-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:15c20829dab95331eaca22d73bf982c27af5f6f4ee296487f693089f14c9381c
+size 2415952592

model-00004-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f1fbb4aab1709e6157757c429fe81af89044472e614011cb0126c0768fa07b9f
+size 4921586920

model-00005-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5226e66474a373786d302a0d5476fb22ec1dd06d646662b5489d91375193354
+size 2415952592

model-00006-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c19b00b01c504b5ed7c54734cd2674f5991b6ccede6e3dabf4c5e6470d0b26fc
+size 4921586920

model-00007-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c92796f9ceb818b0b99e2555745b601a4ea675911d3771d95378c707536e854a
+size 2415952592

model-00008-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:adff9f393f6cb79fa5f63137def18e47f45e8c7e8963ca484f4f6fe8a740aa3f
+size 4921586920

model-00009-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2f941c207dea0029b5a67a1ec9daddf1032604729dedfce18ea034d95860f92a
+size 2415952592

model-00010-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:82d20209f1a39f7144d4d5d9377c0184d2b8bbfa17f4811338283a38c2ea67dc
+size 4921586920

model-00011-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cdf9c912b0adf7ae3caab22d47fedcfd5f8a8c37d233de4c080b4b09985be742
+size 2415952592

model-00012-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:242514dff28a5ab8a593aba0607512b6a94e53f2ba6f7badfbef91ccd2e7a52d
+size 4921586920

model-00013-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7c4874abdcba687dd6d59d48199f19abc4ef871b28eaf89acc1e9bf2bbb4bdab
+size 2415952592

model-00014-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8d7bb0fa02b6ca6f34807f086a2fdac2bdc51cda7e9b47f51b4c4afd83370832
+size 4921586920

model-00015-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6bc35bc417b8800dff5e6df21c38e6c3a92d276be5b6bee5c8683489515d2032
+size 2415952592

model-00016-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:49d86930ef278d5fc6d253ff5201cb76401e2220ed26fca7a39b32c61c5a5506
+size 4921586920

model-00017-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ab165541d9f168197328fdc4bad7972349711c89ba329f119d51c233cd1c4f36
+size 2415952592

model-00018-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0d25162dd8a5208962373b69d4f2967fe2d72c464a74483f59494c7830c744df
+size 4921586920

model-00019-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:07bd87d28dd83a66b38cf8e607a09f0a2d1898e190fee637f0fed0a85ad76988
+size 2415952592

model-00020-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e399a3ff34e4223c23d7e5ab3447d34c456cfa46adf049a5951a260a5c328295
+size 4921587440

model-00021-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ef24d6b9c20d15edc626bbc346f46d4276444e92cfae0523f8ea0c8fb2b852a5
+size 2415952848

model-00022-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c01102007cbb381892cdb1ffddac26c56a7aa1dd2758377445e71f2f1f449dc
+size 4921587440

model-00023-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:185fea902110d1d4392b8ad29e94bf1f5bb3fbc973d0e3f283197801c64e8a3b
+size 2415952848

model-00024-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a803530f9b01189ba67449bd5c80f929e8d8411519b17f3be7cd0b2569d21fc5
+size 4921587440

model-00025-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:45e1f62799cfe403472a2ed9ab3d734aaa912e38badeb6376f84a74ad49b63f8
+size 2415952848

model-00026-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9d3ef7f65f08065bb6c58bc62a2d767286c502d42101c1dc5f5985ae3cd24865
+size 4921587440

model-00027-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:223fa0f2a0cff7a5d0011487d1d2addd104c890efcbf04621e8ba5e6b428c810
+size 2415952848

model-00028-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:51b8cad4f2f48dc1fa726c70b5293caf7dacb4dfc99e2e055e2f7487025efba2
+size 4921587440

model-00029-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:885fc83f7126cb72e078d4870c1af193577c8c4e45c2e56b54f4f9117fd5d792
+size 2415952848

model-00030-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:297adb4bda27dfbd4e6c64badf21cc5e85213abac09613b2f6f9aa2dd5f7a957
+size 4921587440

model-00031-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:232b50813d0fa718d01ef56a9be8888eed79e15e91b9bf978a16a930667b0159
+size 2415952848

model-00032-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:30f4ebd8e4b1df38e9a69fdcce8216f7cef502874f74242c0286ff75ea69c9ed
+size 4921587440

model-00033-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e9519844cb2a92a7c99790518d01436e074efcd7f878a7d05b40212df7e850c
+size 2415952848

model-00034-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d7d247b20113672e2cf658b28463b5160fa25718b7ff439a5ff1400aa1d66d6b
+size 4921587440

model-00035-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c72881de05143dbc7059548c566faada73311c1f6171d0a783472b062e0f9a14
+size 2415952848

model-00036-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ed64e1421ab8625efd597875c290f09fd15a125f8645acbeae06c4308a961f92
+size 4921587440

model-00037-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:71cb120d31a1a65985b708fa97f4d0b8b1a0b7316d883148f40efb028d7a597c
+size 2415952848

model-00038-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:56789a80ac8970477a229d6674061c453fcbbc7c3226a88346ee38605597d516
+size 4921587440

model-00039-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fadfd40a31aff82f7f89c1d56571da39abce28bdbcb1ba54832fec68033645e2
+size 2415952848

model-00040-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:061abc7e57ae12d7c5b0c20c859837bad230e24c6d2c6217f5d8f62d48a04f8e
+size 4921587440

model-00041-of-00130.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0a8f4e03827310b4a7ede4cfb48fe2d139f5d3370d0c1e4af8e1a48f6a1c9852
+size 2415952848