Upload folder using huggingface_hub

README.md

@@ -0,0 +1,46 @@
+  ---
+  license: other
+  license_name: minimax-community
+  license_link: https://huggingface.co/MiniMaxAI/MiniMax-M3/blob/main/LICENSE
+  base_model: MiniMaxAI/MiniMax-M3
+  tags:
+  - fp8
+  - modelopt
+  - quantized
+  ---
+
+  # MiniMax-M3-FP8 (per-tensor static, experts-quantized)
+
+  Built with MiniMax M3.
+
+  FP8 W8A8 quantization of [MiniMaxAI/MiniMax-M3](https://huggingface.co/MiniMaxAI/MiniMax-M3)
+  produced with NVIDIA TensorRT Model Optimizer (per-tensor static scales,
+  calibrated on a 512-sample proprietary domain-matched dataset). Scope:
+
+  - **Routed MoE experts: FP8** (weights + static input scales) — 97% of parameters.
+  - **Attention projections: BF16** — vLLM cannot load per-tensor scales on
+    M3's fused QKV, so attention is kept high-precision by design.
+  - **KV cache: no checkpoint-side scheme** — use `--kv-cache-dtype` as desired.
+
+  ## IMPORTANT: serving requires a vLLM fix
+
+  vLLM releases to date drop MiniMax-M3's swigluoai activation constants
+  (`swiglu_alpha=1.702`, `swiglu_beta=1.0`) in the ModelOpt FP8 MoE path, which
+  makes this checkpoint (and any FP8 M3 checkpoint) produce incoherent output
+  while loading without error. Fixes: vllm-project/vllm#49473 (ModelOpt path)
+  and vllm-project/vllm#46845 (compressed-tensors path). Serve only with a build
+  containing #49473 or with the equivalent patch applied.
+
+  vllm serve <this-repo> --tensor-parallel-size 8 --enable-expert-parallel
+    --trust-remote-code --language-model-only
+
+  ## Quality
+
+  On an internal SWE-bench-based agentic coding evaluation (300 instances,
+  identical harness and hardware), this checkpoint ties the official
+  `MiniMaxAI/MiniMax-M3-MXFP8` cumulative pass@1-3 (77 = 77) and resolves more
+  instances on the first attempt (pass@1 54 vs 51).
+
+  This model inherits the MiniMax Community License of the base model,
+  including its use restrictions and the commercial-use authorization
+  requirement for large entities.

added_tokens.json

@@ -0,0 +1,63 @@
+{
+  "]!p~[": 200000,
+  "<fim_prefix>": 200001,
+  "<fim_middle>": 200002,
+  "<fim_suffix>": 200003,
+  "<fim_pad>": 200004,
+  "<reponame>": 200005,
+  "<filename>": 200006,
+  "<gh_stars>": 200007,
+  "<issue_start>": 200008,
+  "<issue_comment>": 200009,
+  "<issue_closed>": 200010,
+  "<jupyter_start>": 200011,
+  "<jupyter_text>": 200012,
+  "<jupyter_code>": 200013,
+  "<jupyter_output>": 200014,
+  "<empty_output>": 200015,
+  "<commit_before>": 200016,
+  "<commit_msg>": 200017,
+  "<commit_after>": 200018,
+  "]~b]": 200019,
+  "[e~[": 200020,
+  "]!d~[": 200021,
+  "<function_call>": 200022,
+  "<code_interpreter>": 200023,
+  "]<]speech[>[": 200024,
+  "]<]image[>[": 200025,
+  "]<]video[>[": 200026,
+  "]<]start of speech[>[": 200027,
+  "]<]end of speech[>[": 200028,
+  "]<]start of image[>[": 200029,
+  "]<]end of image[>[": 200030,
+  "]<]start of video[>[": 200031,
+  "]<]end of video[>[": 200032,
+  "]<]vision pad[>[": 200033,
+  "]~!b[": 200034,
+  "<jupyter_error>": 200035,
+  "<add_file>": 200036,
+  "<delete_file>": 200037,
+  "<rename_file>": 200038,
+  "<edit_file>": 200039,
+  "<commit_message>": 200040,
+  "<empty_source_file>": 200041,
+  "<repo_struct>": 200042,
+  "<code_context>": 200043,
+  "<file_content>": 200044,
+  "<source_files>": 200045,
+  "<pr_start>": 200046,
+  "<review_comment>": 200047,
+  "<filepath>": 200048,
+  "<file_sep>": 200049,
+  "<think>": 200050,
+  "</think>": 200051,
+  "<tool_call>": 200052,
+  "</tool_call>": 200053,
+  "]<]frame[>[": 200054,
+  "]<]start of frame[>[": 200055,
+  "]<]end of frame[>[": 200056,
+  "<|content_altered_placeholder|>": 200057,
+  "]<]minimax[>[": 200058,
+  "<mm:think>": 200059,
+  "</mm:think>": 200060
+}

chat_template.jinja

@@ -0,0 +1,247 @@
+{# ---------- special token variables ---------- #}
+{%- set ns_token               = ']<]minimax[>['                  -%}
+{%- set bod_token              = ']~!b['                          -%}
+{%- set bos_token              = ']~b]'                           -%}
+{%- set eos_token              = '[e~['                           -%}
+{%- set toolcall_begin_token   = ns_token ~ '<tool_call>'         -%}
+{%- set toolcall_end_token     = ns_token ~ '</tool_call>'        -%}
+{%- set think_begin_token      = '<mm:think>'                     -%}
+{%- set think_end_token        = '</mm:think>'                    -%}
+{%- set image_token            = ']<]image[>['                    -%}
+{%- set video_token            = ']<]video[>['                    -%}
+{#- Thinking mode: "enabled" / "disabled" / "adaptive" / not defined -#}
+{#- Recursive XML renderer for tool_call arguments ======================== -#}
+{#- None values are intentionally skipped in mapping iteration so that
+    `<key>null</key>` (which would round-trip to the literal string "null")
+    never appears in the rendered tool_call. The convention is: omit the
+    field entirely. The top-level `_args` loop applies the same rule.
+    The `val is none` branch below is a safety net only — upstream cleaning
+    (drop_none_in_tool_arguments) should ensure no None ever reaches here. -#}
+{%- macro to_xml(val, ns) -%}
+{%- if val is mapping -%}
+{%- for k, v in val.items() if v is not none -%}
+{{ ns }}<{{ k }}>{{ to_xml(v, ns) }}{{ ns }}</{{ k }}>
+{%- endfor -%}
+{%- elif val is iterable and val is not string -%}
+{%- for item in val -%}
+{{ ns }}<item>{{ to_xml(item, ns) }}{{ ns }}</item>
+{%- endfor -%}
+{%- elif val is none -%}
+{#- Should be unreachable when upstream cleaning is applied. -#}
+{%- elif val is boolean -%}
+{{ val | tojson }}
+{%- else -%}
+{{ val }}
+{%- endif -%}
+{%- endmacro -%}
+{#- 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 mapping and item.type == 'image' -%}
+                {{- image_token }}
+            {%- elif item is mapping and item.type == 'video' -%}
+                {{- video_token}}
+            {%- elif item is string -%}
+                {{- item }}
+            {%- endif -%}
+        {%- endfor -%}
+    {%- elif content is none -%}
+        {{- '' }}
+    {%- 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 -%}
+        {{- 'Your model version is MiniMax-M3, developed by MiniMax. Knowledge cutoff: January 2026. Founded in early 2022, MiniMax is a global AI foundation model company committed to advancing the frontiers of AI towards AGI.' }}
+    {%- endif -%}
+
+    {#- Thinking mode instructions -#}
+    {{- '\n\n<thinking_instructions>\n' }}
+    {{- 'You have a thinking capability that allows you to reason step by step before responding. When thinking is enabled, wrap your reasoning in ' ~ think_begin_token ~ think_end_token ~ ' tags before your response. When thinking is disabled, begin your response directly after the ' ~ think_end_token ~ ' prefix. When thinking is adaptive, decide on your own whether to think for the current turn.\n' }}
+    {%- if thinking_mode is defined -%}
+        {%- if thinking_mode == "enabled" -%}
+            {{- 'Current thinking mode: enabled. You MUST think step by step before every response, including after receiving function/tool results.\n' }}
+        {%- elif thinking_mode == "disabled" -%}
+            {{- 'Current thinking mode: disabled. Do not output any thinking process.\n' }}
+        {%- elif thinking_mode == "adaptive" -%}
+            {{- 'Current thinking mode: adaptive. You are encouraged to think for complex decision-making, multi-step reasoning, or when analyzing function/tool results.\n' }}
+        {%- endif -%}
+    {%- else -%}
+        {{- 'Current thinking mode: adaptive. You are encouraged to think for complex decision-making, multi-step reasoning, or when analyzing function/tool results.\n' }}
+    {%- endif -%}
+    {{- '</thinking_instructions>' }}
+{%- endmacro -%}
+{%- macro build_developer_message(developer_message) -%}
+    {%- if developer_message and developer_message.content -%}
+        {{- visible_text(developer_message.content) }}
+    {%- else -%}
+        {%- if model_identity is not defined -%}
+            {%- set model_identity = "You are a helpful assistant." -%}
+        {%- endif -%}
+        {{- model_identity }}
+    {%- endif -%}
+{%- endmacro -%}
+{#- Main Template Logic ================================================= -#}
+{#- Role mapping: root -> system sp (high priority), system/developer -> developer sp (low priority) -#}
+{%- set system_message = none -%}
+{%- set developer_message = none -%}
+{%- set conversation_messages = messages -%}
+{%- if messages and messages[0].role == "root" -%}
+    {%- set system_message = messages[0] -%}
+    {%- set conversation_messages = messages[1:] -%}
+    {%- if conversation_messages and conversation_messages[0].role in ["system", "developer"] -%}
+        {%- set developer_message = conversation_messages[0] -%}
+        {%- set conversation_messages = conversation_messages[1:] -%}
+    {%- endif -%}
+{%- elif messages and messages[0].role in ["system", "developer"] -%}
+    {%- set developer_message = messages[0] -%}
+    {%- set conversation_messages = messages[1:] -%}
+{%- endif -%}
+{#- Render system sp (higher priority, root role only) -#}
+{{- bod_token ~ bos_token ~ 'system' ~ '\n' }}
+{{- build_system_message(system_message) }}
+{{- eos_token ~ '\n' }}
+
+{#- Render developer sp (lower priority: system/developer role + tools) -#}
+{{- bos_token ~ 'developer' ~ '\n' }}
+{{- build_developer_message(developer_message) }}
+{%- 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' }}
+    {{- 'To call tools, wrap all invocations in a single ' ~ toolcall_begin_token ~ toolcall_end_token ~ ' block. Parameter values containing nested objects or arrays are recursively expanded into XML elements. Example:\n' }}
+    {{- '\n' ~ toolcall_begin_token ~ '\n' }}
+    {{- ns_token + '<invoke name="tool-name-1">' }}
+    {{- ns_token + '<param-1>value-1' + ns_token + '</param-1>' }}
+    {{- ns_token + '<param-2>' }}
+    {{- ns_token + '<item>' }}
+    {{- ns_token + '<key-a>val-a' + ns_token + '</key-a>' }}
+    {{- ns_token + '<key-b>val-b' + ns_token + '</key-b>' }}
+    {{- ns_token + '</item>' }}
+    {{- ns_token + '</param-2>' }}
+    {{- ns_token + '</invoke>\n' }}
+    {{- ns_token + '<invoke name="tool-name-2">' }}
+    {{- ns_token + '<param-1>value-1' + ns_token + '</param-1>' }}
+    {{- ns_token + '</invoke>\n' }}
+    {{- toolcall_end_token }}
+{%- endif -%}
+{{- eos_token ~ '\n' }}
+
+{#- Render messages -#}
+{%- set last_tool_call = namespace(name=none) -%}
+{%- for message in conversation_messages -%}
+    {%- if message.role == 'assistant' -%}
+        {{- bos_token ~ '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_end_token in content %}
+                {%- set reasoning_content = content.split(think_end_token)[0].strip('\n').split(think_begin_token)[-1].strip('\n') %}
+                {%- set content = content.split(think_end_token)[-1].strip('\n') %}
+            {%- endif %}
+        {%- endif %}
+
+        {%- if reasoning_content -%}
+            {#- Render thinking for every assistant turn (all-turn visible) -#}
+            {{- think_begin_token ~ reasoning_content ~ think_end_token }}
+        {%- else -%}
+            {#- No thinking rendered → prefix with think_end_token -#}
+            {{- think_end_token }}
+        {%- endif -%}
+
+        {%- if content -%}
+            {{- content }}
+        {%- endif -%}
+        {%- if message.tool_calls -%}
+            {{- toolcall_begin_token ~ '\n' }}
+
+            {%- for tool_call in message.tool_calls -%}
+                {%- if tool_call.function -%}
+                    {%- set tool_call = tool_call.function -%}
+                {%- endif -%}
+{{- ns_token + '<invoke name="' + tool_call.name + '">' }}
+{%- set _args = tool_call.arguments -%}
+{%- for k, v in _args.items() if v is not none %}
+{{- ns_token + '<' + k + '>' -}}
+{{- to_xml(v, ns_token) -}}
+{{- ns_token + '</' + k + '>' }}
+{%- endfor -%}
+{{- ns_token + '</invoke>' ~ '\n' }}
+            {%- endfor -%}
+
+            {{- toolcall_end_token }}
+            {%- if message.tool_calls[-1].function -%}
+                {%- set last_tool_call.name = message.tool_calls[-1].function.name -%}
+            {%- else -%}
+                {%- set last_tool_call.name = message.tool_calls[-1].name -%}
+            {%- endif -%}
+        {%- else -%}
+            {%- set last_tool_call.name = none -%}
+        {%- endif -%}
+        {{- eos_token ~ '\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') -%}
+            {{- bos_token ~ 'tool' }}
+        {%- endif -%}
+        {{- '\n<response>' }}
+        {%- if message.content is string -%}
+            {{- message.content }}
+        {%- else -%}
+            {%- for tr in message.content -%}
+                {%- if tr is mapping and tr.type is defined and tr.type == 'image' -%}
+                    {{- image_token }}
+                {%- elif tr is mapping and tr.type is defined and tr.type == 'video' -%}
+                    {{- video_token }}
+                {%- else -%}
+                    {{- tr.output if tr.output is defined else (tr.text if tr.type == 'text' and tr.text is defined else tr) }}
+                {%- endif -%}
+            {%- endfor -%}
+        {%- endif -%}
+        {{- '</response>' }}
+        {%- if loop.last or (conversation_messages[loop.index0 + 1].role != 'tool') -%}
+            {{- eos_token ~ '\n' -}}
+        {%- endif -%}
+
+    {%- elif message.role == 'user' -%}
+        {{- bos_token ~ 'user' ~ '\n' }}
+        {{- visible_text(message.content) }}
+        {{- eos_token ~ '\n' }}
+    {%- endif -%}
+{%- endfor -%}
+
+{#- Generation prompt -#}
+{%- if add_generation_prompt -%}
+{{- bos_token ~ 'ai' ~ '\n' }}
+{%- if thinking_mode is defined and thinking_mode == "disabled" -%}
+    {{- think_end_token }}
+{%- elif thinking_mode is defined and thinking_mode == "adaptive" -%}
+    {#- adaptive: no prefix, let model decide -#}
+{%- elif thinking_mode is defined and thinking_mode == "enabled" -%}
+    {#- enabled or not defined: default to think -#}
+    {{- think_begin_token }}
+{%- else -%}
+    {#- adaptive: no prefix, let model decide -#}
+{%- endif -%}
+{%- endif -%}

config.json

@@ -0,0 +1,388 @@
+{
+  "architectures": [
+    "MiniMaxM3SparseForConditionalGeneration"
+  ],
+  "auto_map": {
+    "AutoConfig": "configuration_minimax_m3_vl.MiniMaxM3VLConfig"
+  },
+  "dtype": "bfloat16",
+  "image_grid_pinpoints": "[(336, 336), (336, 672), (336, 1008), (336, 1344), (336, 1680), (336, 2016), (672, 336), (672, 672), (672, 1008), (672, 1344), (672, 1680), (672, 2016), (1008, 336), (1008, 672), (1008, 1008), (1008, 1344), (1008, 1680), (1008, 2016), (1344, 336), (1344, 672), (1344, 1008), (1344, 1344), (1344, 1680), (1344, 2016), (1680, 336), (1680, 672), (1680, 1008), (1680, 1344), (1680, 1680), (1680, 2016), (2016, 336), (2016, 672), (2016, 1008), (2016, 1344), (2016, 1680), (2016, 2016)]",
+  "image_seq_length": 576,
+  "image_token_index": 200025,
+  "img_token_compression_config": {
+    "image_token_compression_method": "patch_merge",
+    "spatial_merge_size": 2,
+    "temporal_patch_size": 2
+  },
+  "merged_hidden_size": 24576,
+  "model_type": "minimax_m3_vl",
+  "multimodal_projector_bias": true,
+  "num_reward_heads": 0,
+  "process_image_mode": "dynamic_res",
+  "projector_hidden_act": "gelu",
+  "projector_hidden_size": 6144,
+  "text_config": {
+    "architectures": [
+      "MiniMaxM3SparseForCausalLM"
+    ],
+    "attention_dropout": 0.0,
+    "attention_output_gate": false,
+    "bos_token_id": 200034,
+    "dense_intermediate_size": 12288,
+    "dtype": "bfloat16",
+    "eos_token_id": 200020,
+    "head_dim": 128,
+    "hidden_act": "swigluoai",
+    "hidden_size": 6144,
+    "index_block_size": 128,
+    "index_head_dim": 128,
+    "index_local_blocks": 1,
+    "index_n_heads": 4,
+    "index_topk_blocks": 16,
+    "initializer_range": 0.02,
+    "intermediate_size": 3072,
+    "layer_types": [
+      "full_attention",
+      "full_attention",
+      "full_attention",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse",
+      "minimax_m3_sparse"
+    ],
+    "max_position_embeddings": 1048576,
+    "mlp_layer_types": [
+      "dense",
+      "dense",
+      "dense",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse",
+      "sparse"
+    ],
+    "model_type": "minimax_m3_vl_text",
+    "n_shared_experts": 1,
+    "num_attention_heads": 64,
+    "num_experts_per_tok": 4,
+    "num_hidden_layers": 60,
+    "num_key_value_heads": 4,
+    "num_local_experts": 128,
+    "num_mtp_modules": 7,
+    "num_nextn_predict_layers": 1,
+    "output_router_logits": false,
+    "pad_token_id": null,
+    "partial_rotary_factor": 0.5,
+    "qk_norm_type": "per_head",
+    "rms_norm_eps": 1e-06,
+    "rope_parameters": {
+      "partial_rotary_factor": 0.5,
+      "rope_theta": 5000000,
+      "rope_type": "default"
+    },
+    "rotary_dim": 64,
+    "routed_scaling_factor": 2.0,
+    "router_aux_loss_coef": 0.001,
+    "router_jitter_noise": 0.0,
+    "scoring_func": "sigmoid",
+    "shared_intermediate_size": 3072,
+    "swiglu_alpha": 1.702,
+    "swiglu_limit": 7.0,
+    "tie_word_embeddings": false,
+    "use_cache": true,
+    "use_gemma_norm": true,
+    "use_qk_norm": true,
+    "use_routing_bias": true,
+    "vocab_size": 200064
+  },
+  "tie_word_embeddings": false,
+  "transformers_version": "5.12.0",
+  "video_token_index": 200026,
+  "vision_config": {
+    "attention_dropout": 0.0,
+    "dtype": "bfloat16",
+    "hidden_act": "gelu",
+    "hidden_size": 1280,
+    "image_size": 2016,
+    "img_token_compression_config": {
+      "image_token_compression_method": "patch_merge",
+      "spatial_merge_size": 2,
+      "temporal_patch_size": 2
+    },
+    "initializer_factor": 1.0,
+    "initializer_range": 0.02,
+    "intermediate_size": 5120,
+    "layer_norm_eps": 1e-05,
+    "model_type": "minimax_m3_vl_vision",
+    "num_attention_heads": 16,
+    "num_channels": 3,
+    "num_hidden_layers": 32,
+    "patch_size": 14,
+    "position_embedding_type": "rope",
+    "projection_dim": 6144,
+    "rope_mode": "3d",
+    "rope_parameters": {
+      "rope_theta": 10000.0,
+      "rope_type": "default"
+    },
+    "spatial_merge_size": 2,
+    "temporal_patch_size": 2,
+    "vision_segment_max_frames": 4,
+    "vocab_size": 32000
+  },
+  "vision_feature_layer": -1,
+  "vision_feature_select_strategy": "full",
+  "quantization_config": {
+    "config_groups": {
+      "group_0": {
+        "input_activations": {
+          "dynamic": false,
+          "num_bits": 8,
+          "type": "float"
+        },
+        "weights": {
+          "dynamic": false,
+          "num_bits": 8,
+          "type": "float"
+        },
+        "targets": [
+          "Linear"
+        ]
+      }
+    },
+    "ignore": [
+      "language_model.lm_head",
+      "language_model.model.embed_tokens",
+      "language_model.model.layers.0.self_attn*",
+      "language_model.model.layers.1.self_attn*",
+      "language_model.model.layers.10.mlp.gate",
+      "language_model.model.layers.10.self_attn*",
+      "language_model.model.layers.11.mlp.gate",
+      "language_model.model.layers.11.self_attn*",
+      "language_model.model.layers.12.mlp.gate",
+      "language_model.model.layers.12.self_attn*",
+      "language_model.model.layers.13.mlp.gate",
+      "language_model.model.layers.13.self_attn*",
+      "language_model.model.layers.14.mlp.gate",
+      "language_model.model.layers.14.self_attn*",
+      "language_model.model.layers.15.mlp.gate",
+      "language_model.model.layers.15.self_attn*",
+      "language_model.model.layers.16.mlp.gate",
+      "language_model.model.layers.16.self_attn*",
+      "language_model.model.layers.17.mlp.gate",
+      "language_model.model.layers.17.self_attn*",
+      "language_model.model.layers.18.mlp.gate",
+      "language_model.model.layers.18.self_attn*",
+      "language_model.model.layers.19.mlp.gate",
+      "language_model.model.layers.19.self_attn*",
+      "language_model.model.layers.2.self_attn*",
+      "language_model.model.layers.20.mlp.gate",
+      "language_model.model.layers.20.self_attn*",
+      "language_model.model.layers.21.mlp.gate",
+      "language_model.model.layers.21.self_attn*",
+      "language_model.model.layers.22.mlp.gate",
+      "language_model.model.layers.22.self_attn*",
+      "language_model.model.layers.23.mlp.gate",
+      "language_model.model.layers.23.self_attn*",
+      "language_model.model.layers.24.mlp.gate",
+      "language_model.model.layers.24.self_attn*",
+      "language_model.model.layers.25.mlp.gate",
+      "language_model.model.layers.25.self_attn*",
+      "language_model.model.layers.26.mlp.gate",
+      "language_model.model.layers.26.self_attn*",
+      "language_model.model.layers.27.mlp.gate",
+      "language_model.model.layers.27.self_attn*",
+      "language_model.model.layers.28.mlp.gate",
+      "language_model.model.layers.28.self_attn*",
+      "language_model.model.layers.29.mlp.gate",
+      "language_model.model.layers.29.self_attn*",
+      "language_model.model.layers.3.mlp.gate",
+      "language_model.model.layers.3.self_attn*",
+      "language_model.model.layers.30.mlp.gate",
+      "language_model.model.layers.30.self_attn*",
+      "language_model.model.layers.31.mlp.gate",
+      "language_model.model.layers.31.self_attn*",
+      "language_model.model.layers.32.mlp.gate",
+      "language_model.model.layers.32.self_attn*",
+      "language_model.model.layers.33.mlp.gate",
+      "language_model.model.layers.33.self_attn*",
+      "language_model.model.layers.34.mlp.gate",
+      "language_model.model.layers.34.self_attn*",
+      "language_model.model.layers.35.mlp.gate",
+      "language_model.model.layers.35.self_attn*",
+      "language_model.model.layers.36.mlp.gate",
+      "language_model.model.layers.36.self_attn*",
+      "language_model.model.layers.37.mlp.gate",
+      "language_model.model.layers.37.self_attn*",
+      "language_model.model.layers.38.mlp.gate",
+      "language_model.model.layers.38.self_attn*",
+      "language_model.model.layers.39.mlp.gate",
+      "language_model.model.layers.39.self_attn*",
+      "language_model.model.layers.4.mlp.gate",
+      "language_model.model.layers.4.self_attn*",
+      "language_model.model.layers.40.mlp.gate",
+      "language_model.model.layers.40.self_attn*",
+      "language_model.model.layers.41.mlp.gate",
+      "language_model.model.layers.41.self_attn*",
+      "language_model.model.layers.42.mlp.gate",
+      "language_model.model.layers.42.self_attn*",
+      "language_model.model.layers.43.mlp.gate",
+      "language_model.model.layers.43.self_attn*",
+      "language_model.model.layers.44.mlp.gate",
+      "language_model.model.layers.44.self_attn*",
+      "language_model.model.layers.45.mlp.gate",
+      "language_model.model.layers.45.self_attn*",
+      "language_model.model.layers.46.mlp.gate",
+      "language_model.model.layers.46.self_attn*",
+      "language_model.model.layers.47.mlp.gate",
+      "language_model.model.layers.47.self_attn*",
+      "language_model.model.layers.48.mlp.gate",
+      "language_model.model.layers.48.self_attn*",
+      "language_model.model.layers.49.mlp.gate",
+      "language_model.model.layers.49.self_attn*",
+      "language_model.model.layers.5.mlp.gate",
+      "language_model.model.layers.5.self_attn*",
+      "language_model.model.layers.50.mlp.gate",
+      "language_model.model.layers.50.self_attn*",
+      "language_model.model.layers.51.mlp.gate",
+      "language_model.model.layers.51.self_attn*",
+      "language_model.model.layers.52.mlp.gate",
+      "language_model.model.layers.52.self_attn*",
+      "language_model.model.layers.53.mlp.gate",
+      "language_model.model.layers.53.self_attn*",
+      "language_model.model.layers.54.mlp.gate",
+      "language_model.model.layers.54.self_attn*",
+      "language_model.model.layers.55.mlp.gate",
+      "language_model.model.layers.55.self_attn*",
+      "language_model.model.layers.56.mlp.gate",
+      "language_model.model.layers.56.self_attn*",
+      "language_model.model.layers.57.mlp.gate",
+      "language_model.model.layers.57.self_attn*",
+      "language_model.model.layers.58.mlp.gate",
+      "language_model.model.layers.58.self_attn*",
+      "language_model.model.layers.59.mlp.gate",
+      "language_model.model.layers.59.self_attn*",
+      "language_model.model.layers.6.mlp.gate",
+      "language_model.model.layers.6.self_attn*",
+      "language_model.model.layers.7.mlp.gate",
+      "language_model.model.layers.7.self_attn*",
+      "language_model.model.layers.8.mlp.gate",
+      "language_model.model.layers.8.self_attn*",
+      "language_model.model.layers.9.mlp.gate",
+      "language_model.model.layers.9.self_attn*",
+      "model.multi_modal_projector*",
+      "model.vision_tower*"
+    ],
+    "quant_algo": "FP8",
+    "producer": {
+      "name": "modelopt",
+      "version": "0.46.0.dev174+gcba8a5c62"
+    },
+    "quant_method": "modelopt"
+  }
+}

configuration_minimax_m3_vl.py

@@ -0,0 +1,111 @@
+"""HuggingFace configs for the MiniMax VL family (M2 VL / M3 VL).
+
+This file is bundled into every converted HF checkpoint so that loading via
+``AutoConfig.from_pretrained(..., trust_remote_code=True)`` works without any
+runtime dependency on sglang or other internal packages — only stock
+``transformers`` is required.
+
+The class definitions intentionally mirror
+``sglang.srt.configs.minimax_vl``; if either side changes, keep them in sync.
+
+The file is named ``configuration_minimax_m3_vl.py`` (matching the legacy
+``model_type="minimax_m3_vl"`` and the converter's ``auto_map`` entry) so
+that ckpts produced by this converter remain loadable by older sglang versions
+that only know the ``MiniMaxM3VL*`` names. The canonical class is
+``MiniMaxM3VLConfig``; ``MiniMaxM3VLConfig`` is a thin BC alias whose only
+purpose is to be referenced from ``auto_map``.
+"""
+
+from typing import Optional
+
+from transformers.configuration_utils import PretrainedConfig
+from transformers.models.auto import CONFIG_MAPPING
+
+
+def _coerce_sub_config(
+    sub_config: Optional[dict], default_model_type: str
+) -> Optional[PretrainedConfig]:
+    """Convert a config dict to a ``PretrainedConfig`` instance.
+
+    If ``model_type`` is registered in HF ``CONFIG_MAPPING`` the corresponding
+    config class is used; otherwise we fall back to a generic
+    ``PretrainedConfig`` so all dict keys still become real attributes (M3's
+    text backbone uses ``model_type="minimax_m2"`` which is not in
+    ``CONFIG_MAPPING``).
+    """
+    if not isinstance(sub_config, dict):
+        return sub_config
+    model_type = sub_config.get("model_type", default_model_type)
+    cls = CONFIG_MAPPING.get(model_type, PretrainedConfig)
+    return cls(**sub_config)
+
+
+class MiniMaxVLBaseConfig(PretrainedConfig):
+    """Base config shared by every MiniMax VL variant.
+
+    Handles vision/text sub-config coercion. Concrete subclasses only need to
+    declare a unique ``model_type`` string.
+    """
+
+    def __init__(
+        self,
+        vision_config: Optional[dict] = None,
+        text_config: Optional[dict] = None,
+        image_token_index: int = 200025,
+        video_token_index: int = 200026,
+        image_seq_length: int = 576,
+        process_image_mode: str = "dynamic_res",
+        projector_hidden_act: str = "gelu",
+        multimodal_projector_bias: bool = True,
+        vision_feature_layer: int = -1,
+        vision_feature_select_strategy: str = "full",
+        img_token_compression_config: Optional[dict] = None,
+        image_grid_pinpoints: Optional[str] = None,
+        **kwargs,
+    ):
+        self.vision_config = _coerce_sub_config(vision_config, "clip_vision_model")
+        self.text_config = _coerce_sub_config(text_config, "mixtral")
+
+        self.image_token_index = image_token_index
+        self.video_token_index = video_token_index
+        self.image_seq_length = image_seq_length
+        self.process_image_mode = process_image_mode
+        self.projector_hidden_act = projector_hidden_act
+        self.multimodal_projector_bias = multimodal_projector_bias
+        self.vision_feature_layer = vision_feature_layer
+        self.vision_feature_select_strategy = vision_feature_select_strategy
+        self.img_token_compression_config = img_token_compression_config or {}
+        self.image_grid_pinpoints = image_grid_pinpoints
+
+        super().__init__(**kwargs)
+
+    def __post_init__(self, **kwargs):
+        super().__post_init__(**kwargs)
+        if hasattr(self, "vision_config"):
+            self.vision_config = _coerce_sub_config(self.vision_config, "clip_vision_model")
+        if hasattr(self, "text_config"):
+            self.text_config = _coerce_sub_config(self.text_config, "mixtral")
+
+
+class MiniMaxM2VLConfig(MiniMaxVLBaseConfig):
+    """MiniMax M2 VL: vision tower + M2 (Mixtral-style MoE) text backbone."""
+
+    model_type = "minimax_m2_vl"
+
+
+class MiniMaxM3VLConfig(MiniMaxVLBaseConfig):
+    """MiniMax M3 VL: vision tower + M3 (mixed sparse/dense MoE) text backbone."""
+
+    model_type = "minimax_m3_vl"
+
+
+class MiniMaxM2MiniVLConfig(MiniMaxM2VLConfig):
+    """Legacy alias kept so old ``model_type="minimax_m2_mini_vl"`` ckpts load."""
+
+    model_type = "minimax_m2_mini_vl"
+
+
+class MiniMaxM3VLConfig(MiniMaxM3VLConfig):
+    """Legacy alias kept so old ``model_type="minimax_m3_vl"`` ckpts load."""
+
+    model_type = "minimax_m3_vl"

generation_config.json

@@ -0,0 +1,8 @@
+{
+  "bos_token_id": 200019,
+  "do_sample": true,
+  "eos_token_id": 200020,
+  "temperature": 1.0,
+  "top_p": 0.95,
+  "transformers_version": "4.46.1"
+}

hf_quant_config.json

@@ -0,0 +1,132 @@
+{
+  "producer": {
+    "name": "modelopt",
+    "version": "0.46.0.dev174+gcba8a5c62"
+  },
+  "quantization": {
+    "quant_algo": "FP8",
+    "exclude_modules": [
+      "language_model.lm_head",
+      "language_model.model.embed_tokens",
+      "language_model.model.layers.0.self_attn*",
+      "language_model.model.layers.1.self_attn*",
+      "language_model.model.layers.10.mlp.gate",
+      "language_model.model.layers.10.self_attn*",
+      "language_model.model.layers.11.mlp.gate",
+      "language_model.model.layers.11.self_attn*",
+      "language_model.model.layers.12.mlp.gate",
+      "language_model.model.layers.12.self_attn*",
+      "language_model.model.layers.13.mlp.gate",
+      "language_model.model.layers.13.self_attn*",
+      "language_model.model.layers.14.mlp.gate",
+      "language_model.model.layers.14.self_attn*",
+      "language_model.model.layers.15.mlp.gate",
+      "language_model.model.layers.15.self_attn*",
+      "language_model.model.layers.16.mlp.gate",
+      "language_model.model.layers.16.self_attn*",
+      "language_model.model.layers.17.mlp.gate",
+      "language_model.model.layers.17.self_attn*",
+      "language_model.model.layers.18.mlp.gate",
+      "language_model.model.layers.18.self_attn*",
+      "language_model.model.layers.19.mlp.gate",
+      "language_model.model.layers.19.self_attn*",
+      "language_model.model.layers.2.self_attn*",
+      "language_model.model.layers.20.mlp.gate",
+      "language_model.model.layers.20.self_attn*",
+      "language_model.model.layers.21.mlp.gate",
+      "language_model.model.layers.21.self_attn*",
+      "language_model.model.layers.22.mlp.gate",
+      "language_model.model.layers.22.self_attn*",
+      "language_model.model.layers.23.mlp.gate",
+      "language_model.model.layers.23.self_attn*",
+      "language_model.model.layers.24.mlp.gate",
+      "language_model.model.layers.24.self_attn*",
+      "language_model.model.layers.25.mlp.gate",
+      "language_model.model.layers.25.self_attn*",
+      "language_model.model.layers.26.mlp.gate",
+      "language_model.model.layers.26.self_attn*",
+      "language_model.model.layers.27.mlp.gate",
+      "language_model.model.layers.27.self_attn*",
+      "language_model.model.layers.28.mlp.gate",
+      "language_model.model.layers.28.self_attn*",
+      "language_model.model.layers.29.mlp.gate",
+      "language_model.model.layers.29.self_attn*",
+      "language_model.model.layers.3.mlp.gate",
+      "language_model.model.layers.3.self_attn*",
+      "language_model.model.layers.30.mlp.gate",
+      "language_model.model.layers.30.self_attn*",
+      "language_model.model.layers.31.mlp.gate",
+      "language_model.model.layers.31.self_attn*",
+      "language_model.model.layers.32.mlp.gate",
+      "language_model.model.layers.32.self_attn*",
+      "language_model.model.layers.33.mlp.gate",
+      "language_model.model.layers.33.self_attn*",
+      "language_model.model.layers.34.mlp.gate",
+      "language_model.model.layers.34.self_attn*",
+      "language_model.model.layers.35.mlp.gate",
+      "language_model.model.layers.35.self_attn*",
+      "language_model.model.layers.36.mlp.gate",
+      "language_model.model.layers.36.self_attn*",
+      "language_model.model.layers.37.mlp.gate",
+      "language_model.model.layers.37.self_attn*",
+      "language_model.model.layers.38.mlp.gate",
+      "language_model.model.layers.38.self_attn*",
+      "language_model.model.layers.39.mlp.gate",
+      "language_model.model.layers.39.self_attn*",
+      "language_model.model.layers.4.mlp.gate",
+      "language_model.model.layers.4.self_attn*",
+      "language_model.model.layers.40.mlp.gate",
+      "language_model.model.layers.40.self_attn*",
+      "language_model.model.layers.41.mlp.gate",
+      "language_model.model.layers.41.self_attn*",
+      "language_model.model.layers.42.mlp.gate",
+      "language_model.model.layers.42.self_attn*",
+      "language_model.model.layers.43.mlp.gate",
+      "language_model.model.layers.43.self_attn*",
+      "language_model.model.layers.44.mlp.gate",
+      "language_model.model.layers.44.self_attn*",
+      "language_model.model.layers.45.mlp.gate",
+      "language_model.model.layers.45.self_attn*",
+      "language_model.model.layers.46.mlp.gate",
+      "language_model.model.layers.46.self_attn*",
+      "language_model.model.layers.47.mlp.gate",
+      "language_model.model.layers.47.self_attn*",
+      "language_model.model.layers.48.mlp.gate",
+      "language_model.model.layers.48.self_attn*",
+      "language_model.model.layers.49.mlp.gate",
+      "language_model.model.layers.49.self_attn*",
+      "language_model.model.layers.5.mlp.gate",
+      "language_model.model.layers.5.self_attn*",
+      "language_model.model.layers.50.mlp.gate",
+      "language_model.model.layers.50.self_attn*",
+      "language_model.model.layers.51.mlp.gate",
+      "language_model.model.layers.51.self_attn*",
+      "language_model.model.layers.52.mlp.gate",
+      "language_model.model.layers.52.self_attn*",
+      "language_model.model.layers.53.mlp.gate",
+      "language_model.model.layers.53.self_attn*",
+      "language_model.model.layers.54.mlp.gate",
+      "language_model.model.layers.54.self_attn*",
+      "language_model.model.layers.55.mlp.gate",
+      "language_model.model.layers.55.self_attn*",
+      "language_model.model.layers.56.mlp.gate",
+      "language_model.model.layers.56.self_attn*",
+      "language_model.model.layers.57.mlp.gate",
+      "language_model.model.layers.57.self_attn*",
+      "language_model.model.layers.58.mlp.gate",
+      "language_model.model.layers.58.self_attn*",
+      "language_model.model.layers.59.mlp.gate",
+      "language_model.model.layers.59.self_attn*",
+      "language_model.model.layers.6.mlp.gate",
+      "language_model.model.layers.6.self_attn*",
+      "language_model.model.layers.7.mlp.gate",
+      "language_model.model.layers.7.self_attn*",
+      "language_model.model.layers.8.mlp.gate",
+      "language_model.model.layers.8.self_attn*",
+      "language_model.model.layers.9.mlp.gate",
+      "language_model.model.layers.9.self_attn*",
+      "model.multi_modal_projector*",
+      "model.vision_tower*"
+    ]
+  }
+}

image_processor.py

@@ -0,0 +1,223 @@
+# Copyright 2023-2024 SGLang Team
+# Licensed under the Apache License, Version 2.0 (the "License");
+"""
+MiniMax VL family HuggingFace-compatible Processor, ImageProcessor, VideoProcessor.
+"""
+import math
+from typing import List, Tuple
+
+import torch
+from torchvision.transforms import InterpolationMode
+from transformers import BatchFeature
+from transformers.image_processing_utils_fast import (
+    BaseImageProcessorFast,
+    group_images_by_shape,
+    reorder_images,
+)
+from transformers.image_utils import PILImageResampling, SizeDict
+from transformers.processing_utils import (
+    ImagesKwargs,
+    Unpack,
+)
+from transformers.utils import TensorType
+
+MAX_RATIO = 200
+
+
+def round_by_factor(number: int, factor: int) -> int:
+    return round(number / factor) * factor
+
+
+def ceil_by_factor(number: int, factor: int) -> int:
+    return math.ceil(number / factor) * factor
+
+
+def floor_by_factor(number: int, factor: int) -> int:
+    return math.floor(number / factor) * factor
+
+
+def smart_resize(
+    height: int,
+    width: int,
+    factor: int = 28,
+    min_pixels: int = 4 * 28 * 28,
+    max_pixels: int = 451584,
+) -> tuple[int, int]:
+    if max(height, width) / min(height, width) > MAX_RATIO:
+        raise ValueError(
+            f"absolute aspect ratio must be smaller than {MAX_RATIO}, "
+            f"got {max(height, width) / min(height, width)}"
+        )
+    h_bar = max(factor, round_by_factor(height, factor))
+    w_bar = max(factor, round_by_factor(width, factor))
+    if h_bar * w_bar > max_pixels:
+        beta = math.sqrt((height * width) / max_pixels)
+        h_bar = floor_by_factor(height / beta, factor)
+        w_bar = floor_by_factor(width / beta, factor)
+    elif h_bar * w_bar < min_pixels:
+        beta = math.sqrt(min_pixels / (height * width))
+        h_bar = ceil_by_factor(height * beta, factor)
+        w_bar = ceil_by_factor(width * beta, factor)
+    return h_bar, w_bar
+
+
+# ==============================================================================
+# MiniMax M3 VL Image Processor Fast (Fast Mode - Torch based)
+# ==============================================================================
+
+
+class MiniMaxM3VLImageProcessorKwargs(ImagesKwargs, total=False):
+    patch_size: int
+    temporal_patch_size: int
+    merge_size: int
+    max_pixels: int
+
+
+class MiniMaxM3VLImageProcessor(BaseImageProcessorFast):
+    do_resize = True
+    resample = PILImageResampling.BICUBIC
+    size = {"height": 672, "width": 672}  # required by base class validation, not used as resize bound
+    default_to_square = False
+    do_rescale = True
+    rescale_factor = 1 / 255
+    do_normalize = True
+    image_mean = [0.48145466, 0.4578275, 0.40821073]
+    image_std = [0.26862954, 0.26130258, 0.27577711]
+    do_convert_rgb = True
+    patch_size = 14
+    temporal_patch_size = 2
+    merge_size = 2
+    max_pixels = 451584             # 672*672
+    valid_kwargs = MiniMaxM3VLImageProcessorKwargs
+    model_input_names = ["pixel_values", "image_grid_thw"]
+
+    def __init__(self, **kwargs: Unpack[MiniMaxM3VLImageProcessorKwargs]):
+        super().__init__(**kwargs)
+
+    def preprocess(
+        self, images, **kwargs: Unpack[MiniMaxM3VLImageProcessorKwargs]
+    ) -> BatchFeature:
+        return super().preprocess(images, **kwargs)
+
+    def _preprocess(
+        self,
+        images: List[torch.Tensor],
+        do_resize: bool,
+        size: SizeDict,
+        resample: PILImageResampling | InterpolationMode | int | None,
+        do_rescale: bool,
+        rescale_factor: float,
+        do_normalize: bool,
+        image_mean: float | List[float] | None,
+        image_std: float | List[float] | None,
+        patch_size: int,
+        temporal_patch_size: int,
+        merge_size: int,
+        max_pixels: int,
+        disable_grouping: bool | None,
+        return_tensors: str | TensorType | None,
+        **kwargs,
+    ) -> BatchFeature:
+        grouped_images, grouped_images_index = group_images_by_shape(
+            images, disable_grouping=disable_grouping
+        )
+        resized_images_grouped = {}
+        factor = patch_size * merge_size
+        for shape, stacked_images in grouped_images.items():
+            height, width = stacked_images.shape[-2:]
+            if do_resize:
+                resized_height, resized_width = smart_resize(
+                    height, width, factor=factor,
+                    max_pixels=max_pixels,
+                )
+                stacked_images = self.resize(
+                    stacked_images,
+                    size=SizeDict(height=resized_height, width=resized_width),
+                    resample=resample,
+                )
+            resized_images_grouped[shape] = stacked_images
+
+        resized_images = reorder_images(resized_images_grouped, grouped_images_index)
+
+        grouped_images, grouped_images_index = group_images_by_shape(
+            resized_images, disable_grouping=disable_grouping
+        )
+        processed_images_grouped = {}
+        processed_grids = {}
+
+        for shape, stacked_images in grouped_images.items():
+            resized_height, resized_width = stacked_images.shape[-2:]
+
+            patches = self.rescale_and_normalize(
+                stacked_images,
+                do_rescale,
+                rescale_factor,
+                do_normalize,
+                image_mean,
+                image_std,
+            )
+            if patches.ndim == 4:
+                patches = patches.unsqueeze(1)
+
+            if patches.shape[1] % temporal_patch_size != 0:
+                repeats = patches[:, -1:].repeat(
+                    1,
+                    temporal_patch_size - (patches.shape[1] % temporal_patch_size),
+                    1,
+                    1,
+                    1,
+                )
+                patches = torch.cat([patches, repeats], dim=1)
+
+            batch_size, grid_t, channel = patches.shape[:3]
+            grid_t = grid_t // temporal_patch_size
+            grid_h, grid_w = resized_height // patch_size, resized_width // patch_size
+
+            patches = patches.view(
+                batch_size,
+                grid_t,
+                temporal_patch_size,
+                channel,
+                grid_h // merge_size,
+                merge_size,
+                patch_size,
+                grid_w // merge_size,
+                merge_size,
+                patch_size,
+            )
+            patches = patches.permute(0, 1, 4, 7, 5, 8, 3, 2, 6, 9)
+
+            flatten_patches = patches.reshape(
+                batch_size,
+                grid_t * grid_h * grid_w,
+                channel * temporal_patch_size * patch_size * patch_size,
+            )
+
+            processed_images_grouped[shape] = flatten_patches
+            processed_grids[shape] = [[grid_t, grid_h, grid_w]] * batch_size
+
+        processed_images = reorder_images(
+            processed_images_grouped, grouped_images_index
+        )
+        processed_grids = reorder_images(processed_grids, grouped_images_index)
+
+        pixel_values = torch.cat(processed_images, dim=0)
+        image_grid_thw = torch.tensor(processed_grids, dtype=torch.long)
+
+        return BatchFeature(
+            data={"pixel_values": pixel_values, "image_grid_thw": image_grid_thw},
+            tensor_type=return_tensors,
+        )
+
+    def get_number_of_image_patches(self, height: int, width: int, images_kwargs=None):
+        images_kwargs = images_kwargs or {}
+        patch_size = images_kwargs.get("patch_size", self.patch_size)
+        merge_size = images_kwargs.get("merge_size", self.merge_size)
+        max_pixels = images_kwargs.get("max_pixels", self.max_pixels)
+
+        resized_height, resized_width = smart_resize(
+            height, width, factor=patch_size * merge_size,
+            max_pixels=max_pixels,
+        )
+        grid_h, grid_w = resized_height // patch_size, resized_width // patch_size
+        return grid_h * grid_w

model-00001-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b20290c34a59238131435ef83800f584356a8b30b1f93c8307ce1a364e2e1c78
+size 10415184664

model-00002-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:48c7dc62fe68c94675d1adaf5948eec1e22880578682857d0320d1084e668f88
+size 10205012344

model-00003-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c5ab88f5729ee3eb4dc98f40a670c87c5af2ecb6bd7be41eed1c0b4ef2643951
+size 10094886160

model-00004-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5038eab4c0e9f1457778f806f0dfc3cfc5bd15948a1811f44c70d5468016e36a
+size 10094886192

model-00005-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f075d509855db896a76ffb1ede3951472939fbeec7accf51fbb819f2c6fd3d52
+size 10205012312

model-00006-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:69ecb368c5b853781544d15160d10075b2b9101f951fbdb41318bca12d433264
+size 10094886760

model-00007-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:24ca910ccd2fca5c49d0ee66e284b558b952847b45722c34930599e59b6f1c7c
+size 10094887832

model-00008-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d2da5ffd7f74481c2592b0d5b26c0de91befc310c8a037d8ea6c0ba253de4911
+size 10205013840

model-00009-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ec25e595642dc0051c5804fb258409a42ad021b3f40750adcd673619fd1ffc8
+size 10094887776

model-00010-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c18dd7bca7fa80f17b83b1db45d52811aff8d25a667d38bcfa7183398e169f88
+size 10094887904

model-00011-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:789961d3166dbe9bd8c69020a4ed5790bfc4fe9f96589468d5aa445b62aae8c2
+size 10205013768

model-00012-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ecce98ac6a3933eebdcb7da393b4c3356091e2f7875193425abb5dc6d4a5d04d
+size 10094887776

model-00013-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4ce8424563417337fecd0f212317fe4feca3d44a393e1be45ce631b589e56760
+size 10094887968

model-00014-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4bcdaa1757e027b5ffc077ac0e7039a62efc6d7789f598003432cd81185bb403
+size 10205013704

model-00015-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0bf0c325be08b094638ecb98fbfbc72ea649d775fd43b05b56b21ca0c41f723e
+size 10094887776

model-00016-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:77a3b734f8dc5047f3ba2414dcfe4fa930ae3e59a220b94b066941e8ed9bddc2
+size 10164118344

model-00017-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9b4dbd50dd547734bdc91fc76ce1b9834a5ec54e34a4a6375b2ab3c598771097
+size 10154658128

model-00018-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4e5df344bc6baede46b2e018c233b55ef1d418199f65a7fbc71bfb4f95f64aaf
+size 10094887776

model-00019-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:63d5e8bb070d7c56e75964e3fce9884b306de9f8d6226adce83ad62e27b4526a
+size 10205013928

model-00020-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:438f0a99ccedf4d973e1a72d621dd44a221ea570800e5e802ca50fa73543aec5
+size 10094887744

model-00021-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9f0eecfb68573d0e400f8820f8e8c440f79ae0800144db8ffb39e785fe69c810
+size 10094887776

model-00022-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e4e59a523c8bec871231c14880375e0e3fa4903dcc06eee81df21238c941bc8a
+size 10205013896

model-00023-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f16ae386d7f9bdb025d2216a527ab933880be78aa1fa0799e8be7a3b6d72ec16
+size 10094887776

model-00024-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:88d505c4efae8f288d7fa19937b5b88d0de89786e207fba8ebb580adab5b811f
+size 10094887832

model-00025-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:171219bb4a5e5373ab4565ea8fd97d4eb0f4e92dc26cfc8fc7169fbc135c23b0
+size 10205013840

model-00026-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7dd159d77f31c03f15aaa2c8d27721cb807029c3a2b73d096a5c2fb0b5f4550f
+size 10094887776

model-00027-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:79b3afce85c45db87ffe8f123a40e01fd834b2a0404e8614857820fe8312f20b
+size 10094887904

model-00028-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:acac4a19b2c2d3cf068e610d6d6ba8cef1b58bb1884a4e0bbfab0553b0435d18
+size 10205013768

model-00029-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b74266f40377577be7dc160224279a91391338418fadc4d2d9559ed93e6d747
+size 10094887776

model-00030-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c016589869ce4feb20d6db9c8110bd5103e24ffca6780be8025031fc19a1a13a
+size 10094887976

model-00031-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:718cd2db8a3185dce85b429221045a7f22a8fa3afd184cd9ee642e672c135ae8
+size 10205013696

model-00032-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6e12c78a5a8b09b671599421244cf77c5eac36f6ce4420d42a8301e6ca865675
+size 10094887776

model-00033-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c301e626fb3d622dbe0d5d6b962304a2d13406251370e4e64d7a3be8b6822c42
+size 10151535144

model-00034-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:373118541fa8fa38b55e8fa8283fc6bc38ce5fbdc8aa1421938296ad2fd1eb52
+size 10148366536

model-00035-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d89262293509b41ffc09b40c4418c3441db20530a2f0f8279fee3bfe368d3905
+size 10094887776

model-00036-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1902479a07ba0ec1a9436e81e3077b16cb0e7b929d45980c08ea3271aebb1b7a
+size 10205013928

model-00037-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e6068eafca8be9303935484eeb97e425b37b574cb56abd4291f1bd9c99ad6e34
+size 10094887744

model-00038-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:08b44dc091e3a4b7f2d7e50c5160d9a0ce766219b109cecd371ec78315c09b53
+size 10094887776

model-00039-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2fbb6265db9663f251a94a1be090219f87a85d6c08b5aa99878d665aed7b86f6
+size 10205013896

model-00040-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:05d43a230c3a400286191713e0d1083cb8a8a5df55aa2ceb2cc6e075f5c6b95d
+size 10094887776

model-00041-of-00044.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3bf0760ddefe9ce8ed94d743736484358883227b5b2969f8db6e62afc32dc1fb
+size 10094887832