Add files using upload-large-folder tool

README.md

@@ -0,0 +1,100 @@
+---
+language:
+- en
+- zh
+license: apache-2.0
+base_model: stepfun-ai/Step-3.7-Flash-NVFP4
+pipeline_tag: image-text-to-text
+library_name: mlx
+tags:
+- mlx
+- jang
+- jang-2l
+- stepfun
+- vision-language
+---
+
+# Step-3.7-Flash-JANG_2L
+
+JANG_2L conversion of [stepfun-ai/Step-3.7-Flash-NVFP4](https://huggingface.co/stepfun-ai/Step-3.7-Flash-NVFP4).
+
+This bundle was built from the public NVFP4 checkpoint. Routed MoE tensors were decoded from ModelOpt NVFP4 (`uint8` payload, `float8_e4m3fn` block scales, fp32 side scales) and then re-quantized into JANG affine `weight/scales/biases` tensors. BF16 attention, shared expert, dense, vision, and projector tensors were handled according to the JANG plan.
+
+## Status
+
+This artifact has a text-only local coherence proof through the bundled `step3p7_mlx.py` bridge, which loads the nested Step3p5 text model using MLX and drops vision tensors for text generation.
+
+Verified locally:
+
+- 67 safetensors shards
+- 2,570 tensors in `model.safetensors.index.json`
+- No missing shard references
+- No raw NVFP4 `weight_scale`, `weight_scale_2`, or `input_scale` sidecars are present in the output index
+- `jang_config.json` capability verification passes
+- Text generation proof passes on a math prompt
+
+Text proof:
+
+```json
+{
+  "prompt": "What is 2+2? Answer with only the number.",
+  "output": "The user asks \"What is 2+2? Answer with only the number.\" So the answer is 4. The user wants only the number. So we should output \"4\". There's no disallowed content. It's a simple arithmetic. So we comply.\\n</think>\\n4",
+  "prompt_tokens": 26,
+  "generated_tokens": 58,
+  "prefill_s": 9.161997079849243,
+  "total_s": 15.426342725753784,
+  "decode_tok_s": 9.25874836391233
+}
+```
+
+Still required before full VLM runtime claims:
+
+- Step3p7 VLM wrapper in the target MLX/vMLX runtime
+- image patch token expansion and vision projector path
+
+## Format
+
+- Format: JANG affine
+- Profile: `JANG_2L`
+- Quantization backend: `mx.quantize`
+- Default group size: `128`
+- Bit widths used: `2`, `3`, `4`, `6`, `8`
+- Vision/projector: BF16 source converted to F16 passthrough for this first artifact
+- Output size: about `82G`
+- Runtime bridge: `step3p7_mlx.py` wraps `mlx_lm.models.step3p5` for text-only proof
+
+Important allocation choices:
+
+- `self_attn.{q,k,v,o,g}_proj`: 8-bit
+- `embed_tokens`: 6-bit
+- routed experts: `gate_proj=4`, `down_proj=3`, `up_proj=2`
+- true router/gate tensors: passthrough where present
+
+## Runtime Metadata
+
+`jang_config.json` stamps:
+
+```json
+{
+  "reasoning_parser": "qwen3",
+  "tool_parser": "step3p5",
+  "think_in_template": true,
+  "supports_tools": true,
+  "supports_thinking": true,
+  "family": "step3p7",
+  "modality": "vision",
+  "cache_type": "kv"
+}
+```
+
+The source chat template opens the assistant generation prompt inside `<think>`. Runtimes should not add a second synthetic reasoning prefix.
+
+## Vision And Audio
+
+The source checkpoint contains the Step vision encoder and `vit_large_projector`. No audio tensors or audio tokenizer files were present in the downloaded checkpoint.
+
+The source config mentions next-token prediction layers, but no MTP/nextn tensors were present in the NVFP4 source. This bundle does not synthesize MTP tensors from config fields.
+
+## Korean
+
+이 번들은 stepfun-ai/Step-3.7-Flash-NVFP4를 JANG_2L 형식으로 변환한 산출물입니다. 텍스트 경로는 `step3p7_mlx.py` 브리지를 통해 로컬 생성 검증을 통과했습니다. 비전 가중치는 포함되어 있지만, 이미지 입력 경로는 아직 별도 런타임 구현과 검증이 필요합니다. 오디오 텐서는 원본 체크포인트에 없었습니다.

chat_template.jinja

@@ -0,0 +1,89 @@
+{% macro render_message_content(message) %}{% if message.content is none %}{{- '' }}{% elif message.content is string %}{{- message.content }}{% elif message.content is mapping %}{{- message.content['value'] if 'value' in message.content else message.content['text'] }}{% elif message.content is iterable %}{% set ns = namespace(needs_text_separator=false) %}{% for item in message.content %}{% if item.type == 'text' %}{% if ns.needs_text_separator %}{{- ' ' }}{% endif %}{{- item['value'] if 'value' in item else item['text'] }}{% set ns.needs_text_separator = true %}{% elif item.type == 'image' %}<im_patch>{% set ns.needs_text_separator = false %}{% endif %}{% endfor %}{% endif %}{% endmacro %}
+{{bos_token}}{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if reasoning_effort is defined %}
+        {{- "Reasoning: " + reasoning_effort + '\n\n' }}
+    {%- endif %}
+    {%- if messages[0].role == 'system' %}
+        {{- render_message_content(messages[0]) + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou have access to the following functions in JSONSchema format:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson(ensure_ascii=False) }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...>\n...\n</function> block must be nested within <tool_call>\n...\n</tool_call> XML tags\n- Required parameters MUST be specified\n</IMPORTANT><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' }}
+        {%- if reasoning_effort is defined %}
+            {{- "Reasoning: " + reasoning_effort + '\n\n' }}
+        {%- endif %}
+        {{- render_message_content(messages[0]) + '<|im_end|>\n' }}
+    {%- elif reasoning_effort is defined %}
+        {{- '<|im_start|>system\n' + "Reasoning: " + reasoning_effort + '\n\n' + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and render_message_content(message) is string and not(render_message_content(message).startswith('<tool_response>') and render_message_content(message).endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- set content = render_message_content(message) %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {%- set role_name = 'observation' if (message.role == "system" and not loop.first and message.name == 'observation') else message.role %}
+        {{- '<|im_start|>' + role_name + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- else %}
+                {%- set reasoning_content = '' %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- if tool_call.arguments is defined %}
+                    {%- set arguments = tool_call.arguments | fromjson if tool_call.arguments is string else tool_call.arguments %}
+                    {%- for args_name, args_value in arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson(ensure_ascii=False) | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>tool_response\n' }}
+        {%- endif %}
+        {{- '<tool_response>' }}
+        {{- content }}
+        {{- '</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n<think>\n' }}
+{%- endif %}

config.json

@@ -0,0 +1,2650 @@
+{
+  "architectures": [
+    "Step3p7ForConditionalGeneration"
+  ],
+  "auto_map": {
+    "AutoConfig": "configuration_step3p7.Step3p7Config",
+    "AutoModelForCausalLM": "modeling_step3p7.Step3p7ForConditionalGeneration",
+    "AutoProcessor": "processing_step3.Step3VLProcessor"
+  },
+  "dtype": "bfloat16",
+  "hidden_size": 4096,
+  "im_end_token": "<im_end>",
+  "im_patch_token": "<im_patch>",
+  "im_start_token": "<im_start>",
+  "image_token_id": 128001,
+  "image_token_len": 169,
+  "max_position_embeddings": 262144,
+  "model_file": "step3p7_mlx.py",
+  "model_type": "step3p7",
+  "patch_token_len": 81,
+  "projector_bias": false,
+  "quantization": {
+    "bits": 2,
+    "format": "jang",
+    "group_size": 128,
+    "lm_head": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.embed_tokens": {
+      "bits": 6,
+      "group_size": 128
+    },
+    "model.layers.0.mlp.down_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.0.mlp.gate_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.0.mlp.up_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.0.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.0.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.0.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.0.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.0.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.1.mlp.down_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.1.mlp.gate_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.1.mlp.up_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.1.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.1.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.1.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.1.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.1.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.10.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.10.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.10.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.10.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.10.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.10.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.10.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.10.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.10.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.10.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.10.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.10.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.11.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.11.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.11.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.11.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.11.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.11.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.11.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.11.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.11.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.11.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.11.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.11.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.12.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.12.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.12.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.12.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.12.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.12.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.12.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.12.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.12.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.12.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.12.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.12.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.13.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.13.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.13.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.13.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.13.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.13.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.13.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.13.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.13.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.13.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.13.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.13.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.14.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.14.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.14.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.14.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.14.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.14.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.14.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.14.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.14.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.14.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.14.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.14.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.15.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.15.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.15.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.15.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.15.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.15.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.15.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.15.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.15.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.15.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.15.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.15.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.16.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.16.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.16.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.16.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.16.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.16.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.16.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.16.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.16.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.16.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.16.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.16.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.17.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.17.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.17.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.17.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.17.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.17.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.17.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.17.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.17.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.17.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.17.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.17.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.18.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.18.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.18.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.18.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.18.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.18.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.18.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.18.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.18.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.18.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.18.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.18.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.19.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.19.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.19.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.19.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.19.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.19.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.19.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.19.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.19.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.19.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.19.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.19.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.2.mlp.down_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.2.mlp.gate_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.2.mlp.up_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.2.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.2.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.2.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.2.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.2.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.20.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.20.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.20.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.20.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.20.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.20.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.20.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.20.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.20.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.20.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.20.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.20.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.21.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.21.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.21.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.21.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.21.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.21.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.21.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.21.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.21.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.21.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.21.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.21.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.22.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.22.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.22.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.22.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.22.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.22.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.22.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.22.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.22.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.22.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.22.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.22.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.23.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.23.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.23.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.23.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.23.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.23.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.23.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.23.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.23.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.23.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.23.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.23.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.24.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.24.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.24.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.24.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.24.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.24.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.24.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.24.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.24.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.24.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.24.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.24.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.25.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.25.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.25.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.25.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.25.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.25.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.25.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.25.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.25.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.25.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.25.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.25.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.26.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.26.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.26.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.26.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.26.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.26.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.26.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.26.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.26.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.26.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.26.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.26.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.27.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.27.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.27.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.27.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.27.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.27.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.27.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.27.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.27.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.27.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.27.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.27.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.28.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.28.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.28.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.28.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.28.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.28.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.28.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.28.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.28.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.28.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.28.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.28.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.29.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.29.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.29.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.29.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.29.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.29.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.29.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.29.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.29.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.29.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.29.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.29.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.3.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.3.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.3.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.3.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.3.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.3.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.3.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.3.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.3.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.3.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.3.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.3.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.30.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.30.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.30.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.30.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.30.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.30.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.30.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.30.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.30.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.30.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.30.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.30.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.31.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.31.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.31.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.31.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.31.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.31.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.31.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.31.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.31.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.31.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.31.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.31.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.32.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.32.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.32.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.32.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.32.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.32.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.32.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.32.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.32.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.32.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.32.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.32.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.33.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.33.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.33.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.33.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.33.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.33.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.33.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.33.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.33.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.33.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.33.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.33.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.34.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.34.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.34.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.34.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.34.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.34.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.34.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.34.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.34.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.34.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.34.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.34.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.35.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.35.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.35.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.35.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.35.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.35.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.35.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.35.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.35.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.35.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.35.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.35.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.36.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.36.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.36.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.36.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.36.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.36.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.36.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.36.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.36.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.36.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.36.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.36.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.37.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.37.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.37.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.37.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.37.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.37.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.37.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.37.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.37.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.37.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.37.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.37.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.38.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.38.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.38.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.38.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.38.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.38.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.38.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.38.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.38.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.38.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.38.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.38.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.39.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.39.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.39.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.39.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.39.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.39.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.39.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.39.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.39.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.39.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.39.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.39.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.4.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.4.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.4.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.4.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.4.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.4.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.4.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.4.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.4.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.4.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.4.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.4.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.40.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.40.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.40.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.40.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.40.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.40.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.40.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.40.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.40.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.40.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.40.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.40.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.41.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.41.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.41.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.41.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.41.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.41.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.41.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.41.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.41.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.41.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.41.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.41.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.42.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.42.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.42.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.42.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.42.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.42.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.42.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.42.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.42.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.42.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.42.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.42.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.43.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.43.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.43.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.43.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.43.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.43.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.43.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.43.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.43.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.43.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.43.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.43.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.44.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.44.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.44.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.44.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.44.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.44.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.44.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.44.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.44.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.44.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.44.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.44.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.5.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.5.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.5.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.5.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.5.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.5.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.5.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.5.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.5.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.5.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.5.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.5.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.6.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.6.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.6.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.6.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.6.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.6.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.6.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.6.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.6.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.6.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.6.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.6.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.7.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.7.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.7.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.7.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.7.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.7.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.7.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.7.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.7.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.7.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.7.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.7.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.8.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.8.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.8.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.8.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.8.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.8.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.8.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.8.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.8.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.8.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.8.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.8.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.9.mlp.gate.gate": {
+      "bits": 8,
+      "group_size": 64
+    },
+    "model.layers.9.mlp.share_expert.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.9.mlp.share_expert.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.9.mlp.share_expert.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.9.mlp.switch_mlp.down_proj": {
+      "bits": 3,
+      "group_size": 128
+    },
+    "model.layers.9.mlp.switch_mlp.gate_proj": {
+      "bits": 4,
+      "group_size": 128
+    },
+    "model.layers.9.mlp.switch_mlp.up_proj": {
+      "bits": 2,
+      "group_size": 128
+    },
+    "model.layers.9.self_attn.g_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.9.self_attn.k_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.9.self_attn.o_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.9.self_attn.q_proj": {
+      "bits": 8,
+      "group_size": 128
+    },
+    "model.layers.9.self_attn.v_proj": {
+      "bits": 8,
+      "group_size": 128
+    }
+  },
+  "quantization_config": {
+    "config_groups": {
+      "group_0": {
+        "input_activations": {
+          "dynamic": false,
+          "group_size": 16,
+          "num_bits": 4,
+          "type": "float"
+        },
+        "targets": [
+          "Linear"
+        ],
+        "weights": {
+          "dynamic": false,
+          "group_size": 16,
+          "num_bits": 4,
+          "type": "float"
+        }
+      }
+    },
+    "ignore": [
+      "lm_head",
+      "model.language_model.layers.0*",
+      "model.language_model.layers.1.*",
+      "model.language_model.layers.10.moe.gate",
+      "model.language_model.layers.10.self_attn*",
+      "model.language_model.layers.10.share_expert*",
+      "model.language_model.layers.11.moe.gate",
+      "model.language_model.layers.11.self_attn*",
+      "model.language_model.layers.11.share_expert*",
+      "model.language_model.layers.12.moe.gate",
+      "model.language_model.layers.12.self_attn*",
+      "model.language_model.layers.12.share_expert*",
+      "model.language_model.layers.13.moe.gate",
+      "model.language_model.layers.13.self_attn*",
+      "model.language_model.layers.13.share_expert*",
+      "model.language_model.layers.14.moe.gate",
+      "model.language_model.layers.14.self_attn*",
+      "model.language_model.layers.14.share_expert*",
+      "model.language_model.layers.15.moe.gate",
+      "model.language_model.layers.15.self_attn*",
+      "model.language_model.layers.15.share_expert*",
+      "model.language_model.layers.16.moe.gate",
+      "model.language_model.layers.16.self_attn*",
+      "model.language_model.layers.16.share_expert*",
+      "model.language_model.layers.17.moe.gate",
+      "model.language_model.layers.17.self_attn*",
+      "model.language_model.layers.17.share_expert*",
+      "model.language_model.layers.18.moe.gate",
+      "model.language_model.layers.18.self_attn*",
+      "model.language_model.layers.18.share_expert*",
+      "model.language_model.layers.19.moe.gate",
+      "model.language_model.layers.19.self_attn*",
+      "model.language_model.layers.19.share_expert*",
+      "model.language_model.layers.2.*",
+      "model.language_model.layers.20.moe.gate",
+      "model.language_model.layers.20.self_attn*",
+      "model.language_model.layers.20.share_expert*",
+      "model.language_model.layers.21.moe.gate",
+      "model.language_model.layers.21.self_attn*",
+      "model.language_model.layers.21.share_expert*",
+      "model.language_model.layers.22.moe.gate",
+      "model.language_model.layers.22.self_attn*",
+      "model.language_model.layers.22.share_expert*",
+      "model.language_model.layers.23.moe.gate",
+      "model.language_model.layers.23.self_attn*",
+      "model.language_model.layers.23.share_expert*",
+      "model.language_model.layers.24.moe.gate",
+      "model.language_model.layers.24.self_attn*",
+      "model.language_model.layers.24.share_expert*",
+      "model.language_model.layers.25.moe.gate",
+      "model.language_model.layers.25.self_attn*",
+      "model.language_model.layers.25.share_expert*",
+      "model.language_model.layers.26.moe.gate",
+      "model.language_model.layers.26.self_attn*",
+      "model.language_model.layers.26.share_expert*",
+      "model.language_model.layers.27.moe.gate",
+      "model.language_model.layers.27.self_attn*",
+      "model.language_model.layers.27.share_expert*",
+      "model.language_model.layers.28.moe.gate",
+      "model.language_model.layers.28.self_attn*",
+      "model.language_model.layers.28.share_expert*",
+      "model.language_model.layers.29.moe.gate",
+      "model.language_model.layers.29.self_attn*",
+      "model.language_model.layers.29.share_expert*",
+      "model.language_model.layers.3.moe.gate",
+      "model.language_model.layers.3.self_attn*",
+      "model.language_model.layers.3.share_expert*",
+      "model.language_model.layers.30.moe.gate",
+      "model.language_model.layers.30.self_attn*",
+      "model.language_model.layers.30.share_expert*",
+      "model.language_model.layers.31.moe.gate",
+      "model.language_model.layers.31.self_attn*",
+      "model.language_model.layers.31.share_expert*",
+      "model.language_model.layers.32.moe.gate",
+      "model.language_model.layers.32.self_attn*",
+      "model.language_model.layers.32.share_expert*",
+      "model.language_model.layers.33.moe.gate",
+      "model.language_model.layers.33.self_attn*",
+      "model.language_model.layers.33.share_expert*",
+      "model.language_model.layers.34.moe.gate",
+      "model.language_model.layers.34.self_attn*",
+      "model.language_model.layers.34.share_expert*",
+      "model.language_model.layers.35.moe.gate",
+      "model.language_model.layers.35.self_attn*",
+      "model.language_model.layers.35.share_expert*",
+      "model.language_model.layers.36.moe.gate",
+      "model.language_model.layers.36.self_attn*",
+      "model.language_model.layers.36.share_expert*",
+      "model.language_model.layers.37.moe.gate",
+      "model.language_model.layers.37.self_attn*",
+      "model.language_model.layers.37.share_expert*",
+      "model.language_model.layers.38.moe.gate",
+      "model.language_model.layers.38.self_attn*",
+      "model.language_model.layers.38.share_expert*",
+      "model.language_model.layers.39.moe.gate",
+      "model.language_model.layers.39.self_attn*",
+      "model.language_model.layers.39.share_expert*",
+      "model.language_model.layers.4.moe.gate",
+      "model.language_model.layers.4.self_attn*",
+      "model.language_model.layers.4.share_expert*",
+      "model.language_model.layers.40.moe.gate",
+      "model.language_model.layers.40.self_attn*",
+      "model.language_model.layers.40.share_expert*",
+      "model.language_model.layers.41.moe.gate",
+      "model.language_model.layers.41.self_attn*",
+      "model.language_model.layers.41.share_expert*",
+      "model.language_model.layers.42.moe.gate",
+      "model.language_model.layers.42.self_attn*",
+      "model.language_model.layers.42.share_expert*",
+      "model.language_model.layers.43.moe.gate",
+      "model.language_model.layers.43.self_attn*",
+      "model.language_model.layers.43.share_expert*",
+      "model.language_model.layers.44.moe.gate",
+      "model.language_model.layers.44.self_attn*",
+      "model.language_model.layers.44.share_expert*",
+      "model.language_model.layers.5.moe.gate",
+      "model.language_model.layers.5.self_attn*",
+      "model.language_model.layers.5.share_expert*",
+      "model.language_model.layers.6.moe.gate",
+      "model.language_model.layers.6.self_attn*",
+      "model.language_model.layers.6.share_expert*",
+      "model.language_model.layers.7.moe.gate",
+      "model.language_model.layers.7.self_attn*",
+      "model.language_model.layers.7.share_expert*",
+      "model.language_model.layers.8.moe.gate",
+      "model.language_model.layers.8.self_attn*",
+      "model.language_model.layers.8.share_expert*",
+      "model.language_model.layers.9.moe.gate",
+      "model.language_model.layers.9.self_attn*",
+      "model.language_model.layers.9.share_expert*",
+      "model.vision_model*",
+      "model.vit_large_projector"
+    ],
+    "kv_cache_scheme": {
+      "dynamic": false,
+      "num_bits": 8,
+      "type": "float"
+    },
+    "producer": {
+      "name": "modelopt",
+      "version": "0.45.0.dev37+g3ad4f4f09.d20260524"
+    },
+    "quant_algo": "NVFP4",
+    "quant_method": "modelopt"
+  },
+  "text_config": {
+    "architectures": [
+      "Step3p5ForCausalLM"
+    ],
+    "att_impl_type": "GQA",
+    "attention_dropout": 0.0,
+    "attention_other_setting": {
+      "attention_type": "sliding_attention",
+      "head_dim": 128,
+      "num_attention_groups": 8,
+      "num_attention_heads": 96,
+      "true_head_dim": 128
+    },
+    "bos_token_id": 0,
+    "dtype": "bfloat16",
+    "eos_token_id": [
+      1,
+      2,
+      128007
+    ],
+    "head_dim": 128,
+    "hidden_size": 4096,
+    "intermediate_size": 11264,
+    "layer_types": [
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention"
+    ],
+    "max_position_embeddings": 262144,
+    "max_seq_len": 262144,
+    "model_type": "step3p5",
+    "moe_every_n_layer": 1,
+    "moe_intermediate_size": 1280,
+    "moe_layer_offset": 0,
+    "moe_layers_enum": "3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44",
+    "moe_num_experts": 288,
+    "moe_router_activation": "sigmoid",
+    "moe_router_scaling_factor": 3.0,
+    "moe_top_k": 8,
+    "need_fp32_gate": true,
+    "norm_expert_weight": true,
+    "num_attention_groups": 8,
+    "num_attention_heads": 64,
+    "num_hidden_layers": 45,
+    "num_nextn_predict_layers": 3,
+    "partial_rotary_factor": 0.5,
+    "partial_rotary_factors": [
+      0.5,
+      1.0,
+      1.0,
+      1.0,
+      0.5,
+      1.0,
+      1.0,
+      1.0,
+      0.5,
+      1.0,
+      1.0,
+      1.0,
+      0.5,
+      1.0,
+      1.0,
+      1.0,
+      0.5,
+      1.0,
+      1.0,
+      1.0,
+      0.5,
+      1.0,
+      1.0,
+      1.0,
+      0.5,
+      1.0,
+      1.0,
+      1.0,
+      0.5,
+      1.0,
+      1.0,
+      1.0,
+      0.5,
+      1.0,
+      1.0,
+      1.0,
+      0.5,
+      1.0,
+      1.0,
+      1.0,
+      0.5,
+      1.0,
+      1.0,
+      1.0,
+      0.5
+    ],
+    "rms_norm_eps": 1e-05,
+    "rope_parameters": {
+      "factor": 2.0,
+      "high_freq_factor": 32.0,
+      "low_freq_factor": 1.0,
+      "original_max_position_embeddings": 131072,
+      "rope_theta": 500000.0,
+      "rope_type": "llama3"
+    },
+    "rope_scaling": {
+      "factor": 2.0,
+      "high_freq_factor": 32.0,
+      "low_freq_factor": 1.0,
+      "original_max_position_embeddings": 131072,
+      "rope_type": "llama3"
+    },
+    "rope_theta": [
+      5000000.0,
+      10000.0,
+      10000.0,
+      10000.0,
+      5000000.0,
+      10000.0,
+      10000.0,
+      10000.0,
+      5000000.0,
+      10000.0,
+      10000.0,
+      10000.0,
+      5000000.0,
+      10000.0,
+      10000.0,
+      10000.0,
+      5000000.0,
+      10000.0,
+      10000.0,
+      10000.0,
+      5000000.0,
+      10000.0,
+      10000.0,
+      10000.0,
+      5000000.0,
+      10000.0,
+      10000.0,
+      10000.0,
+      5000000.0,
+      10000.0,
+      10000.0,
+      10000.0,
+      5000000.0,
+      10000.0,
+      10000.0,
+      10000.0,
+      5000000.0,
+      10000.0,
+      10000.0,
+      10000.0,
+      5000000.0,
+      10000.0,
+      10000.0,
+      10000.0,
+      5000000.0,
+      10000.0,
+      10000.0,
+      10000.0
+    ],
+    "share_expert_dim": 1280,
+    "sink": false,
+    "sliding_window": 512,
+    "swiglu_limits": [
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      7,
+      7
+    ],
+    "swiglu_limits_shared": [
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      16,
+      16
+    ],
+    "torch_dtype": "bfloat16",
+    "use_cache": true,
+    "use_head_wise_attn_gate": true,
+    "use_mfa": false,
+    "use_moe": true,
+    "use_moe_router_bias": true,
+    "use_qk_norm": false,
+    "use_rope_layers": [],
+    "vocab_size": 128896,
+    "yarn_only_types": [
+      "full_attention"
+    ]
+  },
+  "transformers_version": "4.56.2",
+  "understand_projector_stride": 2,
+  "use_cache": true,
+  "use_im_start_end": "true",
+  "vision_config": {
+    "heads": 16,
+    "hidden_act": "quick_gelu",
+    "image_size": 728,
+    "layer_norm_eps": 1e-05,
+    "layers": 47,
+    "ls_init_value": 0.1,
+    "mlp_ratio": 5.833333333333333,
+    "model_type": "perception_encoder",
+    "num_channels": 3,
+    "output_dim": null,
+    "patch_size": 14,
+    "pool_type": "none",
+    "ues_cls_token": false,
+    "use_abs_posemb": true,
+    "use_cls_token": false,
+    "use_ln_post": false,
+    "use_ln_pre": true,
+    "use_rope2d": true,
+    "width": 1536
+  },
+  "vision_select_layer": -1
+}

configuration_step3p7.py

@@ -0,0 +1,207 @@
+from typing import Any, Optional, Sequence, Union
+
+from transformers.configuration_utils import PretrainedConfig
+
+class StepRoboticsVisionEncoderConfig(PretrainedConfig):
+    model_type = "perception_encoder"
+
+    def __init__(
+        self,
+        width=1536,
+        layers=47,
+        heads=16,
+        num_channels=3,
+        image_size=728,
+        mlp_ratio = 8960/1536,
+        patch_size=14,
+        hidden_act="quick_gelu",
+        layer_norm_eps=1e-5,
+        ues_cls_token=False,
+        use_cls_token: Optional[bool] = None,
+        use_ln_pre=True,
+        use_ln_post=False,
+        use_abs_posemb=True,
+        use_rope2d=True,
+        ls_init_value=0.1,
+        **kwargs,
+    ):
+        self.width = width
+        self.layers = layers
+        self.heads = heads
+        self.num_channels = num_channels
+        self.patch_size = patch_size
+        self.image_size = image_size
+        self.mlp_ratio = mlp_ratio
+        self.layer_norm_eps = layer_norm_eps
+        self.hidden_act = hidden_act
+        if use_cls_token is None:
+            use_cls_token = ues_cls_token
+        self.ues_cls_token = use_cls_token
+        self.use_cls_token = use_cls_token
+        self.use_ln_pre = use_ln_pre
+        self.ls_init_value = ls_init_value
+        self.use_ln_post = use_ln_post
+        self.use_abs_posemb = use_abs_posemb
+        self.use_rope2d = use_rope2d
+        super().__init__(**kwargs)
+
+
+class Step3p7TextConfig(PretrainedConfig):
+    model_type = "step3p5"
+    architectures = ["Step3p5ForCausalLM"]
+
+    def __init__(
+        self,
+        hidden_size: int = 4096,
+        intermediate_size: int = 11264,
+        num_attention_heads: int = 64,
+        num_attention_groups: int = 8,
+        num_hidden_layers: int = 45,
+        max_seq_len: int = 128000,
+        vocab_size: int = 128815,
+        rms_norm_eps: float = 1e-5,
+        moe_intermediate_size: int = 1280,
+        moe_num_experts: int = 288,
+        moe_top_k: int = 8,
+        rope_theta: float = 10000,
+        rope_scaling: Optional[dict[str, Any]] = None,
+        max_position_embeddings: int = 128000,
+        share_expert_dims: int = 1280,
+        share_expert_dim: Optional[int] = None,
+        head_dim: int = 128,
+        norm_expert_weight: bool = True,
+        layer_types: list[str] = None,
+        sliding_window: Optional[int] = None,
+        pad_token_id: int = 1,
+        attention_dropout: float = 0.0,
+        use_head_wise_attn_gate: bool = False,
+        use_moe_router_bias: bool = False,
+        moe_router_activation: str = "softmax",
+        moe_router_scaling_factor: float = 1.0,
+        need_fp32_gate: bool = False,
+        attention_other_setting: Optional[dict[str, Any]] = None,
+        swiglu_limits: Optional[list[Optional[float]]] = None,
+        swiglu_limits_shared: Optional[list[Optional[float]]] = None,
+        use_rope_layers: Optional[list[bool]] = None,
+        yarn_only_types: Optional[list[str]] = None,
+        moe_layers_enum: tuple[int] = (3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14,
+                                       15, 16, 17, 18, 19, 20, 21, 22, 23, 24,
+                                       25, 26, 27, 28, 29, 30, 31, 32, 33, 34,
+                                       35, 36, 37, 38, 39, 40, 41, 42, 43, 44),
+        **kwargs,
+    ) -> None:
+        torch_dtype = kwargs.get("torch_dtype")
+        trim_layer_types = _normalize_per_layer_values(layer_types,
+                                                  num_hidden_layers)
+        if isinstance(rope_scaling, dict):
+            rope_scaling = dict(rope_scaling)
+        if share_expert_dim is None:
+            share_expert_dim = share_expert_dims
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.num_attention_heads = num_attention_heads
+        self.num_attention_groups = num_attention_groups
+        self.num_hidden_layers = num_hidden_layers
+        self.max_seq_len = max_seq_len
+        self.vocab_size = vocab_size
+        self.rms_norm_eps = rms_norm_eps
+        self.moe_intermediate_size = moe_intermediate_size
+        self.moe_num_experts = moe_num_experts
+        self.moe_top_k = moe_top_k
+        self.rope_theta = rope_theta
+        self.rope_scaling = rope_scaling
+        self.max_position_embeddings = max_position_embeddings
+        self.share_expert_dim = share_expert_dim
+        self.head_dim = head_dim
+        self.norm_expert_weight = norm_expert_weight
+        self.moe_layers_enum = moe_layers_enum
+        self.layer_types = trim_layer_types
+        self.sliding_window = sliding_window
+        self.pad_token_id = pad_token_id
+        self.attention_dropout = attention_dropout
+        self.use_head_wise_attn_gate = use_head_wise_attn_gate
+        self.use_moe_router_bias = use_moe_router_bias
+        self.moe_router_activation = moe_router_activation
+        self.moe_router_scaling_factor = moe_router_scaling_factor
+        self.need_fp32_gate = need_fp32_gate
+        self.attention_other_setting = attention_other_setting
+        self.swiglu_limits = swiglu_limits
+        self.swiglu_limits_shared = swiglu_limits_shared
+        self.use_rope_layers = use_rope_layers
+        self.yarn_only_types = yarn_only_types
+        super().__init__(**kwargs)
+        if torch_dtype is not None:
+            self.torch_dtype = torch_dtype
+        self.layer_types = layer_types
+
+    def to_dict(self):
+        output = super().to_dict()
+        torch_dtype = getattr(self, "torch_dtype", None)
+        if torch_dtype is not None:
+            output["torch_dtype"] = torch_dtype
+        return output
+
+
+def _normalize_per_layer_values(
+    values: Optional[Sequence[Any]],
+    num_hidden_layers: int,
+) -> Optional[list[Any]]:
+    if values is None:
+        return None
+    normalized = list(values)
+    if not normalized:
+        return normalized
+    if len(normalized) < num_hidden_layers:
+        normalized.extend([normalized[-1]] *
+                          (num_hidden_layers - len(normalized)))
+    # Some checkpoints keep MTP/spec layer entries after the decoder layers.
+    # This config only builds num_hidden_layers decoder layers, and HF strict
+    # validation requires per-layer fields to match that decoder count.
+    return normalized[:num_hidden_layers]
+
+class Step3p7Config(PretrainedConfig):
+    # This loader is a compatibility shim for original Step VL checkpoints
+    # whose top-level config model_type is `step3p7`.
+    model_type = "step3p7"
+
+    def __init__(
+        self,
+        vision_config: Optional[Union[dict, StepRoboticsVisionEncoderConfig]] = None,
+        text_config: Optional[Union[dict, Step3p7TextConfig]] = None,
+        understand_projector_stride: int = 2,
+        projector_bias: bool = False,
+        image_token_id: int = 151679,
+        **kwargs,
+    ) -> None:
+        shared_rope_scaling = kwargs.get("rope_scaling")
+        if isinstance(shared_rope_scaling, dict):
+            shared_rope_scaling = dict(shared_rope_scaling)
+
+        if vision_config is None:
+            vision_config = StepRoboticsVisionEncoderConfig()
+        elif isinstance(vision_config, dict):
+            vision_config = StepRoboticsVisionEncoderConfig(**vision_config)
+        self.vision_config = vision_config
+
+        if text_config is None:
+            text_config = Step3p7TextConfig(rope_scaling=shared_rope_scaling)
+        elif isinstance(text_config, dict):
+            text_config = dict(text_config)
+            if shared_rope_scaling is not None and "rope_scaling" not in text_config:
+                text_config["rope_scaling"] = shared_rope_scaling
+            text_config = Step3p7TextConfig(**text_config)
+        elif shared_rope_scaling is not None and text_config.rope_scaling is None:
+            text_config.rope_scaling = dict(shared_rope_scaling)
+        self.text_config = text_config
+
+        rope_scaling = kwargs.get("rope_scaling")
+        if isinstance(rope_scaling, dict):
+            kwargs["rope_scaling"] = dict(rope_scaling)
+
+        self.understand_projector_stride = understand_projector_stride
+        self.projector_bias = projector_bias
+        self.hidden_size = text_config.hidden_size
+        self.max_position_embeddings = text_config.max_position_embeddings
+        self.image_token_id = image_token_id
+        # Help Auto classes find the correct implementation when saving/loading.
+        super().__init__(**kwargs)

generation_config.json

@@ -0,0 +1,10 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 0,
+  "eos_token_id": [
+    1,
+    2,
+    128007
+  ],
+  "transformers_version": "4.56.2"
+}

hf_quant_config.json

@@ -0,0 +1,145 @@
+{
+    "producer": {
+        "name": "modelopt",
+        "version": "0.45.0.dev37+g3ad4f4f09.d20260524"
+    },
+    "quantization": {
+        "quant_algo": "NVFP4",
+        "kv_cache_quant_algo": "FP8",
+        "group_size": 16,
+        "exclude_modules": [
+            "lm_head",
+            "model.language_model.layers.0*",
+            "model.language_model.layers.1.*",
+            "model.language_model.layers.10.moe.gate",
+            "model.language_model.layers.10.self_attn*",
+            "model.language_model.layers.10.share_expert*",
+            "model.language_model.layers.11.moe.gate",
+            "model.language_model.layers.11.self_attn*",
+            "model.language_model.layers.11.share_expert*",
+            "model.language_model.layers.12.moe.gate",
+            "model.language_model.layers.12.self_attn*",
+            "model.language_model.layers.12.share_expert*",
+            "model.language_model.layers.13.moe.gate",
+            "model.language_model.layers.13.self_attn*",
+            "model.language_model.layers.13.share_expert*",
+            "model.language_model.layers.14.moe.gate",
+            "model.language_model.layers.14.self_attn*",
+            "model.language_model.layers.14.share_expert*",
+            "model.language_model.layers.15.moe.gate",
+            "model.language_model.layers.15.self_attn*",
+            "model.language_model.layers.15.share_expert*",
+            "model.language_model.layers.16.moe.gate",
+            "model.language_model.layers.16.self_attn*",
+            "model.language_model.layers.16.share_expert*",
+            "model.language_model.layers.17.moe.gate",
+            "model.language_model.layers.17.self_attn*",
+            "model.language_model.layers.17.share_expert*",
+            "model.language_model.layers.18.moe.gate",
+            "model.language_model.layers.18.self_attn*",
+            "model.language_model.layers.18.share_expert*",
+            "model.language_model.layers.19.moe.gate",
+            "model.language_model.layers.19.self_attn*",
+            "model.language_model.layers.19.share_expert*",
+            "model.language_model.layers.2.*",
+            "model.language_model.layers.20.moe.gate",
+            "model.language_model.layers.20.self_attn*",
+            "model.language_model.layers.20.share_expert*",
+            "model.language_model.layers.21.moe.gate",
+            "model.language_model.layers.21.self_attn*",
+            "model.language_model.layers.21.share_expert*",
+            "model.language_model.layers.22.moe.gate",
+            "model.language_model.layers.22.self_attn*",
+            "model.language_model.layers.22.share_expert*",
+            "model.language_model.layers.23.moe.gate",
+            "model.language_model.layers.23.self_attn*",
+            "model.language_model.layers.23.share_expert*",
+            "model.language_model.layers.24.moe.gate",
+            "model.language_model.layers.24.self_attn*",
+            "model.language_model.layers.24.share_expert*",
+            "model.language_model.layers.25.moe.gate",
+            "model.language_model.layers.25.self_attn*",
+            "model.language_model.layers.25.share_expert*",
+            "model.language_model.layers.26.moe.gate",
+            "model.language_model.layers.26.self_attn*",
+            "model.language_model.layers.26.share_expert*",
+            "model.language_model.layers.27.moe.gate",
+            "model.language_model.layers.27.self_attn*",
+            "model.language_model.layers.27.share_expert*",
+            "model.language_model.layers.28.moe.gate",
+            "model.language_model.layers.28.self_attn*",
+            "model.language_model.layers.28.share_expert*",
+            "model.language_model.layers.29.moe.gate",
+            "model.language_model.layers.29.self_attn*",
+            "model.language_model.layers.29.share_expert*",
+            "model.language_model.layers.3.moe.gate",
+            "model.language_model.layers.3.self_attn*",
+            "model.language_model.layers.3.share_expert*",
+            "model.language_model.layers.30.moe.gate",
+            "model.language_model.layers.30.self_attn*",
+            "model.language_model.layers.30.share_expert*",
+            "model.language_model.layers.31.moe.gate",
+            "model.language_model.layers.31.self_attn*",
+            "model.language_model.layers.31.share_expert*",
+            "model.language_model.layers.32.moe.gate",
+            "model.language_model.layers.32.self_attn*",
+            "model.language_model.layers.32.share_expert*",
+            "model.language_model.layers.33.moe.gate",
+            "model.language_model.layers.33.self_attn*",
+            "model.language_model.layers.33.share_expert*",
+            "model.language_model.layers.34.moe.gate",
+            "model.language_model.layers.34.self_attn*",
+            "model.language_model.layers.34.share_expert*",
+            "model.language_model.layers.35.moe.gate",
+            "model.language_model.layers.35.self_attn*",
+            "model.language_model.layers.35.share_expert*",
+            "model.language_model.layers.36.moe.gate",
+            "model.language_model.layers.36.self_attn*",
+            "model.language_model.layers.36.share_expert*",
+            "model.language_model.layers.37.moe.gate",
+            "model.language_model.layers.37.self_attn*",
+            "model.language_model.layers.37.share_expert*",
+            "model.language_model.layers.38.moe.gate",
+            "model.language_model.layers.38.self_attn*",
+            "model.language_model.layers.38.share_expert*",
+            "model.language_model.layers.39.moe.gate",
+            "model.language_model.layers.39.self_attn*",
+            "model.language_model.layers.39.share_expert*",
+            "model.language_model.layers.4.moe.gate",
+            "model.language_model.layers.4.self_attn*",
+            "model.language_model.layers.4.share_expert*",
+            "model.language_model.layers.40.moe.gate",
+            "model.language_model.layers.40.self_attn*",
+            "model.language_model.layers.40.share_expert*",
+            "model.language_model.layers.41.moe.gate",
+            "model.language_model.layers.41.self_attn*",
+            "model.language_model.layers.41.share_expert*",
+            "model.language_model.layers.42.moe.gate",
+            "model.language_model.layers.42.self_attn*",
+            "model.language_model.layers.42.share_expert*",
+            "model.language_model.layers.43.moe.gate",
+            "model.language_model.layers.43.self_attn*",
+            "model.language_model.layers.43.share_expert*",
+            "model.language_model.layers.44.moe.gate",
+            "model.language_model.layers.44.self_attn*",
+            "model.language_model.layers.44.share_expert*",
+            "model.language_model.layers.5.moe.gate",
+            "model.language_model.layers.5.self_attn*",
+            "model.language_model.layers.5.share_expert*",
+            "model.language_model.layers.6.moe.gate",
+            "model.language_model.layers.6.self_attn*",
+            "model.language_model.layers.6.share_expert*",
+            "model.language_model.layers.7.moe.gate",
+            "model.language_model.layers.7.self_attn*",
+            "model.language_model.layers.7.share_expert*",
+            "model.language_model.layers.8.moe.gate",
+            "model.language_model.layers.8.self_attn*",
+            "model.language_model.layers.8.share_expert*",
+            "model.language_model.layers.9.moe.gate",
+            "model.language_model.layers.9.self_attn*",
+            "model.language_model.layers.9.share_expert*",
+            "model.vision_model*",
+            "model.vit_large_projector"
+        ]
+    }
+}

jang_config.json

@@ -0,0 +1,70 @@
+{
+  "architecture": {
+    "has_audio": false,
+    "has_mtp_tensors": false,
+    "has_vision": true,
+    "text_model_type": "step3p5",
+    "type": "step3p7"
+  },
+  "capabilities": {
+    "cache_type": "kv",
+    "family": "step3p7",
+    "modality": "vision",
+    "reasoning_parser": "qwen3",
+    "supports_thinking": true,
+    "supports_tools": true,
+    "think_in_template": true,
+    "tool_parser": "step3p5"
+  },
+  "format": "jang",
+  "format_version": "2.0",
+  "mxtq_bits": {
+    "attention": 8,
+    "embedding": 6,
+    "routed_expert": {
+      "down_proj": 3,
+      "gate_proj": 4,
+      "up_proj": 2
+    }
+  },
+  "quantization": {
+    "actual_bits": 3.398,
+    "bit_widths_used": [
+      2,
+      3,
+      4,
+      6,
+      8
+    ],
+    "block_size": 128,
+    "method": "jang-importance",
+    "passthrough_bit_widths_used": [
+      16,
+      32
+    ],
+    "profile": "JANG_2L",
+    "quantization_backend": "mx.quantize",
+    "source_weight_decode": "modelopt-nvfp4-for-routed-moe",
+    "target_bits": 2.0,
+    "total_quantized_bits": 669270114304,
+    "total_source_bits": 3151291744256
+  },
+  "runtime": {
+    "requires": [
+      "step3p7-vlm-wrapper",
+      "step3p5-text-runtime",
+      "full-and-sliding-kv-cache",
+      "head-wise-attention-gate",
+      "qk-rmsnorm",
+      "kv-scale-sidecars",
+      "image-patch-processing"
+    ],
+    "shard_count": 67,
+    "total_shard_bytes": 87621944984
+  },
+  "source_model": {
+    "dtype": "nvfp4+bf16",
+    "hub_id": "stepfun-ai/Step-3.7-Flash-NVFP4",
+    "name": "Step-3.7-Flash-NVFP4"
+  }
+}

model-00001-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5b88593183efb5896bdeef84490dfa878117932160cf3230c0c5789dbaa3d46e
+size 1144607880

model-00004-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8573cbf5d160da516142285db152e3d1876943ca35399d228c663d5bf8d0c62e
+size 1585825616

model-00005-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:733c4efe2bfc6cb51ec149e6d99dbedbf1d1a3c53a465fad65dea67ea2f9879a
+size 1339963408

model-00012-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3885bb3a2200821fe5b15c6c95a00ed599fdf79966334fcf371061dca436d31e
+size 1150967584

model-00013-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:842aee57a603677db6b838c577391d6d24be5929de07aa63d963bb506efa16a3
+size 1423239760

model-00016-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fcd7c45a1b9103dd4cb51c4fdfe9910d550513900f2af18ad8935e79a87a1480
+size 1165028216

model-00017-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2559644091cf2dd2f94d46ea46dfecaa56b8fdf800681de3465e9f7e56bb7371
+size 1346111856

model-00018-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:48ffc0acf2acee48ee0036f60940666c396c1401a9f3fdb597d25623b70ff0fb
+size 1416833416

model-00019-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:380ea0e6e62ef7e292a726f62ca91646036632505042c91ba4178faee5bc3389
+size 1123882448

model-00020-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:968d152f31d258863bc0111602286aede7046a48d0c16eec49eab23399af6625
+size 1347358536

model-00021-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d3a87a2791d54afc254b438f3f536ac37d26554ea717f517e6851edaf1f0ae35
+size 1150967592

model-00024-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:939d77dc9873753c4007400d2f5303239ab3d7ee4cfef4ce5ed3c6d45d14b374
+size 1416833416

model-00025-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c22ad29194eadf79192145c80c0b8adca58adb17156f874f46b4c30793ba5b52
+size 1123874136

model-00027-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b9b79dc8819dc1448b7940e860bbfdc1d495fa3c4da8efc86b8a988e25975235
+size 1416833416

model-00032-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cf3f72cd66b241ce7e5eed200194b804ffddadb0bdbafba08fd2a2a2b16c7938
+size 1421986048

model-00033-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bd934ec3a9d3f24d0ac9f867987cb2a732e215df256c3545e5eb08c0b498d687
+size 1416833416

model-00036-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9fe431a441ff5218f4be3cb2fc4d6649cf540b573a40c6413b206ea1a0f92c61
+size 1416833416

model-00037-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:daa0ef4517f7fa47973a81e556bf947431112675232d5be6a71df966cae08b91
+size 1123890768

model-00038-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e48e303487e60a94e39ac6bd9507a46984a107689c6518c3443b33e129be767
+size 1346111856

model-00039-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4ddafeca14d6b660665f05c93d27e2422917e220d3e68f61b113a0c02ede3316
+size 1416833416

model-00040-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:508457293fdc5d853581fa2595dd425ce225bcf6da13a1024142aede3b874c66
+size 1158620632

model-00041-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4509e0ec408fe787cfcf26e957c59d485aeb800e5cabe6981dd39e034381cc8f
+size 1347358536

model-00044-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9461d83672ed03051de5fe05e6b63ec52e91147b2de62cb08e5852340f59745d
+size 1346111856

model-00045-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d6e2bf216b3dcefe2c8283ce0f6cb86c2323c2af6f0ee11f2e0f60a6c457d521
+size 1416833416

model-00052-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:88905dd2b383bc57a55f80cb4169f7e93edcf8459b12ef6e50ef3b50880ac600
+size 1271490120

model-00053-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5dc0f262e77977417682aeafdf5a36176da698819f65a3d5c5bdddca215b1287
+size 1233242360

model-00056-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1eacde34bf9e5524d3577991d2c26527a2af17b7afffcead26735ec3c1c55498
+size 1347357280

model-00057-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a2f2fd04b8d407595f9ad4bb868f7e6ee438dbf56d466da92fad594058abfc26
+size 1339712544

model-00058-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dcefc98844b20f9cee85fb362645799ef8de0e70d03cf1cef8666ecdd6981efe
+size 1234496104

model-00059-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e48527de7312987a4195201d741c578073ee75ef046a751a4e121e15f089354a
+size 1346111856

model-00060-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c5f05a4f37f9726b52c3e50e599fecf5e59b4a8f1f2fd4a8f955ec49bc80e43e
+size 1416833416

model-00061-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c17b661d43814171df00a41f1a619566a00d89e13e59e36d2baf2814f6953d7d
+size 1123882448

model-00064-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:38de06338a12ac7cba633527c89ad1d686e7a31628b5bc5ee317db61e8687277
+size 1233242360

model-00065-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af2afdb70d9c9d9c4562405b0a16cdd8ce95fa5998925777dffa1276a7b7c0b2
+size 1416833416

model-00066-of-00067.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:09c29930d31f204a6e33a2de4f97727e4b396578bba40ba9023c13a0d25d5ef6
+size 1158620632

modeling_step3p7.py

@@ -0,0 +1,1405 @@
+# Copyright 2025 The LLAMA4 and HuggingFace Inc. 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.
+import copy
+import inspect
+from dataclasses import dataclass
+from typing import Callable, Literal, Optional, Tuple, TypedDict, Union
+
+from PIL import Image
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from transformers.activations import ACT2FN
+from transformers.cache_utils import Cache, DynamicCache
+from transformers.generation import GenerationMixin
+from transformers.masking_utils import (
+    create_causal_mask,
+    create_sliding_window_causal_mask,
+)
+from transformers.modeling_flash_attention_utils import FlashAttentionKwargs
+from transformers.modeling_layers import GradientCheckpointingLayer
+from transformers.modeling_outputs import BaseModelOutputWithPast, ModelOutput
+from transformers.modeling_rope_utils import ROPE_INIT_FUNCTIONS, dynamic_rope_update
+from transformers.modeling_utils import ALL_ATTENTION_FUNCTIONS, PreTrainedModel
+from transformers.processing_utils import Unpack
+from transformers.utils import TransformersKwargs, can_return_tuple, logging
+from .configuration_step3p7 import Step3p7Config, Step3p7TextConfig
+from .vision_encoder import StepRoboticsVisionEncoder
+
+
+logger = logging.get_logger(__name__)
+_MASK_INPUT_EMBEDS_ARG = (
+    "inputs_embeds"
+    if "inputs_embeds" in inspect.signature(create_causal_mask).parameters
+    else "input_embeds"
+)
+
+__all__ = [
+    "Step3p7Model",
+]
+
+
+class StepVLImagePixelInputs(TypedDict):
+    type: Literal["pixel_values"]
+    pixel_values: torch.Tensor
+    patch_pixel_values: Optional[torch.Tensor]
+    num_patches: list[int]
+
+
+class StepVLImageEmbeddingInputs(TypedDict):
+    type: Literal["image_embeds"]
+    image_embeds: torch.Tensor
+
+
+StepVLImageInputs = Union[StepVLImagePixelInputs, StepVLImageEmbeddingInputs]
+
+
+def _flatten_embeddings(embeddings) -> torch.Tensor:
+    """
+    Recursively flattens and concatenates NestedTensors on all but the last
+    dimension.
+    """
+
+    if isinstance(embeddings, torch.Tensor):
+        # Flatten all but the last dimension.
+        return embeddings.flatten(0, -2)
+
+    return torch.cat(tuple(_flatten_embeddings(t) for t in embeddings))
+
+def _embedding_count_expression(embeddings) -> str:
+    """
+    Constructs a debugging representation of the number of embeddings in the
+    NestedTensors.
+    """
+
+    if isinstance(embeddings, torch.Tensor):
+        return " x ".join([str(dim) for dim in embeddings.shape[:-1]])
+
+    return " + ".join(_embedding_count_expression(inner) for inner in embeddings)
+
+
+def _merge_multimodal_embeddings(
+    inputs_embeds: torch.Tensor,
+    is_multimodal: torch.Tensor,
+    multimodal_embeddings,
+) -> torch.Tensor:
+    """
+    Merge ``multimodal_embeddings`` into ``inputs_embeds`` by overwriting the
+    positions in ``inputs_embeds`` corresponding to placeholder tokens in
+    ``input_ids``.
+    Note:
+        This updates ``inputs_embeds`` in place.
+    """
+    num_expected_tokens = is_multimodal.sum().item()
+    assert isinstance(num_expected_tokens, int)
+
+    flattened = _flatten_embeddings(multimodal_embeddings)
+    if flattened.shape[0] != num_expected_tokens:
+        expr = _embedding_count_expression(multimodal_embeddings)
+        raise ValueError(
+            f"Attempted to assign {expr} = {flattened.shape[0]} "
+            f"multimodal tokens to {num_expected_tokens} placeholders"
+        )
+
+    is_multimodal = is_multimodal.to(inputs_embeds.device)
+    flattened = flattened.to(inputs_embeds.device)
+    inputs_embeds[is_multimodal] = flattened
+    return inputs_embeds
+
+def merge_multimodal_embeddings(
+    input_ids: torch.Tensor,
+    inputs_embeds: torch.Tensor,
+    multimodal_embeddings,
+    placeholder_token_id: Union[int, list[int]],
+) -> torch.Tensor:
+    """
+    Merge ``multimodal_embeddings`` into ``inputs_embeds`` by overwriting the
+    positions in ``inputs_embeds`` corresponding to placeholder tokens in
+    ``input_ids``.
+    
+    ``placeholder_token_id`` can be a list of token ids (e.g, token ids 
+    of img_start, img_break, and img_end tokens) when needed: This means 
+    the order of these tokens in the ``input_ids`` MUST MATCH the order of 
+    their embeddings in ``multimodal_embeddings`` since we need to 
+    slice-merge instead of individually scattering.
+    For example, if input_ids is "TTTTTSIIIBIIIBIIIETTT", where
+    - T is text token
+    - S is image start token
+    - I is image embedding token
+    - B is image break token
+    - E is image end token.
+    
+    Then the image embeddings (that correspond to I's) from vision encoder 
+    must be padded with embeddings of S, B, and E in the same order of 
+    input_ids for a correct embedding merge.
+    Note:
+        This updates ``inputs_embeds`` in place.
+    """
+    if isinstance(placeholder_token_id, list):
+        placeholder_token_id = torch.tensor(
+            placeholder_token_id, device=input_ids.device
+        )
+        return _merge_multimodal_embeddings(
+            inputs_embeds,
+            torch.isin(input_ids, placeholder_token_id),
+            multimodal_embeddings,
+        )
+
+    return _merge_multimodal_embeddings(
+        inputs_embeds,
+        (input_ids == placeholder_token_id),
+        multimodal_embeddings,
+    )
+
+
+class Step3p7PreTrainedModel(PreTrainedModel):
+    # Link this model family to its configuration class so PreTrainedModel.from_pretrained
+    # can load the config instead of failing with a NoneType error.
+    config_class = Step3p7Config
+    supports_gradient_checkpointing = True
+    _skip_keys_device_placement = ["past_key_values"]
+    _keys_to_ignore_on_load_unexpected = [
+        r"model\.layers\.45\.*",
+        r"model\.layers\.46\.*",
+        r"model\.layers\.47\.*",
+    ]
+    _supports_flash_attn = False
+    _supports_sdpa = True
+    _supports_flex_attn = True
+    _supports_static_cache = True
+    _supports_attention_backend = True
+
+    @classmethod
+    def from_pretrained(
+        cls, pretrained_model_name_or_path, *model_args, **kwargs
+    ):
+        key_mapping = getattr(cls, "_checkpoint_conversion_mapping", None)
+        if key_mapping is not None and kwargs.get("key_mapping") is None:
+            # Transformers only applies checkpoint renaming when key_mapping is
+            # passed explicitly; inheriting the class attribute alone is not enough.
+            kwargs["key_mapping"] = copy.deepcopy(key_mapping)
+        return super().from_pretrained(
+            pretrained_model_name_or_path, *model_args, **kwargs
+        )
+
+
+class Step3p7RotaryEmbedding(nn.Module):
+    def __init__(self, config: Step3p7TextConfig, device=None, layer_idx=None):
+        super().__init__()
+        self.layer_idx = layer_idx
+        self.max_seq_len_cached = config.max_position_embeddings
+        self.original_max_seq_len = config.max_position_embeddings
+
+        rope_theta = config.rope_theta
+        if isinstance(rope_theta, list):
+            rope_theta = rope_theta[0 if layer_idx is None else layer_idx]
+
+        partial_rotary_factor = getattr(config, "partial_rotary_factor", 1.0)
+        partial_rotary_factors = getattr(config, "partial_rotary_factors", None)
+        if partial_rotary_factors is not None:
+            partial_rotary_factor = partial_rotary_factors[
+                0 if layer_idx is None else layer_idx
+            ]
+
+        self.rope_theta = rope_theta
+        self.partial_rotary_factor = partial_rotary_factor
+
+        self.config = copy.copy(config)
+        self.config.rope_theta = rope_theta
+        self.config.partial_rotary_factor = partial_rotary_factor
+
+        if config.rope_parameters is not None:
+            self.config.rope_parameters = copy.deepcopy(config.rope_parameters)
+            self.config.rope_parameters["rope_theta"] = rope_theta
+            self.config.rope_parameters["partial_rotary_factor"] = (
+                partial_rotary_factor
+            )
+            self.rope_type = self.config.rope_parameters.get(
+                "rope_type", self.config.rope_parameters.get("type")
+            )
+        else:
+            self.rope_type = "default"
+
+        self.rope_init_fn = self.compute_default_rope_parameters
+        if self.rope_type != "default":
+            self.rope_init_fn = ROPE_INIT_FUNCTIONS[self.rope_type]
+        inv_freq, self.attention_scaling = self.rope_init_fn(
+            self.config, device
+        )
+
+        self.register_buffer("inv_freq", inv_freq, persistent=False)
+        self.original_inv_freq = self.inv_freq
+
+    @torch.no_grad()
+    @dynamic_rope_update  # power user: used with advanced RoPE types (e.g. dynamic rope)
+    def forward(self, x, position_ids):
+        inv_freq_expanded = (
+            self.inv_freq[None, :, None]
+            .float()
+            .expand(position_ids.shape[0], -1, 1)
+            .to(x.device)
+        )
+        position_ids_expanded = position_ids[:, None, :].float().to(x.device)
+
+        device_type = (
+            x.device.type
+            if isinstance(x.device.type, str) and x.device.type != "mps"
+            else "cpu"
+        )
+        with torch.autocast(
+            device_type=device_type, enabled=False
+        ):  # Force float32
+            freqs = (
+                inv_freq_expanded.float() @ position_ids_expanded.float()
+            ).transpose(1, 2)
+            emb = torch.cat((freqs, freqs), dim=-1)
+            cos = emb.cos() * self.attention_scaling
+            sin = emb.sin() * self.attention_scaling
+
+        return cos.to(dtype=x.dtype), sin.to(dtype=x.dtype)
+
+    @staticmethod
+    def compute_default_rope_parameters(
+        config: Step3p7TextConfig | None = None,
+        device: Optional["torch.device"] = None,
+    ) -> tuple["torch.Tensor", float]:
+        """
+        Computes the inverse frequencies according to the original RoPE implementation
+        Args:
+            config ([`~transformers.PreTrainedConfig`]):
+                The model configuration.
+            device (`torch.device`):
+                The device to use for initialization of the inverse frequencies.
+            seq_len (`int`, *optional*):
+                The current sequence length. Unused for this type of RoPE.
+        Returns:
+            Tuple of (`torch.Tensor`, `float`), containing the inverse frequencies for the RoPE embeddings and the
+            post-processing scaling factor applied to the computed cos/sin (unused in this type of RoPE).
+        """
+        base = config.rope_theta
+        partial_rotary_factor = getattr(
+            config, "partial_rotary_factor", 1.0
+        )
+        head_dim = (
+            getattr(config, "head_dim", None)
+            or config.hidden_size // config.num_attention_heads
+        )
+        dim = int(head_dim * partial_rotary_factor)
+
+        attention_factor = 1.0  # Unused in this type of RoPE
+
+        # Compute the inverse frequencies
+        inv_freq = 1.0 / (
+            base
+            ** (
+                torch.arange(0, dim, 2, dtype=torch.int64).to(
+                    device=device, dtype=torch.float
+                )
+                / dim
+            )
+        )
+        return inv_freq, attention_factor
+
+def rotate_half(x):
+    """Rotates half the hidden dims of the input."""
+    x1 = x[..., :x.shape[-1] // 2]
+    x2 = x[..., x.shape[-1] // 2:]
+    return torch.cat((-x2, x1), dim=-1)
+
+
+def apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, unsqueeze_dim=1):
+    """Applies Rotary Position Embedding to the query and key tensors.
+
+    Args:
+        q (`torch.Tensor`): The query tensor.
+        k (`torch.Tensor`): The key tensor.
+        cos (`torch.Tensor`): The cosine part of the rotary embedding.
+        sin (`torch.Tensor`): The sine part of the rotary embedding.
+        position_ids (`torch.Tensor`, *optional*):
+            Deprecated and unused.
+        unsqueeze_dim (`int`, *optional*, defaults to 1):
+            The 'unsqueeze_dim' argument specifies the dimension along which to unsqueeze cos[position_ids] and
+            sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example, note
+            that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim]. Then, if q and
+            k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1 makes
+            cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly, if q and k have
+            the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2.
+    Returns:
+        `tuple(torch.Tensor)` comprising of the query and key tensors rotated using the Rotary Position Embedding.
+    """
+    rotary_dim = cos.shape[-1]
+    q_rot, q_pass = q[..., :rotary_dim], q[..., rotary_dim:]
+    k_rot, k_pass = k[..., :rotary_dim], k[..., rotary_dim:]
+
+    # Apply rotary embeddings on the first half or full tensor
+    q_embed = (q_rot * cos) + (rotate_half(q_rot) * sin)
+    k_embed = (k_rot * cos) + (rotate_half(k_rot) * sin)
+
+    # Concatenate back to full shape
+    q_embed = torch.cat([q_embed, q_pass], dim=-1)
+    k_embed = torch.cat([k_embed, k_pass], dim=-1)
+    return q_embed, k_embed
+
+
+def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
+    """
+    This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch,
+    num_key_value_heads, seqlen, head_dim) to (batch, num_attention_heads, seqlen, head_dim)
+    """
+    batch, num_key_value_heads, slen, head_dim = hidden_states.shape
+    if n_rep == 1:
+        return hidden_states
+    hidden_states = hidden_states[:, :, None, :, :].expand(
+        batch, num_key_value_heads, n_rep, slen, head_dim
+    )
+    return hidden_states.reshape(
+        batch, num_key_value_heads * n_rep, slen, head_dim
+    )
+
+
+# Adapted from transformers.models.llama.modeling_llama.eager_attention_forward.
+# Llama4 does not cast attention weights to fp32 here.
+def eager_attention_forward(
+    module: nn.Module,
+    query: torch.Tensor,
+    key: torch.Tensor,
+    value: torch.Tensor,
+    attention_mask: Optional[torch.Tensor],
+    scaling: float,
+    dropout: float = 0.0,
+    **kwargs,
+):
+    key_states = repeat_kv(key, module.num_key_value_groups)
+    value_states = repeat_kv(value, module.num_key_value_groups)
+    # breakpoint()
+    attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling
+    if attention_mask is not None:
+        causal_mask = attention_mask[:, :, :, :key_states.shape[-2]]
+        attn_weights = attn_weights + causal_mask
+
+    attn_weights = nn.functional.softmax(attn_weights, dim=-1)
+    attn_weights = nn.functional.dropout(
+        attn_weights, p=dropout, training=module.training
+    )
+    attn_output = torch.matmul(attn_weights, value_states)
+    attn_output = attn_output.transpose(1, 2).contiguous()
+
+    return attn_output, attn_weights
+
+
+@dataclass
+class Step3p7CausalLMOutputWithPast(ModelOutput):
+    r"""
+    loss (`torch.FloatTensor` of shape `(1,)`, *optional*, returned when `labels` is provided):
+        Language modeling loss (for next-token prediction).
+    logits (`torch.FloatTensor` of shape `(batch_size, sequence_length, config.vocab_size)`):
+        Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).
+    past_key_values (`Cache`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`):
+        Tuple of `tuple(torch.FloatTensor)` of length `config.n_layers`, with each tuple having 2 tensors of shape
+        `(batch_size, num_heads, sequence_length, embed_size_per_head)`)
+        Contains pre-computed hidden-states (key and values in the self-attention blocks) that can be used (see
+        `past_key_values` input) to speed up sequential decoding.
+    """
+
+    loss: Optional[torch.FloatTensor] = None
+    last_hidden_state: Optional[torch.FloatTensor] = None
+    logits: torch.FloatTensor = None
+    past_key_values: Optional[list[torch.FloatTensor]] = None
+    hidden_states: Optional[tuple[torch.FloatTensor]] = None
+    attentions: Optional[tuple[torch.FloatTensor]] = None
+
+
+class Step3p7MLP(nn.Module):
+    def __init__(self, config, intermediate_size=None, swiglu_limit=None):
+        super().__init__()
+        self.config = config
+        self.hidden_size = config.hidden_size
+        self.intermediate_size = (
+            intermediate_size
+            if intermediate_size is not None
+            else config.intermediate_size
+        )
+        self.gate_proj = nn.Linear(self.hidden_size,
+                                   self.intermediate_size,
+                                   bias=False)
+        self.up_proj = nn.Linear(self.hidden_size,
+                                 self.intermediate_size,
+                                 bias=False)
+        self.down_proj = nn.Linear(self.intermediate_size,
+                                   self.hidden_size,
+                                   bias=False)
+        self.act_fn = ACT2FN["silu"]
+        self.limit = swiglu_limit
+
+    def forward(self, x):
+        up = self.up_proj(x)
+        gate = self.act_fn(self.gate_proj(x))
+        if self.limit is not None:
+            gate = gate.clamp(min=None, max=self.limit)
+            up = up.clamp(min=-self.limit, max=self.limit)
+
+        return self.down_proj(gate * up)
+
+
+def sigmoid_routing_function(gating_output: torch.Tensor, topk: int,
+                             renormalize: bool):
+    gating_output = gating_output.float()
+    gate_prob = torch.sigmoid(gating_output)
+    gate_prob = gate_prob / gate_prob.sum(dim=-1, keepdim=True)
+    topk_prob, indices = torch.topk(gate_prob, k=topk, dim=1)
+    expert_topk_weight = topk_prob
+    if renormalize:
+        expert_topk_weight = expert_topk_weight / torch.sum(
+            expert_topk_weight, dim=-1, keepdim=True)
+    return expert_topk_weight, indices
+
+
+def softmax_routing_function(gating_output: torch.Tensor, top_k: int,
+                             renormalize: bool):
+    gating_output = gating_output.float()
+    gate_prob = torch.softmax(gating_output, dim=-1)
+    gate_prob = gate_prob / gate_prob.sum(dim=-1, keepdim=True)
+    topk_prob, indices = torch.topk(gate_prob, k=top_k, dim=1)
+    expert_topk_weight = topk_prob
+    if renormalize:
+        expert_topk_weight = expert_topk_weight / torch.sum(
+            expert_topk_weight, dim=-1, keepdim=True)
+    return expert_topk_weight, indices.to(torch.int32)
+
+
+class MoELinear(nn.Module):
+
+    def __init__(self, num_experts, in_features, out_features):
+        super().__init__()
+        self.num_experts = num_experts
+        self.in_features = in_features
+        self.out_features = out_features
+        self.weight = nn.Parameter(
+            torch.empty(num_experts, out_features, in_features))
+
+    def forward(self, x, expert_id):
+        x = F.linear(x.float(), self.weight[expert_id].float())
+        return x
+
+
+class Step3p7MoEMLP(nn.Module):
+
+    def __init__(self, config, swiglu_limit=None):
+        super().__init__()
+        self.num_experts = config.moe_num_experts
+        self.top_k = config.moe_top_k
+        self.hidden_size = config.hidden_size
+        self.moe_intermediate_size = config.moe_intermediate_size
+
+        self.use_moe_router_bias = config.use_moe_router_bias
+        if self.use_moe_router_bias:
+            self.router_bias = nn.Parameter(torch.zeros(config.moe_num_experts,
+                                                        dtype=torch.float32),
+                                            requires_grad=False)
+            self.custom_routing_function = self.router_bias_func
+        elif config.moe_router_activation == "sigmoid":
+            self.custom_routing_function = sigmoid_routing_function
+        else:
+            self.custom_routing_function = None
+        self.need_fp32_gate = config.need_fp32_gate
+        self.routed_scaling_factor = getattr(config,
+                                             "moe_router_scaling_factor", 1.0)
+        
+        # gating
+        self.gate = nn.Linear(self.hidden_size, self.num_experts, bias=False)
+            
+        self.act_fn = ACT2FN["silu"]
+        self.limit = swiglu_limit
+
+        self.up_proj = MoELinear(self.num_experts, self.hidden_size,
+                                 self.moe_intermediate_size)
+        self.gate_proj = MoELinear(self.num_experts, self.hidden_size,
+                                   self.moe_intermediate_size)
+        self.down_proj = MoELinear(self.num_experts,
+                                   self.moe_intermediate_size,
+                                   self.hidden_size)
+
+    def router_bias_func(self, gating_output: torch.Tensor, topk: int,
+                         renormalize: bool):
+        gate_prob = torch.sigmoid(gating_output.float())
+        gate_prob_with_bias = gate_prob + self.router_bias.unsqueeze(0)
+        _, indices = torch.topk(gate_prob_with_bias, k=topk, dim=1)
+        topk_prob = torch.gather(gate_prob, 1, indices)
+        expert_topk_weight = topk_prob
+        if renormalize:
+            expert_topk_weight = expert_topk_weight / (
+                torch.sum(expert_topk_weight, dim=-1, keepdim=True) + 1e-20)
+        return expert_topk_weight, indices
+
+    def get_expert_output(self, inputs: torch.Tensor, expert_id):
+        #if self.limit is None:
+        up = self.up_proj(inputs, expert_id)
+        gate = self.act_fn(self.gate_proj(inputs, expert_id))
+        if self.limit is not None:
+            gate = gate.clamp(min=None, max=self.limit)
+            up = up.clamp(min=-self.limit, max=self.limit)
+
+        return self.down_proj(gate * up, expert_id)
+
+    def forward(self, hidden_states):
+        """ """
+        batch_size, sequence_length, hidden_dim = hidden_states.shape
+        hidden_states = hidden_states.view(-1, hidden_dim)
+        if self.need_fp32_gate:
+            router_logits = torch.matmul(
+                hidden_states.to(torch.float32),
+                self.gate.weight.t().to(torch.float32),
+            )
+        else:
+            # router_logits: (batch * sequence_length, n_experts)
+            router_logits = self.gate(hidden_states)
+
+        if self.custom_routing_function:
+            routing_weights, selected_experts = self.custom_routing_function(
+                router_logits, self.top_k, renormalize=True)
+        else:
+            routing_weights = F.softmax(router_logits,
+                                        dim=1,
+                                        dtype=torch.float)
+            routing_weights, selected_experts = torch.topk(routing_weights,
+                                                           self.top_k,
+                                                           dim=-1)
+
+        routing_weights = routing_weights * self.routed_scaling_factor
+
+        final_hidden_states = torch.zeros(
+            (batch_size * sequence_length, hidden_dim),
+            dtype=hidden_states.dtype,
+            device=hidden_states.device)
+
+        # One hot encode the selected experts to create an expert mask
+        # this will be used to easily index which expert is going to be sollicitated
+        expert_mask = torch.nn.functional.one_hot(
+            selected_experts, num_classes=self.num_experts).permute(2, 1, 0)
+
+        # Loop over all available experts in the model and perform the computation on each expert
+        for expert_idx in range(self.num_experts):
+            idx, top_x = torch.where(expert_mask[expert_idx])
+
+            # Index the correct hidden states and compute the expert hidden state for
+            # the current expert. We need to make sure to multiply the output hidden
+            # states by `routing_weights` on the corresponding tokens (top-1 and top-2)
+            current_state = hidden_states[None, top_x].reshape(-1, hidden_dim)
+            current_hidden_states = (
+                self.get_expert_output(current_state, expert_idx) *
+                routing_weights[top_x, idx, None])
+
+            # However `index_add_` only support torch tensors for indexing so we'll use
+            # the `top_x` tensor here.
+            final_hidden_states.index_add_(
+                0, top_x, current_hidden_states.to(hidden_states.dtype))
+        final_hidden_states = final_hidden_states.reshape(
+            batch_size, sequence_length, hidden_dim)
+        return final_hidden_states
+
+
+class Step3p7RMSNorm(nn.Module):
+
+    def __init__(
+        self,
+        hidden_size: int,
+        eps: float = 1e-5,
+    ) -> None:
+        super().__init__()
+        self.weight = nn.Parameter(torch.ones(hidden_size))
+        self.variance_epsilon = eps
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        dtype = x.dtype
+        x = x.float()
+        variance = x.pow(2).mean(dim=-1, keepdim=True)
+        normed = x * torch.rsqrt(variance + self.variance_epsilon)
+        normed = normed * (self.weight.float() + 1)
+        return normed.to(dtype)
+class Step3p7Attention(nn.Module):
+
+    def __init__(self, config: Step3p7TextConfig, layer_idx):
+        super().__init__()
+        self.config = config
+        self.layer_idx = layer_idx
+        self.num_attention_heads = config.num_attention_heads
+        self.num_key_value_heads = config.num_attention_groups
+
+        layer_types = getattr(config, "layer_types", [])
+        if layer_types:
+            enable_sliding_window = layer_types[
+                self.layer_idx] == "sliding_attention"
+        else:
+            enable_sliding_window = self.layer_idx % 2 == 0
+        
+        yarn_only_types = getattr(config, "yarn_only_types", None)
+        if yarn_only_types and layer_types[
+                self.layer_idx] not in yarn_only_types:
+            config.rope_parameters = None
+        else:
+            config.rope_parameters = getattr(config, "rope_scaling", None)
+
+        self.sliding_window = config.sliding_window
+        if enable_sliding_window:
+            self.num_attention_heads = config.attention_other_setting[
+                "num_attention_heads"]
+            self.num_key_value_heads = config.attention_other_setting[
+                "num_attention_groups"]
+
+        if self.sliding_window is not None and enable_sliding_window:
+            self.sliding_window = (self.sliding_window)
+        else:
+            self.sliding_window = None
+        self.head_dim = getattr(config, "head_dim",
+                        config.hidden_size // self.num_attention_heads)
+        self.num_key_value_groups = self.num_attention_heads // self.num_key_value_heads
+
+        self.rotary_emb = Step3p7RotaryEmbedding(config, layer_idx=layer_idx)
+
+        self.q_size = self.num_attention_heads * self.head_dim
+        self.kv_size = self.num_key_value_heads * self.head_dim
+        self.scaling = self.head_dim**-0.5
+
+        self.q_proj = nn.Linear(config.hidden_size, self.q_size, bias=False)
+        self.k_proj = nn.Linear(config.hidden_size, self.kv_size, bias=False)
+        self.v_proj = nn.Linear(config.hidden_size, self.kv_size, bias=False)
+        self.o_proj = nn.Linear(self.q_size, config.hidden_size, bias=False)
+        self.attention_dropout = getattr(config, "attention_dropout", 0.0)
+        self.q_norm = Step3p7RMSNorm(self.head_dim,
+                                    eps=config.rms_norm_eps)
+        self.k_norm = Step3p7RMSNorm(self.head_dim,
+                                    eps=config.rms_norm_eps)
+
+        self.use_head_wise_attn_gate = config.use_head_wise_attn_gate
+        if self.use_head_wise_attn_gate:
+            self.g_proj = nn.Linear(config.hidden_size,
+                                    self.num_attention_heads,
+                                    bias=False)
+
+        self.use_rope = True
+        use_rope_layers = getattr(config, "use_rope_layers", None)
+        if use_rope_layers:
+            self.use_rope = use_rope_layers[self.layer_idx]
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor],
+        past_key_value: Optional[Cache] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        **kwargs: Unpack[FlashAttentionKwargs],
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor],
+               Optional[Tuple[torch.Tensor]]]:
+        input_shape = hidden_states.shape[:-1]
+        hidden_shape = (*input_shape, -1, self.head_dim)
+
+        query_states = self.q_norm(
+            self.q_proj(hidden_states).view(hidden_shape)).transpose(1, 2)
+        key_states = self.k_norm(
+            self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)
+        value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(
+            1, 2)
+        if self.use_head_wise_attn_gate:
+            gate_states = self.g_proj(hidden_states)
+        cos, sin = self.rotary_emb(hidden_states, position_ids)
+
+        # cos, sin = position_embeddings
+        query_states, key_states = apply_rotary_pos_emb(
+            query_states, key_states, cos, sin)
+
+        # query_states, key_states = apply_rotary_pos_emb(query_norm_states, key_norm_states, cos, sin)
+        if past_key_value is not None:
+            # sin and cos are specific to RoPE models; position_ids needed for the static cache
+            cache_kwargs = {
+                "sin": sin,
+                "cos": cos,
+                "cache_position": cache_position
+            }
+            key_states, value_states = past_key_value.update(
+                key_states, value_states, self.layer_idx, cache_kwargs)
+
+        attention_interface: Callable = eager_attention_forward
+        # TODO: considering FP8；
+        # RuntimeError: Expected attn_mask dtype to be bool or float or to match query dtype,
+        # but got attn_mask.dtype: long int and  query.dtype: c10::BFloat16 instead.
+        if self.config._attn_implementation != "eager":
+            attention_interface = ALL_ATTENTION_FUNCTIONS[
+                self.config._attn_implementation]
+
+        attn_output, attn_weights = attention_interface(
+            self,
+            query_states,
+            key_states,
+            value_states,
+            attention_mask,
+            dropout=0.0 if not self.training else self.attention_dropout,
+            scaling=self.scaling,
+            sliding_window=self.sliding_window,  # main diff with Llama
+            **kwargs,
+        )
+        attn_output = attn_output.reshape(*input_shape, -1)
+        if self.use_head_wise_attn_gate:
+            output = attn_output.view(
+                *attn_output.shape[:-1], self.num_attention_heads,
+                self.head_dim) * gate_states.unsqueeze(-1).sigmoid()
+            attn_output = output.view(*attn_output.shape)
+        attn_output = self.o_proj(attn_output)
+
+        return attn_output, attn_weights
+
+
+class Step3p7DecoderLayer(GradientCheckpointingLayer):
+
+    def __init__(self, config, layer_idx):
+        super().__init__()
+        self.hidden_size = config.hidden_size
+        self.layer_idx = layer_idx
+        self.self_attn = Step3p7Attention(config, layer_idx)
+        layer_types = getattr(config, "layer_types", None) or []
+        if layer_types:
+            self.attention_type = layer_types[layer_idx]
+        else:
+            self.attention_type = (
+                "sliding_attention" if layer_idx % 2 == 0 else "full_attention"
+            )
+
+        moe_layers_enum = getattr(config, "moe_layers_enum", None)
+        if moe_layers_enum is not None:
+            if isinstance(moe_layers_enum, str):
+                moe_layers_idx = [
+                    int(i) for i in moe_layers_enum.split(',') if i.strip()
+                ]
+            else:
+                moe_layers_idx = [int(i) for i in moe_layers_enum]
+        else:
+            moe_layers_idx = [i for i in range(1, config.num_hidden_layers)]
+        self.is_moe_layer = layer_idx in moe_layers_idx
+        self.use_moe = False
+
+        if config.swiglu_limits_shared and config.swiglu_limits_shared[
+                layer_idx] is not None and config.swiglu_limits_shared[
+                    layer_idx] != 0:
+            swiglu_limit_shared = config.swiglu_limits_shared[layer_idx]
+        else:
+            swiglu_limit_shared = None
+        if config.swiglu_limits and config.swiglu_limits[
+                layer_idx] is not None and config.swiglu_limits[layer_idx] != 0:
+            swiglu_limit = config.swiglu_limits[layer_idx]
+        else:
+            swiglu_limit = None
+        if self.is_moe_layer:
+            self.moe = Step3p7MoEMLP(config, swiglu_limit=swiglu_limit)  #
+            self.share_expert = Step3p7MLP(
+                config,
+                intermediate_size=config.share_expert_dim,
+                swiglu_limit=swiglu_limit_shared)
+            self.use_moe = True
+        else:
+            self.mlp = Step3p7MLP(config,
+                                 intermediate_size=config.intermediate_size,
+                                 swiglu_limit=swiglu_limit_shared)
+
+        self.input_layernorm = Step3p7RMSNorm(
+            config.hidden_size,
+            eps=config.rms_norm_eps)
+        self.post_attention_layernorm = Step3p7RMSNorm(
+            config.hidden_size,
+            eps=config.rms_norm_eps)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[tuple[torch.Tensor]] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        **kwargs: Unpack[FlashAttentionKwargs],
+    ) -> torch.FloatTensor:
+        residual = hidden_states
+        hidden_states = self.input_layernorm(hidden_states)
+        hidden_states, _ = self.self_attn(
+            hidden_states=hidden_states,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_value=past_key_value,
+            cache_position=cache_position,
+            **kwargs,
+        )
+        hidden_states = residual + hidden_states
+        
+        # Fully Connected
+        residual = hidden_states
+        hidden_states = self.post_attention_layernorm(hidden_states)
+        if self.use_moe:
+            share_output = self.share_expert(hidden_states)
+            moe_output = self.moe(hidden_states)
+            ffn_output = moe_output + share_output
+        else:
+            ffn_output = self.mlp(hidden_states)
+        if isinstance(ffn_output, tuple):
+            hidden_states, _ = ffn_output
+        else:
+            hidden_states = ffn_output
+
+        hidden_states = residual + hidden_states
+        return hidden_states
+
+
+class Step3p7TextPreTrainedModel(PreTrainedModel):
+    # Link this model family to its configuration class so PreTrainedModel.from_pretrained
+    # can load the config instead of failing with a NoneType error.
+    config_class = Step3p7TextConfig
+    supports_gradient_checkpointing = True
+    _skip_keys_device_placement = ["past_key_values"]
+    _keys_to_ignore_on_load_unexpected = [
+        r"model\.layers\.45\.*",
+        r"model\.layers\.46\.*",
+        r"model\.layers\.47\.*",
+    ]
+    _supports_flash_attn = False
+    _supports_sdpa = True
+    _supports_flex_attn = True
+    _supports_static_cache = True
+    _supports_attention_backend = True
+
+
+class Step3p7TextModel(Step3p7TextPreTrainedModel, GenerationMixin):
+    _no_split_modules = ["Step3p7DecoderLayer"]
+    base_model_prefix = "model"
+    _tied_weights_keys = ["lm_head.weight"]
+    config: Step3p7TextConfig
+
+    def __init__(self, config: Step3p7TextConfig):
+        super().__init__(config)
+        self.padding_idx = config.pad_token_id
+        self.vocab_size = config.vocab_size
+
+        self.embed_tokens = nn.Embedding(config.vocab_size, config.hidden_size,
+                                         self.padding_idx)
+        self.layers = nn.ModuleList([
+            Step3p7DecoderLayer(config, layer_idx)
+            for layer_idx in range(config.num_hidden_layers)
+        ])
+        self.norm = Step3p7RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.gradient_checkpointing = False
+        layer_types = self.config.layer_types or []
+        self.has_sliding_layers = (not layer_types or
+                                   "sliding_attention" in layer_types)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    
+    def get_input_embeddings(self, input_ids):
+        return self.embed_tokens(input_ids)
+
+    @can_return_tuple
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Cache] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        **kwargs: Unpack[TransformersKwargs],
+    ) -> Union[tuple, BaseModelOutputWithPast]:
+        output_attentions = (
+            output_attentions
+            if output_attentions is not None
+            else self.config.output_attentions
+        )
+        output_hidden_states = (
+            output_hidden_states
+            if output_hidden_states is not None
+            else self.config.output_hidden_states
+        )
+        use_cache = use_cache if use_cache is not None else self.config.use_cache
+        return_dict = (
+            return_dict
+            if return_dict is not None
+            else getattr(self.config, "return_dict", True)
+        )
+        if (input_ids is None) ^ (inputs_embeds is not None):
+            raise ValueError(
+                "You must specify exactly one of input_ids or inputs_embeds")
+
+        if self.gradient_checkpointing and self.training and use_cache:
+            logger.warning_once(
+                "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`."
+            )
+            use_cache = False
+
+        if inputs_embeds is None:
+            inputs_embeds = self.embed_tokens(
+                input_ids.to(self.embed_tokens.weight.device))
+
+        if use_cache and past_key_values is None:
+            past_key_values = DynamicCache()
+
+        if cache_position is None:
+            past_seen_tokens = past_key_values.get_seq_length(
+            ) if past_key_values is not None else 0
+            cache_position = torch.arange(past_seen_tokens,
+                                          past_seen_tokens +
+                                          inputs_embeds.shape[1],
+                                          device=inputs_embeds.device)
+
+        if position_ids is None:
+            position_ids = cache_position.unsqueeze(0)
+
+        hidden_states = inputs_embeds
+
+        # It may already have been prepared by e.g. `generate`
+        if not isinstance(causal_mask_mapping := attention_mask, dict):
+            # Prepare mask arguments
+            mask_kwargs = {
+                "config": self.config,
+                "attention_mask": attention_mask,
+                "past_key_values": past_key_values,
+                "position_ids": position_ids,
+            }
+            mask_kwargs[_MASK_INPUT_EMBEDS_ARG] = inputs_embeds
+            # Create the masks
+            causal_mask_mapping = {
+                "full_attention": create_causal_mask(**mask_kwargs),
+            }
+
+            # The sliding window alternating layers are not always activated depending on the config
+            if self.has_sliding_layers:
+                causal_mask_mapping[
+                    "sliding_attention"] = create_sliding_window_causal_mask(
+                        **mask_kwargs)
+
+        # # create position embeddings to be shared across the decoder layers
+        # decoder layers
+        all_hidden_states = () if output_hidden_states else None
+        all_self_attns = () if output_attentions else None
+        for decoder_layer in self.layers[:self.config.num_hidden_layers]:
+            if output_hidden_states:
+                all_hidden_states += (hidden_states, )
+
+            layer_outputs = decoder_layer(
+                hidden_states,
+                attention_mask=causal_mask_mapping[
+                    decoder_layer.attention_type],
+                position_ids=position_ids,
+                past_key_value=past_key_values,
+                output_attentions=output_attentions,
+                use_cache=use_cache,
+                cache_position=cache_position,
+                **kwargs,
+            )
+
+            hidden_states = layer_outputs
+
+        hidden_states = self.norm(hidden_states)
+
+        return BaseModelOutputWithPast(
+            last_hidden_state=hidden_states,
+            past_key_values=past_key_values if use_cache else None,
+            hidden_states=all_hidden_states,
+            attentions=all_self_attns,
+        )
+
+
+class Step3p7Model(Step3p7PreTrainedModel, GenerationMixin):
+    config: Step3p7Config
+    _tied_weights_keys = ["lm_head.weight"]
+    base_model_prefix = ""
+
+    def __init__(self, config: Step3p7Config):
+        super().__init__(config)
+        self.vision_model = StepRoboticsVisionEncoder(config.vision_config)
+        self.language_model = Step3p7TextModel(config.text_config)
+        self.vocab_size = config.text_config.vocab_size
+        self.vit_large_projector = nn.Linear(
+                config.vision_config.width * 4,
+                config.text_config.hidden_size,                
+                bias=config.projector_bias) 
+        self.image_placeholder_token_id = config.image_token_id
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(
+        self,
+        input_ids: torch.Tensor,
+        multimodal_embeddings  = None,
+    ) -> torch.Tensor:
+        # breakpoint()
+        input_ids = input_ids.squeeze(0)
+        if multimodal_embeddings is None:
+            inputs_embeds = self.language_model.get_input_embeddings(input_ids)
+        else:
+            is_text = input_ids != self.config.image_token_id
+            text_ids = input_ids[is_text]
+            text_embeds = self.language_model.get_input_embeddings(text_ids)
+                     
+            inputs_embeds = torch.empty(input_ids.shape[0],
+                                        text_embeds.shape[-1],
+                                        dtype=text_embeds.dtype,
+                                        device=text_embeds.device)
+            inputs_embeds[is_text] = text_embeds
+            inputs_embeds = merge_multimodal_embeddings(
+                input_ids, inputs_embeds, multimodal_embeddings,
+                self.config.image_token_id)
+        inputs_embeds = inputs_embeds.unsqueeze(0)
+        return inputs_embeds
+       
+
+    def set_input_embeddings(self, value):
+        return self.language_model.set_input_embeddings(value)
+
+    def set_decoder(self, decoder):
+        self.language_model = decoder
+
+    def get_decoder(self):
+        return self.language_model
+    
+    def _parse_and_validate_image_input(
+            self, **kwargs: object) -> Optional[StepVLImageInputs]:
+        pixel_values = kwargs.pop("pixel_values", None)
+        patch_pixel_values = kwargs.pop("patch_pixel_values", None)
+        num_patches = kwargs.pop("num_patches", None)
+        image_embeds = kwargs.pop("image_embeds", None)
+
+        if pixel_values is None and image_embeds is None:
+            return None
+
+        if pixel_values is not None:
+            # pixel_values = flatten_bn(pixel_values, concat=True)
+            if pixel_values.dim() >= 3:
+                pixel_values = pixel_values.view(-1, *pixel_values.shape[-3:])
+            if patch_pixel_values is not None:
+                # patch_pixel_values = flatten_bn(patch_pixel_values,
+                #                                 concat=True)
+                patch_pixel_values = patch_pixel_values.view(
+                    -1, *patch_pixel_values.shape[-3:])
+                # Handle empty patch_pixel_values by setting to None
+                if patch_pixel_values.shape[0] == 0:
+                    patch_pixel_values = None
+
+            return StepVLImagePixelInputs(
+                type="pixel_values",
+                pixel_values=pixel_values.to(self.dtype).to(self.device),
+                patch_pixel_values=patch_pixel_values.to(self.dtype).to(
+                    self.device) if patch_pixel_values is not None else None,
+                num_patches=num_patches,
+            )
+
+        if image_embeds is not None:
+            if image_embeds.dim() == 2 or image_embeds.dim() >= 3:
+                image_embeds = image_embeds.view(-1, image_embeds.shape[-1])
+            else:
+                raise ValueError(
+                    f"Unexpected shape for image_embeds: {image_embeds.shape}")
+
+            return StepVLImageEmbeddingInputs(
+                type="image_embeds",
+                image_embeds=image_embeds.to(self.dtype).to(self.device),
+            )
+        return None
+    
+    def _process_image_features(self,
+                                image_features: torch.Tensor) -> torch.Tensor:
+        B, P = image_features.shape[:2]
+        HW = int(P ** 0.5)
+        image_features = image_features.permute(0, 2, 1).view(B, -1, HW, HW)
+        image_features = self.vision_model.vit_downsampler1(image_features)
+        image_features = self.vision_model.vit_downsampler2(image_features)
+
+        B, C, HW, HW = image_features.shape
+        image_features = image_features.view(B, -1, HW * HW).permute(0, 2, 1)
+        image_features = self.vit_large_projector(image_features)
+        return image_features
+
+    def _get_vision_model_output(self,
+                                 input_tensor: torch.Tensor) -> torch.Tensor:
+        return self.vision_model(input_tensor)
+
+    def _process_image_input(
+            self, image_input: StepVLImageInputs) -> tuple[torch.Tensor, ...]:
+
+        if image_input["type"] == "image_embeds":
+            image_features = image_input["image_embeds"]
+        else:
+            image_features = self._get_vision_model_output(
+                image_input["pixel_values"])
+            patch_image_features = self._get_vision_model_output(
+                image_input["patch_pixel_values"]
+            ) if image_input["patch_pixel_values"] is not None else None
+            num_patches = image_input["num_patches"]
+
+        image_features = self._process_image_features(image_features)
+        patch_image_features = self._process_image_features(
+            patch_image_features) if patch_image_features is not None else None
+
+        merged_image_features = []
+        cur_patch_idx = 0
+        for i, num_patch in enumerate(num_patches):
+            cur_feature = []
+            if num_patch > 0:
+                patch_slice = patch_image_features[
+                    cur_patch_idx:cur_patch_idx + num_patch]
+                cur_feature.append(patch_slice.view(-1, patch_slice.shape[-1]))
+            cur_feature.append(image_features[i].view(
+                -1, image_features.shape[-1]))
+            cur_patch_idx += num_patch
+            merged_image_features.append(
+                torch.cat(cur_feature) if len(cur_feature) >
+                1 else cur_feature[0])
+    
+        return merged_image_features
+    
+    def get_multimodal_embeddings(self, **kwargs):
+        # breakpoint()
+        image_input = self._parse_and_validate_image_input(**kwargs)
+        if image_input is None:
+            return None
+        vision_embeddings = self._process_image_input(image_input)
+        return vision_embeddings
+
+    @can_return_tuple
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        logits_to_keep: Union[int, torch.Tensor] = 0,
+        images: Optional[list[Image.Image]] = None,
+        **kwargs: Unpack[TransformersKwargs],
+    ) -> Union[tuple, Step3p7CausalLMOutputWithPast]:
+        r"""
+        labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Labels for computing the masked language modeling loss. Indices should either be in `[0, ...,
+            config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored
+            (masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`.
+        Example:
+        ```python
+        >>> from transformers import AutoTokenizer, Llama4ForCausalLM
+        >>> model = Llama4ForCausalLM.from_pretrained("meta-llama4/Llama4-2-7b-hf")
+        >>> tokenizer = AutoTokenizer.from_pretrained("meta-llama4/Llama4-2-7b-hf")
+        >>> prompt = "Hey, are you conscious? Can you talk to me?"
+        >>> inputs = tokenizer(prompt, return_tensors="pt")
+        >>> # Generate
+        >>> generate_ids = model.generate(inputs.input_ids, max_length=30)
+        >>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
+        "Hey, are you conscious? Can you talk to me?\nI'm not conscious, but I can talk to you."
+        ```"""
+        output_attentions = (
+            output_attentions
+            if output_attentions is not None
+            else self.config.output_attentions
+        )
+        output_hidden_states = (
+            output_hidden_states
+            if output_hidden_states is not None
+            else self.config.output_hidden_states
+        )
+        return_dict = (
+            return_dict if return_dict is not None else self.config.use_return_dict
+        )
+
+        if inputs_embeds is None:
+            input_ids = input_ids
+            vision_embeddings = self.get_multimodal_embeddings(**kwargs)
+            inputs_embeds = self.get_input_embeddings(input_ids,
+                                                      vision_embeddings)
+            input_ids = None
+        # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn)
+        outputs = self.language_model(
+            input_ids=None,
+            position_ids=position_ids,
+            attention_mask=attention_mask,
+            past_key_values=past_key_values,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=True,
+            cache_position=cache_position,
+            **kwargs,
+        )
+
+        output = Step3p7CausalLMOutputWithPast(
+            last_hidden_state=outputs.last_hidden_state,
+            past_key_values=outputs.past_key_values,
+            attentions=outputs.attentions,
+        )
+        return output if return_dict else output.to_tuple()
+
+
+class Step3p7ForConditionalGeneration(Step3p7PreTrainedModel, GenerationMixin):
+    _checkpoint_conversion_mapping = {
+        "^vision_model": "model.vision_model",
+        r"^model(?!\.(language_model|vision_model))": "model.language_model",
+        "^vit_large_projector": "model.vit_large_projector",
+    }
+    _tied_weights_keys = ["lm_head.weight"]
+    config: Step3p7Config
+
+    def __init__(self, config: Step3p7Config):
+        super().__init__(config)
+        self.model = Step3p7Model(config)
+        self.lm_head = nn.Linear(config.hidden_size,
+                                config.text_config.vocab_size,
+                                bias=False)
+
+        self.post_init()
+
+    def get_input_embeddings(self):
+        return self.model.get_input_embeddings()
+
+    def set_input_embeddings(self, value):
+        self.model.set_input_embeddings(value)
+
+    def get_output_embeddings(self):
+        return self.model.get_output_embeddings()
+
+    def set_output_embeddings(self, new_embeddings):
+        self.model.set_output_embeddings(new_embeddings)
+
+    def set_decoder(self, decoder):
+        self.model.set_decoder(decoder)
+
+    def get_decoder(self):
+        return self.model.get_decoder()
+
+    @property
+    def language_model(self):
+        return self.model.language_model
+
+    @property
+    def visual(self):
+        return self.model.vision_model
+
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        pixel_values: Optional[torch.Tensor] = None,
+        num_patches=None,
+        patch_pixel_values=None,
+        patch_newline_mask=None,
+        image_embeds: Optional[torch.FloatTensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Cache] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        **kwargs: Unpack[TransformersKwargs],
+    ) -> Union[tuple, Step3p7CausalLMOutputWithPast]:
+        output_attentions = (
+            output_attentions
+            if output_attentions is not None
+            else self.config.output_attentions
+        )
+        output_hidden_states = (
+            output_hidden_states
+            if output_hidden_states is not None
+            else self.config.output_hidden_states
+        )
+
+        outputs = self.model(
+            input_ids=input_ids,
+            num_patches=num_patches,
+            patch_pixel_values=patch_pixel_values,
+            patch_newline_mask=patch_newline_mask,
+            position_ids=position_ids,
+            attention_mask=attention_mask,
+            past_key_values=past_key_values,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+            cache_position=cache_position,
+            **kwargs,
+        )
+
+        hidden_states = outputs.last_hidden_state
+        logits = self.lm_head(hidden_states)
+
+        los = None
+        if labels is not None:
+            loss = self.loss_function(
+                logits=logits, labels=labels, vocab_size=self.config.vocab_size
+            )
+
+        return Step3p7CausalLMOutputWithPast(
+            logits=logits,
+        )
+
+
+    def prepare_inputs_for_generation(
+        self,
+        input_ids,
+        past_key_values=None,
+        inputs_embeds=None,
+        pixel_values=None,
+        patch_pixel_values=None,
+        num_patches=None,
+        image_embeds=None,
+        attention_mask=None,
+        cache_position=None,
+        logits_to_keep=None,
+        **kwargs,
+    ):
+        model_inputs = super().prepare_inputs_for_generation(
+            input_ids,
+            past_key_values=past_key_values,
+            inputs_embeds=inputs_embeds,
+            attention_mask=attention_mask,
+            cache_position=cache_position,
+            logits_to_keep=logits_to_keep,
+            **kwargs,
+        )
+
+        generation_cache_position = model_inputs.get("cache_position", cache_position)
+        is_prefill = past_key_values is None
+        if generation_cache_position is not None and generation_cache_position.numel() > 0:
+            is_prefill = generation_cache_position[0].item() == 0
+
+        if is_prefill:
+            # During cached decoding, input ids no longer contain image tokens,
+            # so pixel values should only be passed at the first step.
+            model_inputs["pixel_values"] = pixel_values
+
+        return model_inputs
+    
+    def _fix_state_dict_key_on_load(self, key: str) -> tuple[str, bool]:
+        if key.startswith("language_model."):
+            return key[len("language_model.") :], True
+        
+        return key, False

processing_step3.py

@@ -0,0 +1,475 @@
+from transformers import BaseImageProcessor, ImageProcessingMixin
+from transformers.processing_utils import ImagesKwargs, MultiModalData, ProcessingKwargs, ProcessorMixin, Unpack, VideosKwargs
+import math
+from typing import Iterable, Optional, Tuple, List, TypedDict, Literal, Union, overload
+
+from PIL import Image
+import torch
+import numpy as np
+import torchvision
+from torch import nn
+from torch.nn import functional as F, LayerNorm
+from torchvision.transforms.functional import InterpolationMode
+from transformers.activations import ACT2FN
+from torchvision import transforms
+from torchvision.transforms.functional import InterpolationMode
+from transformers.feature_extraction_utils import BatchFeature, TensorType
+from transformers.image_utils import ImageInput
+from transformers.processing_utils import ProcessingKwargs, ProcessorMixin, Unpack
+from transformers.tokenization_utils_tokenizers import TokenizersBackend
+from math import ceil
+from itertools import product
+
+
+
+MAX_IMAGE_SIZE: int = 3024
+
+class Step3VLImagePixelInputs(TypedDict):
+    type: Literal["pixel_values"]
+    pixel_values: torch.Tensor
+    patch_pixel_values: Optional[torch.Tensor]
+    num_patches: list[int]
+
+
+class Step3VLImageEmbeddingInputs(TypedDict):
+    type: Literal["image_embeds"]
+    image_embeds: torch.Tensor
+
+
+ImageWithPatches = tuple[Image.Image, list[Image.Image], list[int] | None]
+
+
+class GPUToTensor(torch.nn.Module):
+
+    def forward(self, raw_image: Union[np.ndarray,
+                                       Image.Image]) -> torch.Tensor:
+        if isinstance(raw_image, Image.Image):
+            return transforms.ToTensor()(raw_image)
+        if raw_image.ndim == 2:
+            raw_image = raw_image[:, :, None].repeat(3, -1)
+        if torch.cuda.is_available():
+            device = torch.device("cuda")
+        else:
+            device = torch.device("cpu")
+        image_tensor = torch.from_numpy(raw_image).to(device)
+        image_tensor = torch.permute(image_tensor, (2, 0, 1)).contiguous()
+        if image_tensor.dtype == torch.uint8:
+            image_tensor = image_tensor.to(torch.float32).div(255)
+        return image_tensor
+
+class Step3VisionProcessor(BaseImageProcessor):
+
+    def __init__(self, size, interpolation_mode="bicubic", patch_size=None):
+        mean = [0.48145466, 0.4578275, 0.40821073]
+        std = [0.26862954, 0.26130258, 0.27577711]
+        patch_size = patch_size if patch_size is not None else size
+
+        self.transform = transforms.Compose([
+            GPUToTensor(),
+            transforms.Normalize(mean, std),
+            transforms.Resize(
+                (size, size),
+                interpolation=InterpolationMode.BICUBIC if interpolation_mode
+                == "bicubic" else InterpolationMode.BILINEAR,
+                antialias=True),
+        ])
+
+        self.patch_transform = transforms.Compose([
+            GPUToTensor(),
+            transforms.Normalize(mean, std),
+            transforms.Resize(
+                (patch_size, patch_size),
+                interpolation=InterpolationMode.BICUBIC if interpolation_mode
+                == "bicubic" else InterpolationMode.BILINEAR,
+                antialias=True),
+        ]) if patch_size is not None else None
+
+    def __call__(self, image, is_patch=False):
+        if is_patch:
+            return {"pixel_values": self.patch_transform(image).unsqueeze(0)}
+        else:
+            return {"pixel_values": self.transform(image).unsqueeze(0)}
+
+class ImagePatcher:
+    def determine_window_size(self, long: int, short: int) -> int:
+        if long <= 728:
+            return short if long / short > 1.5 else 0
+        return min(short, 504) if long / short > 4 else 504
+    def slide_window(
+        self,
+        width: int,
+        height: int,
+        sizes: list[tuple[int, int]],
+        steps: list[tuple[int, int]],
+        img_rate_thr: float = 0.6,
+    ) -> tuple[list[tuple[int, int, int, int]], tuple[int, int]]:
+        assert 1 >= img_rate_thr >= 0, "The `in_rate_thr` should lie in 0~1"
+        windows = []
+        # Sliding windows.
+        for size, step in zip(sizes, steps):
+            size_w, size_h = size
+            step_w, step_h = step
+
+            x_num = 1 if width <= size_w else ceil((width - size_w) / step_w +
+                                                   1)
+            x_start = [step_w * i for i in range(x_num)]
+            if len(x_start) > 1 and x_start[-1] + size_w > width:
+                x_start[-1] = width - size_w
+
+            y_num = 1 if height <= size_h else ceil((height - size_h) /
+                                                    step_h + 1)
+            y_start = [step_h * i for i in range(y_num)]
+            if len(y_start) > 1 and y_start[-1] + size_h > height:
+                y_start[-1] = height - size_h
+
+            start = np.array(list(product(y_start, x_start)), dtype=int)
+            start[:, [0, 1]] = start[:, [1, 0]]
+            windows.append(np.concatenate([start, start + size], axis=1))
+        windows = np.concatenate(windows, axis=0)
+
+        return [(int(box[0]), int(box[1]), int(box[2] - box[0]),
+                 int(box[3] - box[1])) for box in windows], (x_num, y_num)
+
+    def square_pad(self, img: Image.Image) -> Image.Image:
+        w, h = img.size
+        if w == h:
+            return img
+        size = max(w, h)
+        padded = Image.new(img.mode, (size, size), 0)
+        padded.paste(img, (0, 0))
+        return padded
+
+    def get_image_size_for_padding(self, img_width: int,
+                                   img_height: int) -> tuple[int, int]:
+        ratio = img_width / img_height
+        if min(img_height, img_width) < 32 and (ratio > 4 or ratio < 1 / 4):
+            new_size = max(img_height, img_width)
+            return new_size, new_size
+        return img_width, img_height
+
+    def get_image_size_for_preprocess(self, img_width: int,
+                                      img_height: int) -> tuple[int, int]:
+
+        if max(img_height, img_width) > MAX_IMAGE_SIZE:
+            scale_factor = MAX_IMAGE_SIZE / max(img_height, img_width)
+            img_width = int(img_width * scale_factor)
+            img_height = int(img_height * scale_factor)
+        return img_width, img_height
+
+    def get_image_size_for_crop(self, img_width: int, img_height: int,
+                                window_size: int):
+        w_ratio = img_width / window_size
+        h_ratio = img_height / window_size
+
+        if w_ratio < 1:
+            width_new = img_width
+        else:
+            decimal_w = w_ratio - img_width // window_size
+            w_ratio = int(w_ratio) + 1 if decimal_w > 0.2 else int(w_ratio)
+            width_new = window_size * w_ratio
+        if h_ratio < 1:
+            height_new = img_height
+        else:
+            decimal_h = h_ratio - img_height // window_size
+            h_ratio = int(h_ratio) + 1 if decimal_h > 0.2 else int(h_ratio)
+            height_new = window_size * h_ratio
+        return int(width_new), int(height_new)
+
+    def patch_crop(self, img: Image.Image, i: int, j: int, th: int, tw: int):
+        target = img.crop((j, i, j + tw, i + th))
+        return target
+
+    def get_num_patches(self, img_width: int,
+                        img_height: int) -> tuple[int, int]:
+        img_width, img_height = self.get_image_size_for_padding(
+            img_width, img_height)
+        img_width, img_height = self.get_image_size_for_preprocess(
+            img_width, img_height)
+        window_size = self.determine_window_size(max(img_height, img_width),
+                                                 min(img_height, img_width))
+        if window_size == 0:
+            return 0, 0
+        else:
+            img_width, img_height = self.get_image_size_for_crop(
+                img_width, img_height, window_size)
+            center_list, (x_num, y_num) = self.slide_window(
+                img_width, img_height, [(window_size, window_size)],
+                [(window_size, window_size)])
+            full_rows = (len(center_list) - 1) // x_num + 1
+            if len(center_list) > 0 and len(center_list) % x_num == 0:
+                full_rows -= 1
+            return len(center_list), full_rows
+
+    def __call__(
+        self, img: Image.Image
+    ) -> tuple[Image.Image, list[Image.Image], list[bool] | None]:
+        img_width, img_height = img.size
+        new_img_width, new_img_height = self.get_image_size_for_padding(
+            img_width, img_height)
+        if new_img_width != img_width or new_img_height != img_height:
+            img = self.square_pad(img)
+            img_width, img_height = img.size
+
+        new_img_width, new_img_height = self.get_image_size_for_preprocess(
+            img_width, img_height)
+        img = img.resize((new_img_width, new_img_height),
+                         Image.Resampling.BILINEAR)
+        window_size = self.determine_window_size(
+            max(new_img_height, new_img_width),
+            min(new_img_height, new_img_width))
+        # return img, [], None
+        if window_size == 0:
+            return img, [], None
+        else:
+            new_img_width, new_img_height = self.get_image_size_for_crop(
+                new_img_width, new_img_height, window_size)
+            if (new_img_width, new_img_height) != (img_width, img_height):
+                img_for_crop = img.resize((new_img_width, new_img_height),
+                                          Image.Resampling.BILINEAR)
+            else:
+                img_for_crop = img
+
+            patches = []
+            newlines = []
+            center_list, (x_num, y_num) = self.slide_window(
+                new_img_width, new_img_height, [(window_size, window_size)],
+                [(window_size, window_size)])
+            for patch_id, center_lf_point in enumerate(center_list):
+                x, y, patch_w, patch_h = center_lf_point
+                big_patch = self.patch_crop(img_for_crop, y, x, patch_h,
+                                            patch_w)
+                patches.append(big_patch)
+                if (patch_id + 1) % x_num == 0:
+                    newlines.append(patch_id)
+
+            if newlines and newlines[-1] == len(patches) - 1:
+                newlines.pop()
+
+            return img, patches, [i in newlines for i in range(len(patches))] if len(patches) > 0 else None
+
+
+
+
+class Step3VLProcessor(ProcessorMixin):
+    # Align ProcessorMixin with our custom components.
+    # We only have an image processor (not a feature extractor) plus a tokenizer.
+    attributes = ["tokenizer"]
+    tokenizer_class = "AutoTokenizer"
+
+    @classmethod
+    def _load_tokenizer_from_pretrained(
+        cls, sub_processor_type, pretrained_model_name_or_path, subfolder="", **kwargs
+    ):
+        return TokenizersBackend.from_pretrained(
+            pretrained_model_name_or_path,
+            subfolder=subfolder,
+            **kwargs,
+        )
+
+    def __init__(
+        self,
+        tokenizer=None,
+        chat_template=None,
+        **kwargs
+    ) -> None:
+        self.image_size = 728
+        self.patch_size = 504
+
+        self.image_preprocessor = Step3VisionProcessor(self.image_size,
+                                                       "bilinear",
+                                                       self.patch_size)
+
+        self.num_image_feature_size = 169
+        self.num_patch_feature_size = 81
+        self.image_token = "<im_patch>"
+        self.image_feature_placeholder = (self.image_token *
+                                          self.num_image_feature_size)
+        self.patch_feature_placeholder = (self.image_token *
+                                          self.num_patch_feature_size)
+        super().__init__(tokenizer=tokenizer, chat_template=chat_template, **kwargs)
+        self.patcher = ImagePatcher()
+        
+    @property
+    def image_token_id(self) -> int:
+        return self.tokenizer.get_vocab()[self.image_token]
+
+    def get_num_image_tokens(self, img_width: int, img_height: int) -> int:
+        num_patches, num_newlines = self.patcher.get_num_patches(
+            img_width, img_height)
+
+        return num_patches * (
+            self.num_patch_feature_size +
+            2) + self.num_image_feature_size + 2 + num_newlines
+
+    def _split_images(self,
+                      images: list[Image.Image]) -> list[ImageWithPatches]:
+        result = []
+        for img in images:
+            result.append(self.patcher(img))
+        return result
+
+    def _convert_images_to_pixel_values(
+        self,
+        images: list[Image.Image],
+        is_patch: bool = False,
+    ) -> list[torch.Tensor]:
+        return [
+            self.image_preprocessor(img, is_patch=is_patch)["pixel_values"]
+            for img in images
+        ]
+
+    def _get_patch_repl(
+        self,
+        num_patches: int,
+        patch_newline_mask: list[bool] | None,
+    ) -> tuple[str, list[int]]:
+        text = ""
+        token_ids = []
+        for i in range(num_patches):
+            assert len(patch_newline_mask) == num_patches
+            text += f"<patch_start>{self.patch_feature_placeholder}<patch_end>"
+            token_ids.extend(
+                [self.tokenizer.convert_tokens_to_ids("<patch_start>")] +
+                [self.image_token_id] * self.num_patch_feature_size +
+                [self.tokenizer.convert_tokens_to_ids("<patch_end>")])
+            if patch_newline_mask and patch_newline_mask[i]:
+                text += "<patch_newline>"
+                token_ids.append(
+                    self.tokenizer.convert_tokens_to_ids("<patch_newline>"))
+        return text, token_ids
+
+    def _get_image_repl(
+        self,
+        num_images: int,
+    ) -> tuple[str, list[int]]:
+        text = f"<im_start>{self.image_feature_placeholder}<im_end>"
+        token_ids = [
+            self.tokenizer.convert_tokens_to_ids("<im_start>")
+        ] + [self.image_token_id] * self.num_image_feature_size + [
+            self.tokenizer.convert_tokens_to_ids("<im_end>")
+        ]
+        return text * num_images, token_ids * num_images
+
+    def _get_image_repl_features(
+        self,
+        num_images: int,
+        num_patches: int,
+        patch_new_line_idx: Optional[list[bool]],
+    ) -> tuple[str, list[int]]:
+        if num_patches > 0:
+            patch_repl, patch_repl_ids = self._get_patch_repl(
+                num_patches, patch_new_line_idx)
+        else:
+            patch_repl = ""
+            patch_repl_ids = []
+        image_repl, image_repl_ids = self._get_image_repl(num_images)
+        return patch_repl + image_repl, patch_repl_ids + image_repl_ids
+
+    def replace_placeholder(self, text: str, placeholder: str,
+                            repls: list[str]) -> str:
+        parts = text.split(placeholder)
+
+        if len(parts) - 1 != len(repls):
+            raise ValueError(
+                "The number of placeholders does not match the number of replacements."  # noqa: E501
+            )
+
+        result = [parts[0]]
+        for i, repl in enumerate(repls):
+            result.append(repl)
+            result.append(parts[i + 1])
+
+        return "".join(result)
+
+    def __call__(
+        self,
+        text: Optional[Union[str, list[str]]] = None,
+        images: ImageInput | None = None,
+        return_tensors: Optional[Union[str, TensorType]] = None,
+        **kwargs,
+    ) -> BatchFeature:
+
+        if images is not None:
+            images = self.image_preprocessor.fetch_images(images)
+        if text is None:
+            text = []
+        if not isinstance(text, list):
+            text = [text]
+        if images is None:
+            images = []
+        elif not isinstance(images, list):
+            images = [images]
+        elif isinstance(images[0], list):
+            images = images[0]
+
+        if len(images) == 0:
+            image_inputs = {}
+            text_inputs = self.tokenizer(text)
+        else:
+            splitted_images_data = self._split_images(images)
+            pixel_values_lst = []
+            patch_pixel_values_lst = []
+            patch_newline_mask_lst = []
+            image_repl_str_lst = []
+            image_repl_ids_lst = []
+            num_patches = []
+            for raw_img, img_patches, patch_newline_mask in splitted_images_data:  # noqa: E501
+                pixel_values_lst.extend(
+                    self._convert_images_to_pixel_values([raw_img]))
+
+                if len(img_patches) > 0:
+                    patch_pixel_values_lst.extend(
+                        self._convert_images_to_pixel_values(img_patches,
+                                                             is_patch=True))
+                num_patches.append(len(img_patches))
+
+                image_repl_str, image_repl_ids = self._get_image_repl_features(
+                    1, len(img_patches), patch_newline_mask)
+                image_repl_str_lst.append(image_repl_str)
+                image_repl_ids_lst.extend(image_repl_ids)
+
+                if patch_newline_mask is not None:
+                    patch_newline_mask_lst.extend(patch_newline_mask)
+
+            image_inputs = {
+                "pixel_values": torch.cat(pixel_values_lst),
+                "num_patches": num_patches,
+            }
+            if patch_pixel_values_lst:
+                image_inputs["patch_pixel_values"] = torch.cat(
+                    patch_pixel_values_lst)
+            if patch_newline_mask_lst:
+                image_inputs["patch_newline_mask"] = torch.tensor(
+                    patch_newline_mask_lst, dtype=torch.bool)
+
+            text = [
+                self.replace_placeholder(t, self.image_token,
+                                         image_repl_str_lst) for t in text
+            ]
+            text_inputs = self.tokenizer(text)
+
+        return BatchFeature(
+            {
+                **text_inputs,
+                **image_inputs,
+            },
+            tensor_type=return_tensors,
+        )
+        
+    # Copied from transformers.models.clip.processing_clip.CLIPProcessor.batch_decode with CLIP->Gemma
+    def batch_decode(self, *args, **kwargs):
+        """
+        This method forwards all its arguments to GemmaTokenizerFast's [`~PreTrainedTokenizer.batch_decode`]. Please
+        refer to the docstring of this method for more information.
+        """
+        return self.tokenizer.batch_decode(*args, **kwargs)
+
+    # Copied from transformers.models.clip.processing_clip.CLIPProcessor.decode with CLIP->Gemma
+    def decode(self, *args, **kwargs):
+        """
+        This method forwards all its arguments to GemmaTokenizerFast's [`~PreTrainedTokenizer.decode`]. Please refer to
+        the docstring of this method for more information.
+        """
+        return self.tokenizer.decode(*args, **kwargs)
+        
+__all__ = ["Step3VLProcessor"]

special_tokens_map.json

@@ -0,0 +1,23 @@
+{
+  "bos_token": {
+    "content": "<｜begin▁of▁sentence｜>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<｜end▁of▁sentence｜>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

step3p7_mlx.py

@@ -0,0 +1,39 @@
+"""Local MLX bridge for Step3p7 bundles.
+
+The current MLX runtime has a Step3p5 text model but no Step3p7 VLM wrapper.
+For text coherence proof, this module exposes a custom ``model_file`` that
+loads the nested ``text_config`` through ``mlx_lm.models.step3p5`` and drops
+vision tensors during sanitize. Vision runtime remains a separate follow-up.
+"""
+
+from __future__ import annotations
+
+from dataclasses import dataclass
+
+from mlx_lm.models import step3p5
+
+
+@dataclass
+class ModelArgs(step3p5.ModelArgs):
+    @classmethod
+    def from_dict(cls, params):
+        text_config = dict(params.get("text_config") or params)
+        text_config["model_type"] = "step3p5"
+        return super().from_dict(text_config)
+
+
+class Model(step3p5.Model):
+    def sanitize(self, weights):
+        text_weights = {}
+        prefix = "model.language_model."
+        for key, value in weights.items():
+            if key.startswith(prefix):
+                text_weights["model." + key[len(prefix):]] = value
+            elif key.startswith("model.vision_model.") or key.startswith("model.vit_large_projector."):
+                continue
+            elif key.startswith("vision_model.") or key.startswith("vit_large_projector."):
+                continue
+            else:
+                text_weights[key] = value
+        return super().sanitize(text_weights)
+

vision_encoder.py

@@ -0,0 +1,452 @@
+from typing import Literal, Optional, Tuple, Union
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from transformers.activations import ACT2FN
+
+
+from .configuration_step3p7 import StepRoboticsVisionEncoderConfig
+
+
+
+def rotate_half(x: torch.Tensor) -> torch.Tensor:
+    """Rotate last dimension halves (used by RoPE)."""
+    x = x.reshape(*x.shape[:-1], -1, 2)
+    x1, x2 = x.unbind(dim=-1)
+    x = torch.stack((-x2, x1), dim=-1)
+    return x.reshape(*x.shape[:-2], -1)
+
+
+def apply_rotary_emb(freqs: torch.Tensor,
+                     t: torch.Tensor,
+                     start_index: int = 0,
+                     scale: float = 1.0,
+                     seq_dim: int = -2) -> torch.Tensor:
+    """Apply 2D rotary embeddings to queries / keys."""
+    dtype = t.dtype
+
+    if t.ndim == 3:
+        seq_len = t.shape[seq_dim]
+        freqs = freqs[-seq_len:]
+
+    rot_dim = freqs.shape[-1]
+    end_index = start_index + rot_dim
+    assert rot_dim <= t.shape[-1], (
+        f"feature dimension {t.shape[-1]} is too small for rot_dim {rot_dim}")
+
+    t_left, t, t_right = (
+        t[..., :start_index],
+        t[..., start_index:end_index],
+        t[..., end_index:],
+    )
+    t = (t * freqs.cos() * scale) + (rotate_half(t) * freqs.sin() * scale)
+    out = torch.cat((t_left, t, t_right), dim=-1)
+    return out.type(dtype)
+
+
+class EncoderRope2D(nn.Module):
+    """Cacheable 2D rotary positional embedding."""
+
+    def __init__(
+        self,
+        dim: int,
+        max_grid_height: int,
+        max_grid_width: int,
+        use_cls_token: bool = False,
+        theta: Union[int, float] = 10000,
+        max_freq: int = 10,
+        num_freqs: int = 1,
+        theta_rescale_factor: float = 1.0,
+    ):
+        super().__init__()
+        self.dim = dim
+        self.max_grid_height = max_grid_height
+        self.max_grid_width = max_grid_width
+        self.use_cls_token = use_cls_token
+        self.theta = theta * theta_rescale_factor**(dim / (dim - 2))
+        self.max_freq = max_freq
+        self.num_freqs = num_freqs
+        cache = self._compute_2d_freqs()
+        self.register_buffer("freqs_cache", cache, persistent=False)
+
+    def _compute_inv_freq(self, base: Union[int, float],
+                          dim: int) -> torch.Tensor:
+
+        freqs = 1.0 / (base**(
+            torch.arange(0, dim, 2)[:(dim // 2)].float() / dim))
+        return freqs
+
+    def _compute_freqs(self, t: torch.Tensor, inv_freq: torch.Tensor):
+        freqs = torch.einsum("..., f -> ... f", t.type(inv_freq.dtype),
+                             inv_freq)
+        freqs = freqs.repeat_interleave(2, dim=-1)
+        return freqs
+
+    def _compute_2d_freqs(self) -> torch.Tensor:
+        grid_h_range = torch.arange(self.max_grid_height, dtype=torch.float)
+        grid_w_range = torch.arange(self.max_grid_width, dtype=torch.float)
+        if self.use_cls_token:
+            grid_h_range += 1
+            grid_w_range += 1
+        inv_freq = self._compute_inv_freq(self.theta, self.dim // 2)
+        freqs_h = self._compute_freqs(grid_h_range, inv_freq)[:, None].expand(
+            self.max_grid_height, self.max_grid_width, -1)
+        freqs_w = self._compute_freqs(grid_w_range, inv_freq)[None, :].expand(
+            self.max_grid_height, self.max_grid_width, -1)
+        freqs = torch.cat([freqs_w, freqs_h], dim=-1).reshape(
+            self.max_grid_height * self.max_grid_width, -1)
+        if self.use_cls_token:
+            freqs = torch.cat([torch.zeros(1, freqs.shape[-1]), freqs], dim=0)
+        freqs = freqs[None, None, ...]
+        return freqs
+
+    def forward(self, q: torch.Tensor, k: torch.Tensor,
+                grid_hw: tuple[int, int]):
+        # If grid matches cached shape we reuse directly to avoid recomputation.
+        if grid_hw[0] != self.max_grid_height or grid_hw[1] != self.max_grid_width:
+            rows = torch.arange(grid_hw[0], device=q.device).view(-1, 1)
+            cols = torch.arange(grid_hw[1], device=q.device).view(1, -1)
+            positions = (rows * self.max_grid_width + cols).reshape(-1).to(
+                torch.long)
+            if self.use_cls_token:
+                positions = torch.cat(
+                    [torch.zeros(1, device=q.device), positions + 1], dim=0)
+            freqs = self.freqs_cache.index_select(2, positions)
+        else:
+            freqs = self.freqs_cache
+        q = apply_rotary_emb(freqs, q)
+        k = apply_rotary_emb(freqs, k)
+        return q, k
+
+
+class EncoderLayerScale(nn.Module):
+    """Per-channel residual scaling used when ls_init_value is set."""
+
+    def __init__(self, dim: int, init_values: float):
+        super().__init__()
+        self.gamma = nn.Parameter(torch.full((dim,), init_values))
+
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:  # (B, L, D)
+        return hidden_states * self.gamma
+
+
+class EncoderMLP(nn.Module):
+    """Feed-forward network used inside each transformer block."""
+
+    def __init__(self, hidden_size: int, intermediate_size: int,
+                 hidden_act: str):
+        super().__init__()
+        self.c_fc = nn.Linear(hidden_size, intermediate_size, bias=True)
+        self.act_fn = ACT2FN[hidden_act]
+        self.c_proj = nn.Linear(intermediate_size, hidden_size, bias=True)
+
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+
+        hidden_states = self.c_proj(self.act_fn(self.c_fc(hidden_states)))
+        return hidden_states
+
+
+class EncoderVisionAttention(nn.Module):
+    """Multi-head self attention with optional 2D RoPE."""
+
+    def __init__(
+        self,
+        hidden_size: int,
+        num_heads: int,
+        max_grid_height: int,
+        max_grid_width: int,
+        use_cls_token: bool = False,
+        use_rope2d: bool = True,
+        rope_theta: Union[int, float] = 10000,
+        rope_max_freq: int = 10,
+        rope_num_freqs: int = 1,
+        rope_theta_rescale_factor: float = 1.0,
+        rope_freqs_for: Literal["lang", "pixel", "constant"] = "lang",
+    ):
+        super().__init__()
+        if hidden_size % num_heads != 0:
+            raise ValueError(
+                f"hidden_size ({hidden_size}) must be divisible by num_heads ({num_heads})."
+            )
+        self.num_heads = num_heads
+        self.head_dim = hidden_size // num_heads
+        self.scale = self.head_dim**-0.5
+        self.in_proj_weight = nn.Parameter(torch.zeros(hidden_size * 3, hidden_size))
+        self.in_proj_bias = nn.Parameter(torch.zeros(hidden_size * 3))
+        self.out_proj = nn.Linear(hidden_size, hidden_size, bias=True)
+
+        self.rope = None
+        if use_rope2d:
+            self.rope = EncoderRope2D(
+                dim=self.head_dim,
+                max_grid_height=max_grid_height,
+                max_grid_width=max_grid_width,
+                use_cls_token=use_cls_token,
+                theta=rope_theta,
+                max_freq=rope_max_freq,
+                num_freqs=rope_num_freqs,
+                theta_rescale_factor=rope_theta_rescale_factor,
+            )
+
+    def forward(self, hidden_states: torch.Tensor, grid_hw: tuple[int, int]) -> torch.Tensor:
+        bsz, seq_len, _ = hidden_states.shape
+        qkv = F.linear(
+            hidden_states,
+            self.in_proj_weight,
+            self.in_proj_bias,
+        )
+        q, k, v = qkv.chunk(3, dim=-1)
+
+        q = q.view(bsz, seq_len, self.num_heads,
+                   self.head_dim).transpose(1, 2)
+        k = k.view(bsz, seq_len, self.num_heads,
+                   self.head_dim).transpose(1, 2)
+        if self.rope is not None:
+            q, k = self.rope(q, k, grid_hw=grid_hw)
+        v = v.view(bsz, seq_len, self.num_heads,
+                   self.head_dim).transpose(1, 2)
+
+        attn_output = F.scaled_dot_product_attention(
+            q, k, v, is_causal=False, scale=self.scale)
+        attn_output = attn_output.transpose(1, 2).reshape(
+            bsz, seq_len, self.num_heads * self.head_dim)
+        return self.out_proj(attn_output)
+
+
+class EncoderVisionBlock(nn.Module):
+    """A single Vision Transformer block (self-attention + MLP)."""
+
+    def __init__(
+        self,
+        hidden_size: int,
+        num_heads: int,
+        mlp_ratio: float,
+        hidden_act: str,
+        layer_norm_eps: float,
+        ls_init_value: Optional[float] = None,
+        max_grid_height: Optional[int] = None,
+        max_grid_width: Optional[int] = None,
+        use_cls_token: bool = False,
+        use_rope2d: bool = True,
+        rope_kwargs: Optional[dict] = None,
+    ):
+        super().__init__()
+        rope_kwargs = rope_kwargs or {}
+        self.attn = EncoderVisionAttention(
+            hidden_size,
+            num_heads,
+            max_grid_height=max_grid_height,
+            max_grid_width=max_grid_width,
+            use_cls_token=use_cls_token,
+            use_rope2d=use_rope2d,
+            **rope_kwargs,
+        )
+        self.ln_1 = nn.LayerNorm(hidden_size, eps=layer_norm_eps)
+        self.ln_2 = nn.LayerNorm(hidden_size, eps=layer_norm_eps)
+
+        intermediate = int(hidden_size * mlp_ratio)
+        self.mlp = EncoderMLP(hidden_size, intermediate, hidden_act)
+
+        self.ls_1 = EncoderLayerScale(hidden_size, ls_init_value)
+        self.ls_2 = EncoderLayerScale(hidden_size, ls_init_value)
+
+    def forward(self, hidden_states: torch.Tensor,
+                grid_hw: tuple[int, int]) -> torch.Tensor:
+        # breakpoint()
+        residual = hidden_states
+        hidden_states = self.ln_1(hidden_states)
+        hidden_states = self.attn(hidden_states, grid_hw=grid_hw)
+        hidden_states = residual + self.ls_1(hidden_states)
+
+        residual = hidden_states
+        hidden_states = self.ln_2(hidden_states)
+        hidden_states = self.mlp(hidden_states)
+        hidden_states = residual + self.ls_2(hidden_states)
+        return hidden_states
+
+
+class EncoderVisionTransformer(nn.Module):
+    """Stack of encoder blocks parameterised by Step35VisionEncoderConfig."""
+
+    def __init__(
+        self,
+        embed_dim: int,
+        depth: int,
+        num_heads: int,
+        mlp_ratio: float,
+        hidden_act: str,
+        layer_norm_eps: float,
+        ls_init_value: Optional[float] = None,
+        max_grid_height: Optional[int] = None,
+        max_grid_width: Optional[int] = None,
+        use_cls_token: bool = False,
+        use_rope2d: bool = True,
+        rope_kwargs: Optional[dict] = None,
+    ):
+        super().__init__()
+        self.layers = depth
+        rope_kwargs = rope_kwargs or {}
+        self.resblocks = nn.ModuleList([
+            EncoderVisionBlock(embed_dim, num_heads, mlp_ratio, hidden_act,
+                               layer_norm_eps,
+                               max_grid_height=max_grid_height,
+                               max_grid_width=max_grid_width,
+                               use_cls_token=use_cls_token,
+                               use_rope2d=use_rope2d,
+                               ls_init_value=ls_init_value,
+                               rope_kwargs=rope_kwargs)
+            for _ in range(depth)
+        ])
+
+    def forward(self,
+                hidden_states: torch.Tensor,
+                grid_hw: tuple[int, int]) -> torch.Tensor:
+        for block in self.resblocks:
+            hidden_states = block(hidden_states, grid_hw=grid_hw)
+        return hidden_states
+
+
+class StepRoboticsVisionEncoder(nn.Module):
+    """
+    Vision encoder built from StepRoboticsVisionEncoderConfig.
+
+    The encoder performs patch embedding followed by a stack of transformer
+    blocks. Only the config fields defined in StepRoboticsVisionEncoderConfig (and
+    StepRoboticVLConfig.vision_config) are expected.
+    """
+
+    def __init__(self, config: StepRoboticsVisionEncoderConfig):
+        super().__init__()
+        self.config = config
+
+        # Align commonly used attributes so downstream code (e.g. StepRoboticVL)
+        # can access them without extra renaming.
+        self.hidden_size = config.width
+        self.num_heads = config.heads
+        self.num_hidden_layers = config.layers
+        self.patch_size = config.patch_size
+        self.image_size = config.image_size
+        self.use_cls_token = getattr(config, "use_cls_token", False)
+        self.use_rope2d = getattr(config, "use_rope2d", True)
+        self.use_abs_posemb = getattr(config, "use_abs_posemb", True)
+        self.layer_norm_eps = config.layer_norm_eps
+        self.mlp_ratio = getattr(config, "mlp_ratio", 8960 / 1536)
+        self.ls_init_value = getattr(config, "ls_init_value", None)
+        self.hidden_act = config.hidden_act
+        self.use_ln_pre = getattr(config, "use_ln_pre", False)
+        self.use_ln_post = getattr(config, "use_ln_post", True)
+
+        # Patch embedding.
+        self.conv1 = nn.Conv2d(in_channels=config.num_channels,
+                               out_channels=self.hidden_size,
+                               kernel_size=self.patch_size,
+                               stride=self.patch_size,
+                               bias=False)
+
+        self.ln_pre = nn.LayerNorm(self.hidden_size, eps=self.layer_norm_eps) if self.use_ln_pre else nn.Identity()
+        self.ln_post =  nn.LayerNorm(self.hidden_size, eps=self.layer_norm_eps) if self.use_ln_post else nn.Identity()
+
+        grid_size = self.image_size // self.patch_size
+        self.base_grid = (grid_size, grid_size)
+
+        if self.use_cls_token:
+            self.class_embedding = nn.Parameter(
+                torch.randn(self.hidden_size) * (self.hidden_size**-0.5))
+        else:
+            self.class_embedding = None
+        
+        if self.use_abs_posemb:
+            self.posemb_grid_size = self.image_size // self.patch_size
+            self.positional_embedding = nn.Parameter(
+                (self.hidden_size**-0.5) * torch.randn(
+                    int(self.use_cls_token) + self.posemb_grid_size**2,
+                    self.hidden_size,
+                ))
+
+        self.transformer = EncoderVisionTransformer(
+            embed_dim=self.hidden_size,
+            depth=self.num_hidden_layers,
+            num_heads=self.num_heads,
+            mlp_ratio=self.mlp_ratio,
+            hidden_act=self.hidden_act,
+            layer_norm_eps=self.layer_norm_eps,
+            ls_init_value=self.ls_init_value,
+            max_grid_height=self.base_grid[0],
+            max_grid_width=self.base_grid[1],
+            use_cls_token=self.use_cls_token,
+            use_rope2d=self.use_rope2d,
+            rope_kwargs={
+                "rope_theta": getattr(config, "rope_theta", 10000),
+                "rope_max_freq": getattr(config, "rope_max_freq", 10),
+                "rope_num_freqs": getattr(config, "rope_num_freqs", 1),
+                "rope_theta_rescale_factor":
+                getattr(config, "rope_theta_rescale_factor", 1.0),
+                "rope_freqs_for": getattr(config, "rope_freqs_for", "lang"),
+            },
+        )
+        self.vit_downsampler1 = nn.Conv2d(self.hidden_size,
+                                          self.hidden_size * 2,
+                                          kernel_size=3,
+                                          stride=2,
+                                          padding=1)
+        self.vit_downsampler2 = nn.Conv2d(self.hidden_size * 2,
+                                          self.hidden_size * 4,
+                                          kernel_size=3,
+                                          stride=2,
+                                          padding=1)
+
+
+    def sample_abs_posemb(self, grid_h: int, grid_w: int):
+        if self.posemb_grid_size == grid_h and self.posemb_grid_size == grid_w:
+            return self.positional_embedding[None, ...]
+
+        pos_embed = self.positional_embedding
+        if self.use_cls_token:
+            cls_token_embed, pos_embed = pos_embed[:1], pos_embed[1:]
+
+        pos_embed = (pos_embed.reshape(1, self.posemb_grid_size,
+                                       self.posemb_grid_size,
+                                       -1).permute(0, 3, 1, 2).contiguous())
+        pos_embed = F.interpolate(pos_embed,
+                                  size=(grid_h, grid_w),
+                                  mode="bilinear",
+                                  align_corners=False)
+        pos_embed = pos_embed.permute(0, 2, 3, 1).reshape(-1, self.hidden_size)
+
+        if self.use_cls_token:
+            pos_embed = torch.cat([cls_token_embed, pos_embed], dim=0)
+
+        return pos_embed[None, ...]
+
+    def forward(self, pixel_values: torch.Tensor) -> torch.Tensor:
+        """
+        Args:
+            pixel_values: Image tensor of shape (B, C, H, W).
+            layer_idx: Negative indices stop after a given block (e.g., -1 uses all blocks).
+            strip_cls_token: If True and cls token is used, remove it from output.
+        """
+        bsz, _, height, width = pixel_values.shape
+        grid_h, grid_w = height // self.patch_size, width // self.patch_size
+
+        hidden_state = self.conv1(pixel_values)  # (B, D, Gh, Gw)
+        hidden_state = hidden_state.flatten(2).transpose(1, 2)  # (B, Gh*Gw, D)
+
+        if self.use_cls_token:
+            cls_token = self.class_embedding.view(1, 1,
+                                                  -1).expand(bsz, -1, -1)
+            hidden_state = torch.cat([cls_token, hidden_state], dim=1)
+
+        if self.use_abs_posemb:
+            pos_emb = self.sample_abs_posemb(grid_h, grid_w)
+            hidden_state = hidden_state + pos_emb
+        hidden_state = self.ln_pre(hidden_state)
+        hidden_state = self.transformer(hidden_state, grid_hw=(grid_h, grid_w))
+
+        if self.use_ln_post:
+            hidden_state = self.ln_post(hidden_state)
+
+        if self.use_cls_token:
+            hidden_state = hidden_state[:, 1:, :]
+
+        return hidden_state