Add files using upload-large-folder tool

README.md

@@ -1,199 +1,168 @@
 ---
+license: apache-2.0
+pipeline_tag: text-generation
 library_name: transformers
-tags: []
+tags:
+- vllm
 ---
 
-# Model Card for Model ID
+<p align="center">
+  <img alt="gpt-oss-20b" src="https://raw.githubusercontent.com/openai/gpt-oss/main/docs/gpt-oss-20b.svg">
+</p>
 
-<!-- Provide a quick summary of what the model is/does. -->
+<p align="center">
+  <a href="https://gpt-oss.com"><strong>Try gpt-oss</strong></a> ·
+  <a href="https://cookbook.openai.com/topic/gpt-oss"><strong>Guides</strong></a> ·
+  <a href="https://openai.com/index/gpt-oss-model-card"><strong>Model card</strong></a> ·
+  <a href="https://openai.com/index/introducing-gpt-oss/"><strong>OpenAI blog</strong></a>
+</p>
 
+<br>
 
+Welcome to the gpt-oss series, [OpenAI’s open-weight models](https://openai.com/open-models) designed for powerful reasoning, agentic tasks, and versatile developer use cases.
 
-## Model Details
+We’re releasing two flavors of these open models:
+- `gpt-oss-120b` — for production, general purpose, high reasoning use cases that fit into a single H100 GPU (117B parameters with 5.1B active parameters)
+- `gpt-oss-20b` — for lower latency, and local or specialized use cases (21B parameters with 3.6B active parameters)
 
-### Model Description
+Both models were trained on our [harmony response format](https://github.com/openai/harmony) and should only be used with the harmony format as it will not work correctly otherwise.
 
-<!-- Provide a longer summary of what this model is. -->
 
-This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
+> [!NOTE]
+> This model card is dedicated to the smaller `gpt-oss-20b` model. Check out [`gpt-oss-120b`](https://huggingface.co/openai/gpt-oss-120b) for the larger model.
 
-- **Developed by:** [More Information Needed]
-- **Funded by [optional]:** [More Information Needed]
-- **Shared by [optional]:** [More Information Needed]
-- **Model type:** [More Information Needed]
-- **Language(s) (NLP):** [More Information Needed]
-- **License:** [More Information Needed]
-- **Finetuned from model [optional]:** [More Information Needed]
+# Highlights
 
-### Model Sources [optional]
+* **Permissive Apache 2.0 license:** Build freely without copyleft restrictions or patent risk—ideal for experimentation, customization, and commercial deployment.  
+* **Configurable reasoning effort:** Easily adjust the reasoning effort (low, medium, high) based on your specific use case and latency needs.  
+* **Full chain-of-thought:** Gain complete access to the model’s reasoning process, facilitating easier debugging and increased trust in outputs. It’s not intended to be shown to end users.  
+* **Fine-tunable:** Fully customize models to your specific use case through parameter fine-tuning.
+* **Agentic capabilities:** Use the models’ native capabilities for function calling, [web browsing](https://github.com/openai/gpt-oss/tree/main?tab=readme-ov-file#browser), [Python code execution](https://github.com/openai/gpt-oss/tree/main?tab=readme-ov-file#python), and Structured Outputs.
+* **Native MXFP4 quantization:** The models are trained with native MXFP4 precision for the MoE layer, making `gpt-oss-120b` run on a single H100 GPU and the `gpt-oss-20b` model run within 16GB of memory.
 
-<!-- Provide the basic links for the model. -->
-
-- **Repository:** [More Information Needed]
-- **Paper [optional]:** [More Information Needed]
-- **Demo [optional]:** [More Information Needed]
-
-## Uses
-
-<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
-
-### Direct Use
-
-<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
-
-[More Information Needed]
-
-### Downstream Use [optional]
-
-<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
-
-[More Information Needed]
-
-### Out-of-Scope Use
-
-<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
-
-[More Information Needed]
-
-## Bias, Risks, and Limitations
-
-<!-- This section is meant to convey both technical and sociotechnical limitations. -->
-
-[More Information Needed]
-
-### Recommendations
-
-<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
-
-Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
-
-## How to Get Started with the Model
-
-Use the code below to get started with the model.
-
-[More Information Needed]
-
-## Training Details
-
-### Training Data
-
-<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
-
-[More Information Needed]
-
-### Training Procedure
-
-<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
-
-#### Preprocessing [optional]
-
-[More Information Needed]
-
-
-#### Training Hyperparameters
-
-- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
-
-#### Speeds, Sizes, Times [optional]
-
-<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
-
-[More Information Needed]
-
-## Evaluation
-
-<!-- This section describes the evaluation protocols and provides the results. -->
-
-### Testing Data, Factors & Metrics
-
-#### Testing Data
-
-<!-- This should link to a Dataset Card if possible. -->
-
-[More Information Needed]
+---
 
-#### Factors
+# Inference examples
 
-<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+## Transformers
 
-[More Information Needed]
+You can use `gpt-oss-120b` and `gpt-oss-20b` with Transformers. If you use the Transformers chat template, it will automatically apply the [harmony response format](https://github.com/openai/harmony). If you use `model.generate` directly, you need to apply the harmony format manually using the chat template or use our [openai-harmony](https://github.com/openai/harmony) package.
 
-#### Metrics
+To get started, install the necessary dependencies to setup your environment:
 
-<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+```
+pip install -U transformers kernels torch 
+```
 
-[More Information Needed]
+Once, setup you can proceed to run the model by running the snippet below:
 
-### Results
+```py
+from transformers import pipeline
+import torch
 
-[More Information Needed]
+model_id = "openai/gpt-oss-20b"
 
-#### Summary
+pipe = pipeline(
+    "text-generation",
+    model=model_id,
+    torch_dtype="auto",
+    device_map="auto",
+)
 
+messages = [
+    {"role": "user", "content": "Explain quantum mechanics clearly and concisely."},
+]
 
+outputs = pipe(
+    messages,
+    max_new_tokens=256,
+)
+print(outputs[0]["generated_text"][-1])
+```
 
-## Model Examination [optional]
+Alternatively, you can run the model via [`Transformers Serve`](https://huggingface.co/docs/transformers/main/serving) to spin up a OpenAI-compatible webserver:
 
-<!-- Relevant interpretability work for the model goes here -->
+```
+transformers serve
+transformers chat localhost:8000 --model-name-or-path openai/gpt-oss-20b
+```
 
-[More Information Needed]
+[Learn more about how to use gpt-oss with Transformers.](https://cookbook.openai.com/articles/gpt-oss/run-transformers)
 
-## Environmental Impact
+## vLLM
 
-<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+vLLM recommends using [uv](https://docs.astral.sh/uv/) for Python dependency management. You can use vLLM to spin up an OpenAI-compatible webserver. The following command will automatically download the model and start the server.
 
-Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+```bash
+uv pip install --pre vllm==0.10.1+gptoss \
+    --extra-index-url https://wheels.vllm.ai/gpt-oss/ \
+    --extra-index-url https://download.pytorch.org/whl/nightly/cu128 \
+    --index-strategy unsafe-best-match
 
-- **Hardware Type:** [More Information Needed]
-- **Hours used:** [More Information Needed]
-- **Cloud Provider:** [More Information Needed]
-- **Compute Region:** [More Information Needed]
-- **Carbon Emitted:** [More Information Needed]
+vllm serve openai/gpt-oss-20b
+```
 
-## Technical Specifications [optional]
+[Learn more about how to use gpt-oss with vLLM.](https://cookbook.openai.com/articles/gpt-oss/run-vllm)
 
-### Model Architecture and Objective
+## PyTorch / Triton
 
-[More Information Needed]
+To learn about how to use this model with PyTorch and Triton, check out our [reference implementations in the gpt-oss repository](https://github.com/openai/gpt-oss?tab=readme-ov-file#reference-pytorch-implementation).
 
-### Compute Infrastructure
+## Ollama
 
-[More Information Needed]
+If you are trying to run gpt-oss on consumer hardware, you can use Ollama by running the following commands after [installing Ollama](https://ollama.com/download).
 
-#### Hardware
+```bash
+# gpt-oss-20b
+ollama pull gpt-oss:20b
+ollama run gpt-oss:20b
+```
 
-[More Information Needed]
+[Learn more about how to use gpt-oss with Ollama.](https://cookbook.openai.com/articles/gpt-oss/run-locally-ollama)
 
-#### Software
+#### LM Studio
 
-[More Information Needed]
+If you are using [LM Studio](https://lmstudio.ai/) you can use the following commands to download.
 
-## Citation [optional]
+```bash
+# gpt-oss-20b
+lms get openai/gpt-oss-20b
+```
 
-<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+Check out our [awesome list](https://github.com/openai/gpt-oss/blob/main/awesome-gpt-oss.md) for a broader collection of gpt-oss resources and inference partners.
 
-**BibTeX:**
+---
 
-[More Information Needed]
+# Download the model
 
-**APA:**
+You can download the model weights from the [Hugging Face Hub](https://huggingface.co/collections/openai/gpt-oss-68911959590a1634ba11c7a4) directly from Hugging Face CLI:
 
-[More Information Needed]
+```shell
+# gpt-oss-20b
+huggingface-cli download openai/gpt-oss-20b --include "original/*" --local-dir gpt-oss-20b/
+pip install gpt-oss
+python -m gpt_oss.chat model/
+```
 
-## Glossary [optional]
+# Reasoning levels
 
-<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+You can adjust the reasoning level that suits your task across three levels:
 
-[More Information Needed]
+* **Low:** Fast responses for general dialogue.  
+* **Medium:** Balanced speed and detail.  
+* **High:** Deep and detailed analysis.
 
-## More Information [optional]
+The reasoning level can be set in the system prompts, e.g., "Reasoning: high".
 
-[More Information Needed]
+# Tool use
 
-## Model Card Authors [optional]
+The gpt-oss models are excellent for:
+* Web browsing (using built-in browsing tools)
+* Function calling with defined schemas
+* Agentic operations like browser tasks
 
-[More Information Needed]
+# Fine-tuning
 
-## Model Card Contact
+Both gpt-oss models can be fine-tuned for a variety of specialized use cases.
 
-[More Information Needed]
+This smaller model `gpt-oss-20b` can be fine-tuned on consumer hardware, whereas the larger [`gpt-oss-120b`](https://huggingface.co/openai/gpt-oss-120b) can be fine-tuned on a single H100 node.

chat_template.json

@@ -0,0 +1,3 @@
+{
+  "chat_template": "{#-\n  In addition to the normal inputs of `messages` and `tools`, this template also accepts the\n  following kwargs:\n  - \"builtin_tools\": A list, can contain \"browser\" and/or \"python\".\n  - \"model_identity\": A string that optionally describes the model identity.\n  - \"reasoning_effort\": A string that describes the reasoning effort, defaults to \"medium\".\n #}\n\n{#- Tool Definition Rendering ============================================== #}\n{%- macro render_typescript_type(param_spec, required_params, is_nullable=false) -%}\n    {%- if param_spec.type == \"array\" -%}\n        {%- if param_spec['items'] -%}\n            {%- if param_spec['items']['type'] == \"string\" -%}\n                {{- \"string[]\" }}\n            {%- elif param_spec['items']['type'] == \"number\" -%}\n                {{- \"number[]\" }}\n            {%- elif param_spec['items']['type'] == \"integer\" -%}\n                {{- \"number[]\" }}\n            {%- elif param_spec['items']['type'] == \"boolean\" -%}\n                {{- \"boolean[]\" }}\n            {%- else -%}\n                {%- set inner_type = render_typescript_type(param_spec['items'], required_params) -%}\n                {%- if inner_type == \"object | object\" or inner_type|length > 50 -%}\n                    {{- \"any[]\" }}\n                {%- else -%}\n                    {{- inner_type + \"[]\" }}\n                {%- endif -%}\n            {%- endif -%}\n            {%- if param_spec.nullable -%}\n                {{- \" | null\" }}\n            {%- endif -%}\n        {%- else -%}\n            {{- \"any[]\" }}\n            {%- if param_spec.nullable -%}\n                {{- \" | null\" }}\n            {%- endif -%}\n        {%- endif -%}\n    {%- elif param_spec.type is defined and param_spec.type is iterable and param_spec.type is not string and param_spec.type is not mapping and param_spec.type[0] is defined -%}\n        {#- Handle array of types like [\"object\", \"object\"] from Union[dict, list] #}\n        {%- if param_spec.type | length > 1 -%}\n            {{- param_spec.type | join(\" | \") }}\n        {%- else -%}\n            {{- param_spec.type[0] }}\n        {%- endif -%}\n    {%- elif param_spec.oneOf -%}\n        {#- Handle oneOf schemas - check for complex unions and fallback to any #}\n        {%- set has_object_variants = false -%}\n        {%- for variant in param_spec.oneOf -%}\n            {%- if variant.type == \"object\" -%}\n                {%- set has_object_variants = true -%}\n            {%- endif -%}\n        {%- endfor -%}\n        {%- if has_object_variants and param_spec.oneOf|length > 1 -%}\n            {{- \"any\" }}\n        {%- else -%}\n            {%- for variant in param_spec.oneOf -%}\n                {{- render_typescript_type(variant, required_params) -}}\n                {%- if variant.description %}\n                    {{- \"// \" + variant.description }}\n                {%- endif -%}\n                {%- if variant.default is defined %}\n                    {{ \"// default: \" + variant.default|tojson }}\n                {%- endif -%}\n                {%- if not loop.last %}\n                    {{- \" | \" }}\n                {% endif -%}\n            {%- endfor -%}\n        {%- endif -%}\n    {%- elif param_spec.type == \"string\" -%}\n        {%- if param_spec.enum -%}\n            {{- '\"' + param_spec.enum|join('\" | \"') + '\"' -}}\n        {%- else -%}\n            {{- \"string\" }}\n            {%- if param_spec.nullable %}\n                {{- \" | null\" }}\n            {%- endif -%}\n        {%- endif -%}\n    {%- elif param_spec.type == \"number\" -%}\n        {{- \"number\" }}\n    {%- elif param_spec.type == \"integer\" -%}\n        {{- \"number\" }}\n    {%- elif param_spec.type == \"boolean\" -%}\n        {{- \"boolean\" }}\n\n    {%- elif param_spec.type == \"object\" -%}\n        {%- if param_spec.properties -%}\n            {{- \"{\n\" }}\n            {%- for prop_name, prop_spec in param_spec.properties.items() -%}\n                {{- prop_name -}}\n                {%- if prop_name not in (param_spec.required or []) -%}\n                    {{- \"?\" }}\n                {%- endif -%}\n                {{- \": \" }}\n                {{ render_typescript_type(prop_spec, param_spec.required or []) }}\n                {%- if not loop.last -%}\n                    {{-\", \" }}\n                {%- endif -%}\n            {%- endfor -%}\n            {{- \"}\" }}\n        {%- else -%}\n            {{- \"object\" }}\n        {%- endif -%}\n    {%- else -%}\n        {{- \"any\" }}\n    {%- endif -%}\n{%- endmacro -%}\n\n{%- macro render_tool_namespace(namespace_name, tools) -%}\n    {{- \"## \" + namespace_name + \"\n\n\" }}\n    {{- \"namespace \" + namespace_name + \" {\n\n\" }}\n    {%- for tool in tools %}\n        {%- set tool = tool.function %}\n        {{- \"// \" + tool.description + \"\n\" }}\n        {{- \"type \"+ tool.name + \" = \" }}\n        {%- if tool.parameters and tool.parameters.properties %}\n            {{- \"(_: {\n\" }}\n            {%- for param_name, param_spec in tool.parameters.properties.items() %}\n                {%- if param_spec.description %}\n                    {{- \"// \" + param_spec.description + \"\n\" }}\n                {%- endif %}\n                {{- param_name }}\n                {%- if param_name not in (tool.parameters.required or []) -%}\n                    {{- \"?\" }}\n                {%- endif -%}\n                {{- \": \" }}\n                {{- render_typescript_type(param_spec, tool.parameters.required or []) }}\n                {%- if param_spec.default is defined -%}\n                    {%- if param_spec.enum %}\n                        {{- \", // default: \" + param_spec.default }}\n                    {%- elif param_spec.oneOf %}\n                        {{- \"// default: \" + param_spec.default }}\n                    {%- else %}\n                        {{- \", // default: \" + param_spec.default|tojson }}\n                    {%- endif -%}\n                {%- endif -%}\n                {%- if not loop.last %}\n                    {{- \",\n\" }}\n                {%- else %}\n                    {{- \"\n\" }}\n                {%- endif -%}\n            {%- endfor %}\n            {{- \"}) => any;\n\n\" }}\n        {%- else -%}\n            {{- \"() => any;\n\n\" }}\n        {%- endif -%}\n    {%- endfor %}\n    {{- \"} // namespace \" + namespace_name }}\n{%- endmacro -%}\n\n{%- macro render_builtin_tools(browser_tool, python_tool) -%}\n    {%- if browser_tool %}\n        {{- \"## browser\n\n\" }}\n        {{- \"// Tool for browsing.\n\" }}\n        {{- \"// The `cursor` appears in brackets before each browsing display: `[{cursor}]`.\n\" }}\n        {{- \"// Cite information from the tool using the following format:\n\" }}\n        {{- \"// `\u3010{cursor}\u2020L{line_start}(-L{line_end})?\u3011`, for example: `\u30106\u2020L9-L11\u3011` or `\u30108\u2020L3\u3011`.\n\" }}\n        {{- \"// Do not quote more than 10 words directly from the tool output.\n\" }}\n        {{- \"// sources=web (default: web)\n\" }}\n        {{- \"namespace browser {\n\n\" }}\n        {{- \"// Searches for information related to `query` and displays `topn` results.\n\" }}\n        {{- \"type search = (_: {\n\" }}\n        {{- \"query: string,\n\" }}\n        {{- \"topn?: number, // default: 10\n\" }}\n        {{- \"source?: string,\n\" }}\n        {{- \"}) => any;\n\n\" }}\n        {{- \"// Opens the link `id` from the page indicated by `cursor` starting at line number `loc`, showing `num_lines` lines.\n\" }}\n        {{- \"// Valid link ids are displayed with the formatting: `\u3010{id}\u2020.*\u3011`.\n\" }}\n        {{- \"// If `cursor` is not provided, the most recent page is implied.\n\" }}\n        {{- \"// If `id` is a string, it is treated as a fully qualified URL associated with `source`.\n\" }}\n        {{- \"// If `loc` is not provided, the viewport will be positioned at the beginning of the document or centered on the most relevant passage, if available.\n\" }}\n        {{- \"// Use this function without `id` to scroll to a new location of an opened page.\n\" }}\n        {{- \"type open = (_: {\n\" }}\n        {{- \"id?: number | string, // default: -1\n\" }}\n        {{- \"cursor?: number, // default: -1\n\" }}\n        {{- \"loc?: number, // default: -1\n\" }}\n        {{- \"num_lines?: number, // default: -1\n\" }}\n        {{- \"view_source?: boolean, // default: false\n\" }}\n        {{- \"source?: string,\n\" }}\n        {{- \"}) => any;\n\n\" }}\n        {{- \"// Finds exact matches of `pattern` in the current page, or the page given by `cursor`.\n\" }}\n        {{- \"type find = (_: {\n\" }}\n        {{- \"pattern: string,\n\" }}\n        {{- \"cursor?: number, // default: -1\n\" }}\n        {{- \"}) => any;\n\n\" }}\n        {{- \"} // namespace browser\n\n\" }}\n    {%- endif -%}\n\n    {%- if python_tool %}\n        {{- \"## python\n\n\" }}\n        {{- \"Use this tool to execute Python code in your chain of thought. The code will not be shown to the user. This tool should be used for internal reasoning, but not for code that is intended to be visible to the user (e.g. when creating plots, tables, or files).\n\n\" }}\n        {{- \"When you send a message containing Python code to python, it will be executed in a stateful Jupyter notebook environment. python will respond with the output of the execution or time out after 120.0 seconds. The drive at '/mnt/data' can be used to save and persist user files. Internet access for this session is UNKNOWN. Depends on the cluster.\n\n\" }}\n    {%- endif -%}\n{%- endmacro -%}\n\n{#- System Message Construction ============================================ #}\n{%- macro build_system_message() -%}\n    {%- if model_identity is not defined %}\n        {%- set model_identity = \"You are ChatGPT, a large language model trained by OpenAI.\" %}\n    {%- endif %}\n    {{- model_identity + \"\n\" }}\n    {{- \"Knowledge cutoff: 2024-06\n\" }}\n    {{- \"Current date: \" + strftime_now(\"%Y-%m-%d\") + \"\n\n\" }}\n    {%- if reasoning_effort is not defined %}\n        {%- set reasoning_effort = \"medium\" %}\n    {%- endif %}\n    {{- \"Reasoning: \" + reasoning_effort + \"\n\n\" }}\n    {%- if builtin_tools %}\n        {{- \"# Tools\n\n\" }}\n        {%- set available_builtin_tools = namespace(browser=false, python=false) %}\n        {%- for tool in builtin_tools %}\n            {%- if tool == \"browser\" %}\n                {%- set available_builtin_tools.browser = true %}\n            {%- elif tool == \"python\" %}\n                {%- set available_builtin_tools.python = true %}\n            {%- endif %}\n        {%- endfor %}\n        {{- render_builtin_tools(available_builtin_tools.browser, available_builtin_tools.python) }}\n    {%- endif -%}\n    {{- \"# Valid channels: analysis, commentary, final. Channel must be included for every message.\" }}\n    {%- if tools -%}\n        {{- \"\nCalls to these tools must go to the commentary channel: 'functions'.\" }}\n    {%- endif -%}\n{%- endmacro -%}\n\n{#- Main Template Logic ================================================= #}\n{#- Set defaults #}\n\n{#- Render system message #}\n{{- \"<|start|>system<|message|>\" }}\n{{- build_system_message() }}\n{{- \"<|end|>\" }}\n\n{#- Extract developer message #}\n{%- if messages[0].role == \"developer\" or messages[0].role == \"system\" %}\n    {%- set developer_message = messages[0].content %}\n    {%- set loop_messages = messages[1:] %}\n{%- else %}\n    {%- set developer_message = \"\" %}\n    {%- set loop_messages = messages %}\n{%- endif %}\n\n{#- Render developer message #}\n{%- if developer_message or tools %}\n    {{- \"<|start|>developer<|message|>\" }}\n    {%- if developer_message %}\n        {{- \"# Instructions\n\n\" }}\n        {{- developer_message }}\n    {%- endif %}\n    {%- if tools -%}\n        {{- \"\n\n\" }}\n        {{- \"# Tools\n\n\" }}\n        {{- render_tool_namespace(\"functions\", tools) }}\n    {%- endif -%}\n    {{- \"<|end|>\" }}\n{%- endif %}\n\n{#- Render messages #}\n{%- set last_tool_call = namespace(name=none) %}\n{%- for message in loop_messages -%}\n    {#- At this point only assistant/user/tool messages should remain #}\n    {%- if message.role == 'assistant' -%}\n        {#- Checks to ensure the messages are being passed in the format we expect #}\n        {%- if \"content\" in message %}\n            {%- if \"<|channel|>analysis<|message|>\" in message.content or \"<|channel|>final<|message|>\" in message.content %}\n                {{- raise_exception(\"You have passed a message containing <|channel|> tags in the content field. Instead of doing this, you should pass analysis messages (the string between '<|message|>' and '<|end|>') in the 'thinking' field, and final messages (the string between '<|message|>' and '<|end|>') in the 'content' field.\") }}\n            {%- endif %}\n        {%- endif %}\n        {%- if \"thinking\" in message %}\n            {%- if \"<|channel|>analysis<|message|>\" in message.thinking or \"<|channel|>final<|message|>\" in message.thinking %}\n                {{- raise_exception(\"You have passed a message containing <|channel|> tags in the thinking field. Instead of doing this, you should pass analysis messages (the string between '<|message|>' and '<|end|>') in the 'thinking' field, and final messages (the string between '<|message|>' and '<|end|>') in the 'content' field.\") }}\n            {%- endif %}\n        {%- endif %}\n        {%- if \"tool_calls\" in message %}\n            {#- We assume max 1 tool call per message, and so we infer the tool call name #}\n            {#- in \"tool\" messages from the most recent assistant tool call name #}\n            {%- set tool_call = message.tool_calls[0] %}\n            {%- if tool_call.function %}\n                {%- set tool_call = tool_call.function %}\n            {%- endif %}\n            {%- if message.content and message.thinking %}\n                {{- raise_exception(\"Cannot pass both content and thinking in an assistant message with tool calls! Put the analysis message in one or the other, but not both.\") }}\n            {%- elif message.content %}\n                {{- \"<|start|>assistant<|channel|>analysis<|message|>\" + message.content + \"<|end|>\" }}\n            {%- elif message.thinking %}\n                {{- \"<|start|>assistant<|channel|>analysis<|message|>\" + message.thinking + \"<|end|>\" }}\n            {%- endif %}\n            {{- \"<|start|>assistant to=\" }}\n            {{- \"functions.\" + tool_call.name + \"<|channel|>commentary \" }}\n            {{- (tool_call.content_type if tool_call.content_type is defined else \"json\") + \"<|message|>\" }}\n            {{- tool_call.arguments|tojson }}\n            {{- \"<|call|>\" }}\n            {%- set last_tool_call.name = tool_call.name %}\n        {%- elif loop.last and not add_generation_prompt %}\n            {#- Only render the CoT if the final turn is an assistant turn and add_generation_prompt is false #}\n            {#- This is a situation that should only occur in training, never in inference. #}\n            {%- if \"thinking\" in message %}\n                {{- \"<|start|>assistant<|channel|>analysis<|message|>\" + message.thinking + \"<|end|>\" }}\n            {%- endif %}\n            {#- <|return|> indicates the end of generation, but <|end|> does not #}\n            {#- <|return|> should never be an input to the model, but we include it as the final token #}\n            {#- when training, so the model learns to emit it. #}\n            {{- \"<|start|>assistant<|channel|>final<|message|>\" + message.content + \"<|return|>\" }}\n        {%- else %}\n            {#- CoT is dropped during all previous turns, so we never render it for inference #}\n            {{- \"<|start|>assistant<|channel|>final<|message|>\" + message.content + \"<|end|>\" }}\n            {%- set last_tool_call.name = none %}\n        {%- endif %}\n    {%- elif message.role == 'tool' -%}\n        {%- if last_tool_call.name is none %}\n            {{- raise_exception(\"Message has tool role, but there was no previous assistant message with a tool call!\") }}\n        {%- endif %}\n        {{- \"<|start|>functions.\" + last_tool_call.name }}\n        {{- \" to=assistant<|channel|>commentary<|message|>\" + message.content|tojson + \"<|end|>\" }}\n    {%- elif message.role == 'user' -%}\n        {{- \"<|start|>user<|message|>\" + message.content + \"<|end|>\" }}\n    {%- endif -%}\n{%- endfor -%}\n\n{#- Generation prompt #}\n{%- if add_generation_prompt -%}\n<|start|>assistant\n{%- endif -%}"
+}

config.json

@@ -46,25 +46,15 @@
   "num_key_value_heads": 8,
   "num_local_experts": 32,
   "output_router_logits": false,
-  "pad_token_id": 199999,
+  "pad_token_id": 200017,
   "quantization_config": {
-    "_load_in_4bit": true,
-    "_load_in_8bit": false,
-    "bnb_4bit_compute_dtype": "bfloat16",
-    "bnb_4bit_quant_storage": "uint8",
-    "bnb_4bit_quant_type": "nf4",
-    "bnb_4bit_use_double_quant": true,
-    "llm_int8_enable_fp32_cpu_offload": false,
-    "llm_int8_has_fp16_weight": false,
-    "llm_int8_skip_modules": [
-      "router",
-      "lm_head",
-      "embed_tokens"
+    "modules_to_not_convert": [
+      "model.layers.*.self_attn",
+      "model.layers.*.mlp.router",
+      "model.embed_tokens",
+      "lm_head"
     ],
-    "llm_int8_threshold": 6.0,
-    "load_in_4bit": true,
-    "load_in_8bit": false,
-    "quant_method": "bitsandbytes"
+    "quant_method": "mxfp4"
   },
   "rms_norm_eps": 1e-05,
   "rope_scaling": {
@@ -80,8 +70,8 @@
   "sliding_window": 128,
   "swiglu_limit": 7.0,
   "tie_word_embeddings": false,
-  "torch_dtype": "float16",
   "transformers_version": "4.55.0",
+  "unsloth_fixed": true,
   "use_cache": true,
   "vocab_size": 201088
 }

generation_config.json

@@ -6,5 +6,5 @@
     199999
   ],
   "pad_token_id": 199999,
-  "transformers_version": "4.55.0"
+  "transformers_version": "4.55.0.dev0"
 }

tokenizer_config.json

@@ -181,5 +181,6 @@
   "pad_token": "<|reserved_200017|>",
   "padding_side": "left",
   "tokenizer_class": "PreTrainedTokenizerFast",
-  "unk_token": null
-}
+  "unk_token": null,
+  "chat_template": "{# Copyright 2025-present Unsloth. Apache 2.0 License. Unsloth chat template fixes. Edited from ggml-org & OpenAI #}\n{#-\n  In addition to the normal inputs of `messages` and `tools`, this template also accepts the\n  following kwargs:\n  - \"builtin_tools\": A list, can contain \"browser\" and/or \"python\".\n  - \"model_identity\": A string that optionally describes the model identity.\n  - \"reasoning_effort\": A string that describes the reasoning effort, defaults to \"medium\".\n #}\n\n{#- Tool Definition Rendering ============================================== #}\n{%- macro render_typescript_type(param_spec, required_params, is_nullable=false) -%}\n    {%- if param_spec.type == \"array\" -%}\n        {%- if param_spec['items'] -%}\n            {%- if param_spec['items']['type'] == \"string\" -%}\n                {{- \"string[]\" }}\n            {%- elif param_spec['items']['type'] == \"number\" -%}\n                {{- \"number[]\" }}\n            {%- elif param_spec['items']['type'] == \"integer\" -%}\n                {{- \"number[]\" }}\n            {%- elif param_spec['items']['type'] == \"boolean\" -%}\n                {{- \"boolean[]\" }}\n            {%- else -%}\n                {%- set inner_type = render_typescript_type(param_spec['items'], required_params) -%}\n                {%- if inner_type == \"object | object\" or inner_type|length > 50 -%}\n                    {{- \"any[]\" }}\n                {%- else -%}\n                    {{- inner_type + \"[]\" }}\n                {%- endif -%}\n            {%- endif -%}\n            {%- if param_spec.nullable -%}\n                {{- \" | null\" }}\n            {%- endif -%}\n        {%- else -%}\n            {{- \"any[]\" }}\n            {%- if param_spec.nullable -%}\n                {{- \" | null\" }}\n            {%- endif -%}\n        {%- endif -%}\n    {%- elif param_spec.type is defined and param_spec.type is iterable and param_spec.type is not string and param_spec.type is not mapping and param_spec.type[0] is defined -%}\n        {#- Handle array of types like [\"object\", \"object\"] from Union[dict, list] #}\n        {%- if param_spec.type | length > 1 -%}\n            {{- param_spec.type | join(\" | \") }}\n        {%- else -%}\n            {{- param_spec.type[0] }}\n        {%- endif -%}\n    {%- elif param_spec.oneOf -%}\n        {#- Handle oneOf schemas - check for complex unions and fallback to any #}\n        {%- set has_object_variants = false -%}\n        {%- for variant in param_spec.oneOf -%}\n            {%- if variant.type == \"object\" -%}\n                {%- set has_object_variants = true -%}\n            {%- endif -%}\n        {%- endfor -%}\n        {%- if has_object_variants and param_spec.oneOf|length > 1 -%}\n            {{- \"any\" }}\n        {%- else -%}\n            {%- for variant in param_spec.oneOf -%}\n                {{- render_typescript_type(variant, required_params) -}}\n                {%- if variant.description %}\n                    {{- \"// \" + variant.description }}\n                {%- endif -%}\n                {%- if variant.default is defined %}\n                    {{ \"// default: \" + variant.default|tojson }}\n                {%- endif -%}\n                {%- if not loop.last %}\n                    {{- \" | \" }}\n                {% endif -%}\n            {%- endfor -%}\n        {%- endif -%}\n    {%- elif param_spec.type == \"string\" -%}\n        {%- if param_spec.enum -%}\n            {{- '\"' + param_spec.enum|join('\" | \"') + '\"' -}}\n        {%- else -%}\n            {{- \"string\" }}\n            {%- if param_spec.nullable %}\n                {{- \" | null\" }}\n            {%- endif -%}\n        {%- endif -%}\n    {%- elif param_spec.type == \"number\" -%}\n        {{- \"number\" }}\n    {%- elif param_spec.type == \"integer\" -%}\n        {{- \"number\" }}\n    {%- elif param_spec.type == \"boolean\" -%}\n        {{- \"boolean\" }}\n\n    {%- elif param_spec.type == \"object\" -%}\n        {%- if param_spec.properties -%}\n            {{- \"{\\n\" }}\n            {%- for prop_name, prop_spec in param_spec.properties.items() -%}\n                {{- prop_name -}}\n                {%- if prop_name not in (param_spec.required or []) -%}\n                    {{- \"?\" }}\n                {%- endif -%}\n                {{- \": \" }}\n                {{ render_typescript_type(prop_spec, param_spec.required or []) }}\n                {%- if not loop.last -%}\n                    {{-\", \" }}\n                {%- endif -%}\n            {%- endfor -%}\n            {{- \"}\" }}\n        {%- else -%}\n            {{- \"object\" }}\n        {%- endif -%}\n    {%- else -%}\n        {{- \"any\" }}\n    {%- endif -%}\n{%- endmacro -%}\n\n{%- macro render_tool_namespace(namespace_name, tools) -%}\n    {{- \"## \" + namespace_name + \"\\n\\n\" }}\n    {{- \"namespace \" + namespace_name + \" {\\n\\n\" }}\n    {%- for tool in tools %}\n        {%- set tool = tool.function %}\n        {{- \"// \" + tool.description + \"\\n\" }}\n        {{- \"type \"+ tool.name + \" = \" }}\n        {%- if tool.parameters and tool.parameters.properties -%}\n            {{- \"(_: \" }}\n            {{- \"{\\n\" }}\n            {%- for param_name, param_spec in tool.parameters.properties.items() %}\n                {{- \"// \" + param_spec.description + \"\\n\" }}\n                {{- param_name }}\n                {%- if param_name not in (tool.parameters.required or []) -%}\n                    {{- \"?\" }}\n                {%- endif -%}\n                {{- \": \" }}\n                {{- render_typescript_type(param_spec, tool.parameters.required or []) }}\n                {%- if param_spec.default is defined -%}\n                    {%- if param_spec.enum %}\n                        {{- \", // default: \" + param_spec.default }}\n                    {%- elif param_spec.oneOf %}\n                        {{- \"// default: \" + param_spec.default }}\n                    {%- else %}\n                        {{- \", // default: \" + param_spec.default|tojson }}\n                    {%- endif -%}\n                {%- endif -%}\n                {%- if not loop.last %}\n                    {{- \",\\n\" }}\n                {%- else %}\n                    {{- \"\\n\" }}\n                {%- endif -%}\n            {%- endfor %}\n            {{- \"}) => any;\\n\\n\" }}\n        {%- else -%}\n            {{- \"() => any;\\n\\n\" }}\n        {%- endif -%}\n    {%- endfor %}\n    {{- \"} // namespace \" + namespace_name }}\n{%- endmacro -%}\n\n{%- macro render_builtin_tools(browser_tool, python_tool) -%}\n    {%- if browser_tool %}\n        {{- \"## browser\\n\\n\" }}\n        {{- \"// Tool for browsing.\\n\" }}\n        {{- \"// The `cursor` appears in brackets before each browsing display: `[{cursor}]`.\\n\" }}\n        {{- \"// Cite information from the tool using the following format:\\n\" }}\n        {{- \"// `【{cursor}†L{line_start}(-L{line_end})?】`, for example: `【6†L9-L11】` or `【8†L3】`.\\n\" }}\n        {{- \"// Do not quote more than 10 words directly from the tool output.\\n\" }}\n        {{- \"// sources=web (default: web)\\n\" }}\n        {{- \"namespace browser {\\n\\n\" }}\n        {{- \"// Searches for information related to `query` and displays `topn` results.\\n\" }}\n        {{- \"type search = (_: {\\n\" }}\n        {{- \"query: string,\\n\" }}\n        {{- \"topn?: number, // default: 10\\n\" }}\n        {{- \"source?: string,\\n\" }}\n        {{- \"}) => any;\\n\\n\" }}\n        {{- \"// Opens the link `id` from the page indicated by `cursor` starting at line number `loc`, showing `num_lines` lines.\\n\" }}\n        {{- \"// Valid link ids are displayed with the formatting: `【{id}†.*】`.\\n\" }}\n        {{- \"// If `cursor` is not provided, the most recent page is implied.\\n\" }}\n        {{- \"// If `id` is a string, it is treated as a fully qualified URL associated with `source`.\\n\" }}\n        {{- \"// If `loc` is not provided, the viewport will be positioned at the beginning of the document or centered on the most relevant passage, if available.\\n\" }}\n        {{- \"// Use this function without `id` to scroll to a new location of an opened page.\\n\" }}\n        {{- \"type open = (_: {\\n\" }}\n        {{- \"id?: number | string, // default: -1\\n\" }}\n        {{- \"cursor?: number, // default: -1\\n\" }}\n        {{- \"loc?: number, // default: -1\\n\" }}\n        {{- \"num_lines?: number, // default: -1\\n\" }}\n        {{- \"view_source?: boolean, // default: false\\n\" }}\n        {{- \"source?: string,\\n\" }}\n        {{- \"}) => any;\\n\\n\" }}\n        {{- \"// Finds exact matches of `pattern` in the current page, or the page given by `cursor`.\\n\" }}\n        {{- \"type find = (_: {\\n\" }}\n        {{- \"pattern: string,\\n\" }}\n        {{- \"cursor?: number, // default: -1\\n\" }}\n        {{- \"}) => any;\\n\\n\" }}\n        {{- \"} // namespace browser\\n\\n\" }}\n    {%- endif -%}\n\n    {%- if python_tool %}\n        {{- \"## python\\n\\n\" }}\n        {{- \"Use this tool to execute Python code in your chain of thought. The code will not be shown to the user. This tool should be used for internal reasoning, but not for code that is intended to be visible to the user (e.g. when creating plots, tables, or files).\\n\\n\" }}\n        {{- \"When you send a message containing Python code to python, it will be executed in a stateful Jupyter notebook environment. python will respond with the output of the execution or time out after 120.0 seconds. The drive at '/mnt/data' can be used to save and persist user files. Internet access for this session is UNKNOWN. Depends on the cluster.\\n\\n\" }}\n    {%- endif -%}\n{%- endmacro -%}\n\n{#- System Message Construction ============================================ #}\n{%- macro build_system_message() -%}\n    {%- if model_identity is not defined %}\n        {{- \"You are ChatGPT, a large language model trained by OpenAI.\" -}}\n    {%- else %}\n        {{- model_identity }}\n    {%- endif %}\n    {{- \"\\nKnowledge cutoff: 2024-06\\n\" }}\n    {{- \"Current date: \" + strftime_now(\"%Y-%m-%d\") + \"\\n\\n\" }}\n    {%- if reasoning_effort is not defined %}\n        {%- set reasoning_effort = \"medium\" %}\n    {%- endif %}\n    {{- \"Reasoning: \" + reasoning_effort + \"\\n\\n\" }}\n    {%- if builtin_tools is defined and builtin_tools is none %}\n        {{- \"# Tools\\n\\n\" }}\n        {%- set available_builtin_tools = namespace(browser=false, python=false) %}\n        {%- for tool in builtin_tools %}\n            {%- if tool == \"browser\" %}\n                {%- set available_builtin_tools.browser = true %}\n            {%- elif tool == \"python\" %}\n                {%- set available_builtin_tools.python = true %}\n            {%- endif %}\n        {%- endfor %}\n        {{- render_builtin_tools(available_builtin_tools.browser, available_builtin_tools.python) }}\n    {%- endif -%}\n    {{- \"# Valid channels: analysis, commentary, final. Channel must be included for every message.\" }}\n    {%- if tools is defined and tools is not none -%}\n        {{- \"\\nCalls to these tools must go to the commentary channel: 'functions'.\" }}\n    {%- endif -%}\n{%- endmacro -%}\n\n{#- Main Template Logic ================================================= #}\n{#- Set defaults #}\n\n{#- Render system message #}\n{{- \"<|start|>system<|message|>\" }}\n{{- build_system_message() }}\n{{- \"<|end|>\" }}\n\n{#- Extract developer message #}\n{%- if messages[0].role == \"developer\" or messages[0].role == \"system\" %}\n    {%- set developer_message = messages[0].content %}\n    {%- set loop_messages = messages[1:] %}\n{%- else %}\n    {%- set developer_message = \"\" %}\n    {%- set loop_messages = messages %}\n{%- endif %}\n\n{#- Render developer message #}\n{%- if developer_message or tools %}\n    {{- \"<|start|>developer<|message|>\" }}\n    {%- if developer_message %}\n        {{- \"# Instructions\\n\\n\" }}\n        {{- developer_message }}\n    {%- endif %}\n    {%- if tools -%}\n        {{- \"\\n\\n\" }}\n        {{- \"# Tools\\n\\n\" }}\n        {{- render_tool_namespace(\"functions\", tools) }}\n    {%- endif -%}\n    {{- \"<|end|>\" }}\n{%- endif %}\n\n{#- Render messages #}\n{%- set last_tool_call = namespace(name=none) %}\n{%- for message in loop_messages -%}\n    {#- At this point only assistant/user/tool messages should remain #}\n    {%- if message.role == 'assistant' -%}\n        {%- if \"tool_calls\" in message %}\n            {#- We assume max 1 tool call per message, and so we infer the tool call name #}\n            {#- in \"tool\" messages from the most recent assistant tool call name #}\n            {%- set tool_call = message.tool_calls[0] %}\n            {%- if tool_call.function %}\n                {%- set tool_call = tool_call.function %}\n            {%- endif %}\n            {%- if message.content %}\n                {{- \"<|start|>assistant<|channel|>analysis<|message|>\" + message.content + \"<|end|>\" }}\n            {%- endif %}\n            {{- \"<|start|>assistant to=\" }}\n            {{- \"functions.\" + tool_call.name + \"<|channel|>commentary json<|message|>\" }}\n            {{- tool_call.arguments|tojson }}\n            {{- \"<|call|>\" }}\n            {%- set last_tool_call.name = tool_call.name %}\n        {%- elif \"thinking\" in message and loop.last and not add_generation_prompt %}\n            {#- Only render the CoT if the final turn is an assistant turn and add_generation_prompt is false #}\n            {#- This is a situation that should only occur in training, never in inference. #}\n            {{- \"<|start|>assistant<|channel|>analysis<|message|>\" + message.thinking + \"<|end|>\" }}\n            {#- <|return|> indicates the end of generation, but <|end|> does not #}\n            {#- <|return|> should never be an input to the model, but we include it as the final token #}\n            {#- when training, so the model learns to emit it. #}\n            {{- \"<|start|>assistant<|channel|>final<|message|>\" + message.content + \"<|return|>\" }}\n            {%- set last_tool_call.name = none %}\n        {%- elif \"thinking\" in message %}\n            {#- CoT is dropped during all previous turns, so we never render it for inference #}\n            {{- \"<|start|>assistant<|channel|>final<|message|>\" + message.content + \"<|end|>\" }}\n            {%- set last_tool_call.name = none %}\n        {%- elif loop.last and not add_generation_prompt %}\n            {#- <|return|> indicates the end of generation, but <|end|> does not #}\n            {#- <|return|> should never be an input to the model, but we include it as the final token #}\n            {#- when training, so the model learns to emit it. #}\n            {{- \"<|start|>assistant<|message|>\" + message.content + \"<|return|>\" }}\n        {%- else %}\n            {{- \"<|start|>assistant<|message|>\" + message.content + \"<|end|>\" }}\n            {%- set last_tool_call.name = none %}\n        {%- endif %}\n    {%- elif message.role == 'tool' -%}\n        {%- if last_tool_call.name is none %}\n            {{- raise_exception(\"Message has tool role, but there was no previous assistant message with a tool call!\") }}\n        {%- endif %}\n        {{- \"<|start|>functions.\" + last_tool_call.name }}\n        {{- \" to=assistant<|channel|>commentary<|message|>\" + message.content|tojson + \"<|end|>\" }}\n    {%- else -%}\n        {{- \"<|start|>user<|message|>\" + message.content + \"<|end|>\" }}\n    {%- endif -%}\n{%- endfor -%}\n\n{#- Generation prompt #}\n{%- if add_generation_prompt -%}\n<|start|>assistant\n{%- endif -%}\n{# Copyright 2025-present Unsloth. Apache 2.0 License. Unsloth chat template fixes. Edited from ggml-org & OpenAI #}"
+}