Model card: variant-specific rewrite (match XS.2-FP8); add LICENSE.md

README.md

@@ -7,7 +7,6 @@ extra_gated_description: >-
 tags:
 - laguna-m.1
 - vllm
-- sglang
 - fp8
 - moe
 license: apache-2.0
@@ -79,7 +78,7 @@ Laguna M.1-FP8 is a 225B total parameter Mixture-of-Experts model with 23B activ
 
 ## Usage
 
-Laguna M.1 has upstream support in vLLM, SGLang, and TRT-LLM thanks to the support of the team at NVIDIA.
+Laguna M.1 has upstream support in vLLM, and TRT-LLM thanks to the support of the team at NVIDIA.
 
 > [!NOTE]
 > For complete usage instructions, see the main [Laguna M.1 model card](https://huggingface.co/poolside/Laguna-M.1).
@@ -88,13 +87,11 @@ Laguna M.1 has upstream support in vLLM, SGLang, and TRT-LLM thanks to the suppo
 
 #### vLLM
 
-The full vLLM recipe is on the main [Laguna M.1 model card](https://huggingface.co/poolside/Laguna-M.1). Quantization is detected automatically from `quantization_config` in this checkpoint, so the same command works with `poolside/Laguna-M.1-FP8` substituted for the model ID. Set `VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER=0` when serving with vLLM.
+The full vLLM recipe is on the main [Laguna M.1 model card](https://huggingface.co/poolside/Laguna-M.1). Quantization is detected automatically from `quantization_config` in this checkpoint, so the same command works with `poolside/Laguna-M.1-FP8` substituted for the model ID. No extra flags required.
 
 ```shell
 pip install 'vllm>=0.21.0'
 
-export VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER=0
-
 vllm serve \
     --model poolside/Laguna-M.1-FP8 \
     --tool-call-parser poolside_v1 \
@@ -104,23 +101,6 @@ vllm serve \
     --default-chat-template-kwargs '{"enable_thinking": true}'
 ```
 
-#### SGLang
-
-The full SGLang recipe is on the [SGLang Cookbook](https://docs.sglang.io/cookbook/autoregressive/Poolside/Laguna-M.1). Quantization is detected automatically, so no extra flags are required. 
-```shell
-git clone https://github.com/sgl-project/sglang.git
-cd sglang
-pip install -e "python[all]"
-
-sglang serve \
-    --model-path poolside/Laguna-M.1-FP8 \
-    --trust-remote-code \
-    --reasoning-parser poolside_v1 \
-    --tool-call-parser poolside_v1 \
-    --tp 8 \
-    --host 0.0.0.0 \
-    --port 30000
-```
 #### TRT-LLM
 
 Laguna is supported in TensorRT-LLM thanks to the team at NVIDIA ([NVIDIA/TensorRT-LLM#13559](https://github.com/NVIDIA/TensorRT-LLM/pull/13559), with partial-RoPE fusion in [#15110](https://github.com/NVIDIA/TensorRT-LLM/pull/15110)). The full recipe is on the main [Laguna M.1 model card](https://huggingface.co/poolside/Laguna-M.1). Quantization is detected automatically from `quantization_config` in this checkpoint, so no extra flags are required.

generation_config.json

@@ -9,10 +9,5 @@
   "pad_token_id": 9,
   "temperature": 1.0,
   "top_p": 1.0,
-  "min_p": 0.0,
-  "tool_call_parser": "poolside_v1",
-  "reasoning_parser": "poolside_v1",
-  "default_chat_template_kwargs": {
-    "enable_thinking": true
-  }
-}
+  "min_p": 0.0
+}

tokenizer_config.json

@@ -571,5 +571,6 @@
   "pad_token": "〈|PAD|〉",
   "sep_token": "〈|SEP|〉",
   "tokenizer_class": "PreTrainedTokenizerFast",
-  "unk_token": "〈|UNK|〉"
-}
+  "unk_token": "〈|UNK|〉",
+  "chat_template": "{% include 'chat_template.jinja' %}"
+}