fix: match LFM2.5-8B-A1B quick start to HF params

CLAUDE.md

@@ -144,7 +144,7 @@ Every inference example must use the correct sampling params from the upstream H
 |---|---|---|---|---|---|
 | LFM2.5-1.2B-Instruct | 0.1 | 50 | — | — | 1.05 |
 | LFM2.5-1.2B-Thinking | 0.1 | 50 | 0.1 | — | 1.05 |
-| LFM2.5-8B-A1B | 0.1 | 50 | — | — | 1.05 |
+| LFM2.5-8B-A1B | 0.2 | 80 | — | — | 1.05 |
 | LFM2-24B-A2B | 0.1 | 50 | — | — | 1.05 |
 | LFM2 text + LFM2.5-JP | 0.3 | — | — | 0.15 | 1.05 |
 | All VL models | 0.1 | — | — | 0.15 | 1.05 |

lfm/models/lfm25-8b-a1b.mdx

@@ -10,7 +10,7 @@
 
 <a href="/lfm/models/text-models" className="back-button">← Back to Text Models</a>
 
 LFM2.5-8B-A1B is Liquid AI's Mixture-of-Experts model, combining 8B total parameters with only 1.5B active parameters per forward pass with a 128K context window and chain of thought reasoning. This model delivers exceptional performance in tool calling and agentic tasks while running on-device.
 
 <div style={{display: 'flex', gap: '0.5rem', margin: '0.5rem 0 1.5rem 0'}}>
   <a href="https://huggingface.co/LiquidAI/LFM2.5-8B-A1B" style={{padding: '0.35rem 0.7rem', borderRadius: '4px', fontSize: '0.85rem', fontWeight: 600, textDecoration: 'none', backgroundColor: '#fbbf24'}}><span style={{color: '#000'}}>HF</span></a>
@@ -39,7 +39,7 @@
 </Card>
 
 <Card title="Tool Calling" icon="wrench">
   Native function calling for agentic workflows
 </Card>
 
 </CardGroup>
@@ -49,15 +49,48 @@
 
 <Tabs>
   <Tab title="Transformers">
-    <TextTransformers modelId="LiquidAI/LFM2.5-8B-A1B" samplingParams="do_sample=True, temperature=0.1, top_k=50, repetition_penalty=1.05, "></TextTransformers>
+    Quick start with Transformers (compatible with `transformers>=5.0.0`):
+
+    ```python
+    from transformers import AutoModelForCausalLM, AutoTokenizer, TextStreamer
+
+    model_id = "LiquidAI/LFM2.5-8B-A1B"
+    model = AutoModelForCausalLM.from_pretrained(
+        model_id,
+        device_map="auto",
+        dtype="bfloat16",
+    #   attn_implementation="flash_attention_2" <- uncomment on compatible GPU
+    )
+    tokenizer = AutoTokenizer.from_pretrained(model_id)
+    streamer = TextStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
+
+    prompt = "What is C. elegans?"
+
+    input_ids = tokenizer.apply_chat_template(
+        [{"role": "user", "content": prompt}],
+        add_generation_prompt=True,
+        return_tensors="pt",
+        tokenize=True,
+    ).to(model.device)
+
+    output = model.generate(
+        input_ids,
+        do_sample=True,
+        temperature=0.2,
+        top_k=80,
+        repetition_penalty=1.05,
+        max_new_tokens=8192,
+        streamer=streamer,
+    )
+    ```
   </Tab>
   <Tab title="llama.cpp">
-    <TextLlamacpp ggufRepo="LiquidAI/LFM2.5-8B-A1B-GGUF" samplingFlags="--temp 0.1 --top-k 50 --repeat-penalty 1.05"></TextLlamacpp>
+    <TextLlamacpp ggufRepo="LiquidAI/LFM2.5-8B-A1B-GGUF" samplingFlags="--temp 0.2 --top-k 80 --repeat-penalty 1.05"></TextLlamacpp>
   </Tab>
   <Tab title="vLLM">
-    <TextVllm modelId="LiquidAI/LFM2.5-8B-A1B" samplingParams="temperature=0.1, top_k=50, repetition_penalty=1.05, "></TextVllm>
+    <TextVllm modelId="LiquidAI/LFM2.5-8B-A1B" samplingParams="temperature=0.2, top_k=80, repetition_penalty=1.05, " maxTokens="8192"></TextVllm>
   </Tab>
   <Tab title="SGLang">
-    <TextSglang modelId="LiquidAI/LFM2.5-8B-A1B" toolCallParser="lfm2"></TextSglang>
+    <TextSglang modelId="LiquidAI/LFM2.5-8B-A1B" toolCallParser="lfm2" samplingParams={'temperature=0.2,\n    extra_body={"top_k": 80, "repetition_penalty": 1.05},'}></TextSglang>
   </Tab>
 </Tabs>

scripts/generate_snippets.py

@@ -76,7 +76,7 @@
     },
     "text-vllm": {
         "component_name": "TextVllm",
-        "props": "{ modelId, samplingParams }",
+        "props": "{ modelId, samplingParams, maxTokens }",
         "replacement_group": "text",
         "source": "config",
         "sections": [
@@ -90,7 +90,7 @@
                  "\n"
                  'llm = LLM(model="${modelId}")\n'
                  "\n"
-                 "sampling_params = SamplingParams(${samplingParams}max_tokens=512)\n"
+                 "sampling_params = SamplingParams(${samplingParams}max_tokens=${maxTokens || 512})\n"
                  "\n"
                  'output = llm.chat("What is machine learning?", sampling_params)\n'
                  "print(output[0].outputs[0].text)"

snippets/quickstart/text-sglang.mdx

@@ -1,4 +1,4 @@
-export const TextSglang = ({ modelId, toolCallParser }) => (
+export const TextSglang = ({ modelId, toolCallParser, samplingParams }) => (
 <div>
 <p><strong>Install:</strong></p>
 <pre className="shiki shiki-themes github-light github-dark" style={{backgroundColor: '#fff', '--shiki-dark-bg': '#24292e', color: '#24292e', '--shiki-dark': '#e1e4e8'}} language="bash">
@@ -32,7 +32,7 @@ client = OpenAI(base_url="http://localhost:30000/v1", api_key="None")
 response = client.chat.completions.create(
     model="${modelId}",
     messages=[{"role": "user", "content": "What is machine learning?"}],
-    temperature=0.3,
+    ${samplingParams || "temperature=0.3,"}
 )
 
 print(response.choices[0].message.content)`.split('\n').map((line, i) => <span key={i} className="line">{line}{'\n'}</span>)}

snippets/quickstart/text-vllm.mdx

@@ -1,4 +1,4 @@
-export const TextVllm = ({ modelId, samplingParams }) => (
+export const TextVllm = ({ modelId, samplingParams, maxTokens }) => (
 <div>
 <p><strong>Install:</strong></p>
 <pre className="shiki shiki-themes github-light github-dark" style={{backgroundColor: '#fff', '--shiki-dark-bg': '#24292e', color: '#24292e', '--shiki-dark': '#e1e4e8'}} language="bash">
@@ -13,7 +13,7 @@ export const TextVllm = ({ modelId, samplingParams }) => (
 
 llm = LLM(model="${modelId}")
 
-sampling_params = SamplingParams(${samplingParams}max_tokens=512)
+sampling_params = SamplingParams(${samplingParams}max_tokens=${maxTokens || 512})
 
 output = llm.chat("What is machine learning?", sampling_params)
 print(output[0].outputs[0].text)`.split('\n').map((line, i) => <span key={i} className="line">{line}{'\n'}</span>)}