GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
SwapnilS-15
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
SwapnilS-15
Model Tree
Qwen/Qwen2.5-0.5B-Instruct
Input Modalities
Output Modalities
Supported Functionality
Frozen test set of 2,929 xLAM questions, greedy decoding, scored by the same code for base and fine-tuned model.
| Metric | Fine-tuned | 95% CI |
|---|---|---|
| exact_match | 78.0% | 76.5–79.4 |
| function_match | 97.3% | 96.7–97.8 |
| argument_f1 | 90.3% | 89.5–91.1 |
| schema_valid | 99.4% | 99.0–99.6 |
Exact match: base 61.6% → fine-tuned 78.0% (592 questions fixed, 112 broken, McNemar p < 0.001).
| Call pattern | n | Exact match |
|---|---|---|
| single | 1,386 | 81.8% |
| parallel_same_tool | 530 | 80.4% |
| multi_distinct_tools | 936 | 71.3% |
| mixed | 77 | 74.0% |
The model was trained on this exact format. It is not Qwen's native <tool_call> format, and it does not read OpenAI-style tools= parameters.
System:
text
You are a function-calling assistant. Your task is to select the appropriate tool and provide the arguments requiredto answer the user's request. Return ONLY valid JSON. The output must be a JSON array of tool calls in this format: [ { "name": "function_name", "arguments": { "argument_name": "value" } }] If none of the available tools can answer the request, return an empty array: [] Do not explain your answer.Do not output markdown.Do not invent tools.User:
text
Available tools: <tools as a JSON list: name, description, parameters> User request: <the request> Return the required tool call(s) as JSON.Assistant: [{"name": "get_weather", "arguments": {"city": "Pune"}}]
python
import jsonfrom transformers import AutoModelForCausalLM, AutoTokenizer repo = "SwapniL-pr1117/qwen2.5-0.5b-instruct-fc"tok = AutoTokenizer.from_pretrained(repo)model = AutoModelForCausalLM.from_pretrained(repo, dtype="auto", device_map="auto") SYSTEM_PROMPT = """<copy the system prompt above>"""tools = [{"name": "get_weather", "description": "Current weather for a city.", "parameters": {"city": {"type": "str", "description": "City name"}}}]user = (f"Available tools:\n\n{json.dumps(tools, indent=2)}\n\n" f"User request:\n\nWhat's the weather in Pune?\n\n" f"Return the required tool call(s) as JSON.") messages = [{"role": "system", "content": SYSTEM_PROMPT}, {"role": "user", "content": user}]inputs = tok.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt", return_dict=True).to(model.device)out = model.generate(**inputs, max_new_tokens=256) # greedy is the defaultprint(tok.decode(out[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True))bash
vllm serve SwapniL-pr1117/qwen2.5-0.5b-instruct-fc --max-model-len 4096Send the same two messages to /v1/chat/completions with temperature=0, then json.loads the reply.
bash
ollama run Avahan/qwen2.5_0.5b-instruct-fc-ChatNo system prompt is baked in: send the system prompt above yourself (plain ollama run gives the chat behaviour described below).
Use greedy decoding. generation_config.json already sets do_sample=False; with OpenAI-compatible
servers pass temperature=0. Sampling and repetition penalties hurt JSON accuracy.
[], but it returned [] in 0% of such cases.
It picks the closest tool and may invent argument values. Guard against this (e.g. check arguments appear in the request).QLoRA (4-bit base) with Unsloth, r=16, alpha=32, all linear layers, 1 epoch on ~48.7k xLAM examples, effective batch 16, lr 0.0002, max length 1536, loss on the answer only. LoRA merged into the bf16 base model.
You are a function-calling assistant.Your task is to select the appropriate tool and provide the arguments requiredto answer the user's request.Return ONLY valid JSON.The output must be a JSON array of tool calls in this format:[ { "name": "function_name", "arguments": { "argument_name": "value" } }]If none of the available tools can answer the request, return an empty array: []Do not explain your answer.Do not output markdown.Do not invent tools.Available tools:<tools as a JSON list: name, description, parameters>User request:<the request>Return the required tool call(s) as JSON.import jsonfrom transformers import AutoModelForCausalLM, AutoTokenizer repo = "SwapniL-pr1117/qwen2.5-0.5b-instruct-fc"tok = AutoTokenizer.from_pretrained(repo)model = AutoModelForCausalLM.from_pretrained(repo, dtype="auto", device_map="auto") SYSTEM_PROMPT = """<copy the system prompt above>"""tools = [{"name": "get_weather", "description": "Current weather for a city.", "parameters": {"city": {"type": "str", "description": "City name"}}}]user = (f"Available tools:\n\n{json.dumps(tools, indent=2)}\n\n" f"User request:\n\nWhat's the weather in Pune?\n\n" f"Return the required tool call(s) as JSON.") messages = [{"role": "system", "content": SYSTEM_PROMPT}, {"role": "user", "content": user}]inputs = tok.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt", return_dict=True).to(model.device)out = model.generate(**inputs, max_new_tokens=256) # greedy is the defaultprint(tok.decode(out[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True))