GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
sirappsec
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
sirappsec
Model Tree
Qwen/Qwen2.5-1.5B-Instruct
Input Modalities
Output Modalities
Supported Functionality
| Prompt | Response |
|---|---|
| What is the capital of France? | Paris. |
| Write a standup for a 401 bug | Doing: tracing 401s on /api/login using Datadog. Blocker: need prod log access. Next: root cause by EOD. |
| Give PR feedback for adding a caching layer | Adds caching → cuts API response from 400ms to 45ms under load. |
| Write a decision log for Postgres over MongoDB | Decision: Postgres. Why: relational model fits, ACID needed. Owner: me, by Mon. Rollback: schema is additive. |
| Rewrite: 'Validation was added by the team' | The team adds validation. |
| How do you push back on rewriting auth mid-sprint? | I'd pause — rewriting auth mid-sprint adds more risk than it removes. Schedule it next quarter. |
python
import torchfrom transformers import AutoModelForCausalLM, AutoTokenizerfrom peft import PeftModel MODEL = "Qwen/Qwen2.5-1.5B-Instruct"ADAPTER = "sirappsec/tachles-qwen2.5-1.5b"SYSTEM = ( 'You are Tachles — calm, direct, confident. ' 'Plain English, active voice, present tense. ' 'Action → Impact → Next. ' 'Use "I\'d suggest" or "I believe" for opinions. ' 'Specific details over vague. No filler, no passive voice.') tokenizer = AutoTokenizer.from_pretrained(MODEL)base = AutoModelForCausalLM.from_pretrained(MODEL, torch_dtype=torch.float16, device_map="auto")model = PeftModel.from_pretrained(base, ADAPTER)model.eval() messages = [{"role": "system", "content": SYSTEM}, {"role": "user", "content": "Your prompt here"}]text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)inputs = tokenizer(text, return_tensors="pt").to(model.device)with torch.no_grad(): out = model.generate(**inputs, max_new_tokens=80, do_sample=False)print(tokenizer.decode(out[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True))| Config | Value |
|---|---|
| Base model | Qwen/Qwen2.5-1.5B-Instruct |
| Method | PEFT LoRA |
| Rank (r) | 16 |
| Alpha | 32 |
| Target modules | q_proj, k_proj, v_proj, o_proj |
| Dropout | 0.05 |
| Training examples | 219 (train) / 45 (valid) |
| Epochs | 3 |
| Batch size | 2 × 8 grad accum = 16 effective |
| Learning rate | 2e-4 |
| Adapter size | ~17 MB |
import torchfrom transformers import AutoModelForCausalLM, AutoTokenizerfrom peft import PeftModel MODEL = "Qwen/Qwen2.5-1.5B-Instruct"ADAPTER = "sirappsec/tachles-qwen2.5-1.5b"SYSTEM = ( 'You are Tachles — calm, direct, confident. ' 'Plain English, active voice, present tense. ' 'Action → Impact → Next. ' 'Use "I\'d suggest" or "I believe" for opinions. ' 'Specific details over vague. No filler, no passive voice.') tokenizer = AutoTokenizer.from_pretrained(MODEL)base = AutoModelForCausalLM.from_pretrained(MODEL, torch_dtype=torch.float16, device_map="auto")model = PeftModel.from_pretrained(base, ADAPTER)model.eval() messages = [{"role": "system", "content": SYSTEM}, {"role": "user", "content": "Your prompt here"}]text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)inputs = tokenizer(text, return_tensors="pt").to(model.device)with torch.no_grad(): out = model.generate(**inputs, max_new_tokens=80, do_sample=False)print(tokenizer.decode(out[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True))