Views
No views yet
q_proj, k_proj, v_proj, o_proj) are trained; the full transformer backbone remains frozen, keeping the adapter extremely lightweight while steering the model's outputs toward coherent, context-aware responses across extended conversations.| Parameter | Value |
|---|---|
| PEFT type | LoRA |
Rank (r) | 16 |
| Alpha | 32 |
| Dropout | 0.05 |
| Target modules | q_proj, k_proj, v_proj, o_proj |
| Bias | none |
| Task type | CAUSAL_LM |
| PEFT version | 0.19.1 |
1from transformers import AutoModelForCausalLM, AutoTokenizer
2from peft import PeftModel
3
4base_model_id = "openai/gpt-oss-20b"
5adapter_id = "AdityaPS/SpaceLLM_Multi_turn"
6
7tokenizer = AutoTokenizer.from_pretrained(adapter_id)
8base_model = AutoModelForCausalLM.from_pretrained(base_model_id, device_map="auto")
9model = PeftModel.from_pretrained(base_model, adapter_id)
10
11messages = [
12 {"role": "user", "content": "Hi, can you help me plan a trip?"},
13 {"role": "assistant", "content": "Of course! Where are you thinking of going?"},
14 {"role": "user", "content": "Somewhere in Japan, maybe Tokyo."},
15]
16inputs = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt").to(model.device)
17outputs = model.generate(inputs, max_new_tokens=256)
18print(tokenizer.decode(outputs[0], skip_special_tokens=True))