GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
Rudraksh001
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
Rudraksh001
Model Tree
Qwen/Qwen2.5-0.5B-Instruct
Input Modalities
Output Modalities
Supported Functionality
| Property | Value |
|---|---|
| Base model | Qwen/Qwen2.5-0.5B-Instruct |
| Fine-tuning | LoRA / PEFT |
| LoRA rank | 16 |
| LoRA alpha | 32 |
| LoRA dropout | 0.05 |
| Target modules | q_proj, k_proj, v_proj, o_proj |
| Trainable parameters | 2,162,688 |
| Total parameters | 496,195,456 |
| Trainable percentage | 0.4359% |
| Training examples | 3,500 |
| Language | English |
This repository contains the LoRA adapter and tokenizer files, not the full base model.
Training was performed using a free Kaggle GPU environment.
The model was trained on the English subset of
JDhruv14/Bhagavad-Gita-QA.
The dataset was split by verse rather than by individual question.
The best checkpoint achieved a validation loss of 1.8104.
A fixed 50-question benchmark was also evaluated using the base model and the fine-tuned model.
| Model | Questions | Successful generations |
|---|---|---|
| Qwen2.5-0.5B-Instruct | 50 | 50/50 |
| Gita-Qwen-0.5B-V2 | 50 | 50/50 |
The benchmark evaluates open-ended responses and is not an exact-match accuracy benchmark.
This is an experimental fine-tune and may produce:
Model outputs should be independently verified against reliable translations, primary texts, or scholarly sources before being relied upon.
This model is intended for:
It is not intended to replace primary or scholarly sources.
python
import torchfrom transformers import AutoTokenizer, AutoModelForCausalLMfrom peft import PeftModel BASE_MODEL = "Qwen/Qwen2.5-0.5B-Instruct"ADAPTER = "Rudraksh001/gita-qwen-0.5b-v2" tokenizer = AutoTokenizer.from_pretrained(BASE_MODEL) base_model = AutoModelForCausalLM.from_pretrained( BASE_MODEL, torch_dtype=torch.float16, device_map="auto") model = PeftModel.from_pretrained( base_model, ADAPTER) model.eval() messages = [ { "role": "user", "content": ( "According to Bhagavad Gita Chapter 2, Verse 47:\n\n" "What is the central teaching of this verse?" ) }] inputs = tokenizer.apply_chat_template( messages, add_generation_prompt=True, return_tensors="pt") input_ids = inputs["input_ids"].to(model.device)attention_mask = inputs["attention_mask"].to(model.device) with torch.no_grad(): outputs = model.generate( input_ids=input_ids, attention_mask=attention_mask, max_new_tokens=150, do_sample=False, pad_token_id=tokenizer.eos_token_id ) answer = tokenizer.decode( outputs[0][input_ids.shape[-1]:], skip_special_tokens=True) print(answer)import torchfrom transformers import AutoTokenizer, AutoModelForCausalLMfrom peft import PeftModel BASE_MODEL = "Qwen/Qwen2.5-0.5B-Instruct"ADAPTER = "Rudraksh001/gita-qwen-0.5b-v2" tokenizer = AutoTokenizer.from_pretrained(BASE_MODEL) base_model = AutoModelForCausalLM.from_pretrained( BASE_MODEL, torch_dtype=torch.float16, device_map="auto") model = PeftModel.from_pretrained( base_model, ADAPTER) model.eval() messages = [ { "role": "user", "content": ( "According to Bhagavad Gita Chapter 2, Verse 47:\n\n" "What is the central teaching of this verse?" ) }] inputs = tokenizer.apply_chat_template( messages, add_generation_prompt=True, return_tensors="pt") input_ids = inputs["input_ids"].to(model.device)attention_mask = inputs["attention_mask"].to(model.device) with torch.no_grad(): outputs = model.generate( input_ids=input_ids, attention_mask=attention_mask, max_new_tokens=150, do_sample=False, pad_token_id=tokenizer.eos_token_id ) answer = tokenizer.decode( outputs[0][input_ids.shape[-1]:], skip_special_tokens=True) print(answer)