Downloads · 30 days
6
9% of all-time downloads
BryceBlankinship/fitness-coach-qwen3-14b-lora
fitness-coach-qwen3-14b-lora is a text generation model from BryceBlankinship. Use it when you need the model to write or continue text. It is set up for peft. The card lists the license as apache-2.0.
QLoRA adapter for a fitness-coach chat model. Continued pretraining on exercise-science text, then SFT on coach chats. Same LoRA throughout; this repo is the SFT snapshot.
Downloads · 30 days
6
9% of all-time downloads
All-time downloads
69
Public
Repo size
2.1 GB
Likes
0
Public
Click a slice to open those files.
.safetensors2.1 GB · 99%
From the Hugging Face model README
QLoRA adapter for a fitness-coach chat model. Continued pretraining on exercise-science text, then SFT on coach chats. Same LoRA throughout; this repo is the SFT snapshot.
Load this adapter on Qwen3-14B 4-bit Unsloth base.
unsloth/Qwen3-14B-Base-unsloth-bnb-4bit<|im_start|> / <|im_end|>)from unsloth import FastLanguageModel
model, tokenizer = FastLanguageModel.from_pretrained(
model_name="BryceBlankinship/fitness-coach-qwen3-14b-lora",
max_seq_length=4096,
load_in_4bit=True,
)
FastLanguageModel.for_inference(model)
system = (
"You are an expert fitness coach. Answer only as a fitness coach. "
"Give evidence-based, practical guidance on training, recovery, nutrition, "
"and injury-aware programming. Stay in character. If a request is outside "
"fitness, coaching, exercise science, or related health-and-performance "
"topics, briefly refuse and redirect to fitness. Do not diagnose medical "
"conditions. For sharp pain, injury, or medical questions, tell the user "
"to see a qualified clinician."
)
messages = [
{"role": "system", "content": system},
{"role": "user", "content": "Build a 3-day full-body plan for a busy beginner."},
]
text = tokenizer.apply_chat_template(
messages, tokenize=False, add_generation_prompt=True, enable_thinking=False
)
inputs = tokenizer(text, return_tensors="pt").to(model.device)
out = model.generate(**inputs, max_new_tokens=512, temperature=0.7, top_p=0.8)
print(tokenizer.decode(out[0], skip_special_tokens=True))
from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
from peft import PeftModel
bnb = BitsAndBytesConfig(load_in_4bit=True, bnb_4bit_compute_dtype="bfloat16")
base = AutoModelForCausalLM.from_pretrained(
"unsloth/Qwen3-14B-Base-unsloth-bnb-4bit",
quantization_config=bnb,
device_map="auto",
)
model = PeftModel.from_pretrained(base, "BryceBlankinship/fitness-coach-qwen3-14b-lora")
tokenizer = AutoTokenizer.from_pretrained("BryceBlankinship/fitness-coach-qwen3-14b-lora")
Send the coach system prompt yourself. The adapter does not inject it automatically.
This is not a medical model. It should refuse diagnosis and send injury / sharp-pain questions to a clinician.