license: apache-2.0
base_model: TinyLlama/TinyLlama-1.1B-Chat-v1.0
tags:
- trl
- sft
- generated_from_trainer
model-index:
- name: TinyLlama_instruct_generation
results: []
TinyLlama_instruct_generation
This model is a fine-tuned version of TinyLlama/TinyLlama-1.1B-Chat-v1.0 on the generator dataset.
Model description
This model has been fine tuned with mosaicml/instruct-v3 dataset with 2 epoch only. Mainly this model is useful for RAG based application
How to use?
from peft import PeftModel
#load the base model
model_path = "TinyLlama/TinyLlama-1.1B-Chat-v1.0"
tokenizer=AutoTokenizer.from_pretrained(model_path)
model = AutoModelForCausalLM.from_pretrained( model_path, torch_dtype = torch.bfloat16, device_map = "auto", trust_remote_code = True )
#load the adapter
model_peft = PeftModel.from_pretrained(model, "azam25/TinyLlama_instruct_generation")
messages = [{
"role": "user",
"content": "Act as a gourmet chef. I have a friend coming over who is a vegetarian.
I want to impress my friend with a special vegetarian dish.
What do you recommend?
Give me two options, along with the whole recipe for each"
}]
def generate_response(message, model):
prompt = tokenizer.apply_chat_template(messages, tokenize=False) encoded_input = tokenizer(prompt, return_tensors="pt", add_special_tokens=True) model_inputs = encoded_input.to('cuda') generated_ids = model.generate(**model_inputs, max_new_tokens=1000, do_sample=True, pad_token_id=tokenizer.eos_token_id) decoded_output = tokenizer.batch_decode(generated_ids) return decoded_output[0]
response = generate_response(messages, model) print(response)
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0002
- train_batch_size: 4
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: constant
- lr_scheduler_warmup_steps: 0.03
- num_epochs: 2
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 1.6386 | 1.0 | 25 | 1.4451 |
| 1.5234 | 2.0 | 50 | 1.3735 |
Framework versions
- Transformers 4.35.2
- Pytorch 2.1.0+cu121
- Datasets 2.16.1
- Tokenizers 0.15.0