---
license: cc-by-nc-4.0
base_model:
- unsloth/llama-3-8b-bnb-4bit
pipeline_tag: text-generation
tags:
- conversational
- llama
- ollama
- unsloth
- gguf
- 8b
---

# Training Process
## Model + LoRA Loading
```python
from unsloth import FastLanguageModel
import torch
max_seq_length = 2048
dtype = None # None for auto detection. Float16 for Tesla T4, V100, Bfloat16 for Ampere+
load_in_4bit = True # Use 4bit quantization to reduce memory usage. Can be False.

model, tokenizer = FastLanguageModel.from_pretrained(
    model_name = "unsloth/llama-3-8b-bnb-4bit",
    max_seq_length = max_seq_length,
    dtype = dtype,
    load_in_4bit = load_in_4bit,
    # token = "hf_...", # use one if using gated models like meta-llama/Llama-2-7b-hf
)

model = FastLanguageModel.get_peft_model(
    model,
    r = 16, # Choose any number > 0 ! Suggested 8, 16, 32, 64, 128
    target_modules = ["q_proj", "k_proj", "v_proj", "o_proj",
                      "gate_proj", "up_proj", "down_proj",],
    lora_alpha = 16,
    lora_dropout = 0, # Supports any, but = 0 is optimized
    bias = "none",    # Supports any, but = "none" is optimized
    use_gradient_checkpointing = "unsloth", # True or "unsloth" for very long context
    random_state = 3407,
    use_rslora = False,  # We support rank stabilized LoRA
    loftq_config = None, # And LoftQ
)
```

## Dataset Preparation
```python
from datasets import load_dataset
dataset = load_dataset(
    "csv",
    data_files = "/content/synth_data.csv",
    split = "train",
)

from unsloth import to_sharegpt
dataset = to_sharegpt(
    dataset,
    merged_prompt = "Labels: {available_entities}\n\nText: {text}\n",
    conversation_extension = 5, # Randomnly combines conversations into 1
    output_column_name = "label",
)

from unsloth import standardize_sharegpt
dataset = standardize_sharegpt(dataset)

chat_template = """{SYSTEM}
USER: {INPUT}
ASSISTANT: {OUTPUT}"""

from unsloth import apply_chat_template
dataset = apply_chat_template(
    dataset,
    tokenizer = tokenizer,
    chat_template = chat_template,
    default_system_message = "NER Task: Label the text based on the available Labels."
)
```

## Training Configuration
```python
from trl import SFTTrainer
from transformers import TrainingArguments
from unsloth import is_bfloat16_supported

trainer = SFTTrainer(
    model = model,
    tokenizer = tokenizer,
    train_dataset = dataset,
    dataset_text_field = "text",
    max_seq_length = max_seq_length,
    dataset_num_proc = 2,
    packing = True, # Can make training 5x faster for short sequences.
    args = TrainingArguments(
        per_device_train_batch_size = 2,
        gradient_accumulation_steps = 4,
        warmup_steps = 5,
        # max_steps = None,
        num_train_epochs = 1,
        learning_rate = 2e-4,
        fp16 = not is_bfloat16_supported(),
        bf16 = is_bfloat16_supported(),
        logging_steps = 1,
        optim = "adamw_8bit",
        weight_decay = 0.01,
        lr_scheduler_type = "linear",
        seed = 3407,
        output_dir = "outputs",
    ),
)

trainer_stats = trainer.train()

# Save to 8bit Q8_0
if False: model.save_pretrained_gguf("model", tokenizer,)
```

## Training Results
![image/png](https://cdn-uploads.huggingface.co/production/uploads/678fa336b0c8266724597a95/hcRhMrr2lC5f2b9enD_1z.png)
- Steps Trained: 26
- Final Loss: 0.1870
- Total Time: 21:04 min
- Full epoch had been 261 steps

# Sample Inference
```python
FastLanguageModel.for_inference(model) # Enable native 2x faster inference
messages = [                    # Change below!
    {"role": "user", "content": 'Labels: ATTR, CITY, CITY_PART, COUNTRY, O, ORG, PER, PHONE, REGION, REL, STREET, WORK_P, WORK_S\n\n'\
                                'Text: "doctors in berlin"'},
]
input_ids = tokenizer.apply_chat_template(
    messages,
    add_generation_prompt = True,
    return_tensors = "pt",
).to("cuda")

from transformers import TextStreamer
text_streamer = TextStreamer(tokenizer, skip_prompt = True)
_ = model.generate(input_ids, streamer = text_streamer, max_new_tokens = 128, pad_token_id = tokenizer.eos_token_id)
```