axolotl version: 0.4.1

base_model: meta-llama/Meta-Llama-3-8B
model_type: LlamaForCausalLM
tokenizer_type: AutoTokenizer

load_in_8bit: true
load_in_4bit: false
strict: false

datasets:
  - path: translation-dataset-v3-train.hf
    type: alpaca
    train_on_split: train

test_datasets:
  - path: translation-dataset-v3-test.hf
    type: alpaca
    split: train

dataset_prepared_path: ./last_run_prepared
output_dir: ./llama_3_translator
hub_model_id: ahmedsamirio/llama_3_translator_v3


sequence_len: 2048
sample_packing: true
pad_to_sequence_len: true
eval_sample_packing: false

adapter: lora
lora_r: 32
lora_alpha: 16
lora_dropout: 0.05
lora_target_linear: true
lora_fan_in_fan_out:
lora_target_modules:
  - gate_proj
  - down_proj
  - up_proj
  - q_proj
  - v_proj
  - k_proj
  - o_proj

wandb_project: en_eg_translator
wandb_entity: ahmedsamirio
wandb_name: llama_3_en_eg_translator_v3

gradient_accumulation_steps: 4
micro_batch_size: 2
num_epochs: 2
optimizer: paged_adamw_32bit
lr_scheduler: cosine
learning_rate: 2e-5

train_on_inputs: false
group_by_length: false
bf16: auto
fp16:
tf32: false

gradient_checkpointing: true
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true

warmup_steps: 10
evals_per_epoch: 10
eval_table_size:
eval_max_new_tokens: 128
saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:
  pad_token: <|end_of_text|>

Egyptian Arabic Translator Llama-3 8B

This model is a fine-tuned version of meta-llama/Meta-Llama-3-8B on the ahmedsamirio/oasst2-9k-translation dataset.

Model description

This model is an attempt to create a small translation model from English to Egyptian Arabic.

Intended uses & limitations

• Translating instruction finetuning and text generation datasets

Inference code

from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline

tokenizer = AutoTokenizer.from_pretrained("ahmedsamirio/Egyptian-Arabic-Translator-Llama-3-8B")
model = AutoModelForCausalLM.from_pretrained("ahmedsamirio/Egyptian-Arabic-Translator-Llama-3-8B")
pipe = pipeline(task='text-generation', model=model, tokenizer=tokenizer)


en_template = """<|begin_of_text|>Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.

### Instruction:
Translate the following text to English.

### Input:
{text}

### Response:
"""

ar_template = """<|begin_of_text|>Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.

### Instruction:
Translate the following text to Arabic.

### Input:
{text}

### Response:
"""

eg_template = """<|begin_of_text|>Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.

### Instruction:
Translate the following text to Egyptian Arabic.

### Input:
{text}

### Response:
"""

text = """Some habits are known as "keystone habits," and these influence the formation of other habits. \
For example, identifying as the type of person who takes care of their body and is in the habit of exercising regularly, \
can also influence eating better and using credit cards less. In business, \
safety can be a keystone habit that influences other habits that result in greater productivity.[17]"""

ar_text = pipe(ar_template.format(text=text), 
               max_new_tokens=256, 
               do_sample=True, 
               temperature=0.3, 
               top_p=0.5)


eg_text = pipe(eg_template.format(text=ar_text), 
               max_new_tokens=256, 
               do_sample=True, 
               temperature=0.3, 
               top_p=0.5)

print("Original Text:" text)
print("\nArabic Translation:", ar_text)
print("\nEgyptian Arabic Translation:", eg_text)

Training and evaluation data

ahmedsamirio/oasst2-9k-translation

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 10
• num_epochs: 2

Training results

Framework versions

• PEFT 0.11.1
• Transformers 4.42.3
• Pytorch 2.1.2+cu118
• Datasets 2.19.1
• Tokenizers 0.19.1