tesy-0.3-hotfix

Uploaded finetuned model

• Developed by: CreitinGameplays
• License: apache-2.0
• Finetuned from model : unsloth/Llama-3.1-8B-Instruct

This llama model was trained 2x faster with Unsloth and Huggingface's TRL library.

<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>

Trained using the following parameters:

model = FastLanguageModel.get_peft_model(
    model,
    r = 16,
    target_modules = ["q_proj", "k_proj", "v_proj", "o_proj",
                      "gate_proj", "up_proj", "down_proj",],
    lora_alpha = 16,
    lora_dropout = 0,
    bias = "none",
    use_gradient_checkpointing = "unsloth",
    random_state = 3407,
    use_rslora = False,
    loftq_config = None,
)

training_args = TrainingArguments(
    per_device_train_batch_size = 16,
    gradient_accumulation_steps = 2,
    warmup_steps = 100,
    num_train_epochs = 2,
    learning_rate = 2e-4,
    fp16 = not torch.cuda.is_bf16_supported(),
    bf16 = torch.cuda.is_bf16_supported(),
    logging_steps = 10,
    optim = "adamw_8bit",
    weight_decay = 0.01,
    lr_scheduler_type = "linear",
    seed = 3407,
    output_dir = OUTPUT_DIR,
    report_to = "none",
    save_strategy = "steps",
    save_steps = 50,
    save_total_limit = 2,
    load_best_model_at_end = False,
)

trainer = SFTTrainer(
    model = model,
    tokenizer = tokenizer,
    train_dataset = dataset,
    dataset_text_field = "text",
    max_seq_length = max_seq_length,
    dataset_num_proc = 2,
    packing = False,
    args = training_args,
)

trainer = train_on_responses_only(
    trainer,
    instruction_part = "<|start_header_id|>user<|end_header_id|>\n\n", # llama
    response_part = "<|start_header_id|>assistant<|end_header_id|>\n\n",
)

Note: this model version was finetuned and saved with model.merge_and_unload() instead of model.save_pretrained_merged() to ensure better compatibility.