typhoonv2-iot-lora-instruct

<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->

<img src="https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>

<details><summary>See axolotl config</summary>

axolotl version: 0.12.2

adapter: qlora
base_model: scb10x/llama-3-typhoon-v1.5-8b-instruct
bf16: true
chat_template: llama3
datasets:
- ds_type: json
  field_messages: messages
  message_property_mappings:
    content: content
    role: role
  path: iot_train_chat.json
  split: train
  type: chat_template
flash_attention: true
fp16: false
gradient_accumulation_steps: 2
gradient_checkpointing: true
learning_rate: 7.0e-05
load_in_4bit: true
logging_steps: 1
lora_alpha: 64
lora_dropout: 0.15
lora_r: 32
lora_target_modules:
- q_proj
- k_proj
- v_proj
- o_proj
- gate_proj
- up_proj
- down_proj
lr_scheduler: cosine
micro_batch_size: 8
model_type: LlamaForCausalLM
num_epochs: 2
optimizer: paged_adamw_8bit
output_dir: ./outputs/typhoon-iot-chat-format
pad_to_sequence_len: true
sample_packing: true
save_steps: 50
save_strategy: steps
sequence_len: 4096
special_tokens:
  bos_token: <|begin_of_text|>
  eos_token: <|end_of_text|>
  pad_token: <|end_of_text|>
tokenizer_type: AutoTokenizer
warmup_steps: 50
weight_decay: 0.015
xformers_attention: false

</details><br>

outputs/typhoon-iot-chat-format

This model is a fine-tuned version of scb10x/llama-3-typhoon-v1.5-8b-instruct on the iot_train_chat.json dataset.

Model description

More information needed

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: 7e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.PAGED_ADAMW_8BIT with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 50
• training_steps: 52

Training results

Framework versions

• PEFT 0.17.0
• Transformers 4.55.2
• Pytorch 2.5.1+cu124
• Datasets 4.0.0
• Tokenizers 0.21.4