typhoon-7b-en-to-th-lora

<!-- 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/OpenAccess-AI-Collective/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>

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

axolotl version: 0.4.0

base_model: ./models/scb10x_typhoon-7b
model_type: MistralForCausalLM
tokenizer_type: LlamaTokenizer
is_mistral_derived_model: true

load_in_8bit: false
load_in_4bit: true
strict: false


datasets:
  - path: ./work/scb-mt-en-th-2020/apdf.csv
    type:
      system_prompt: ""
      field_system: system
      field_instruction: en_text
      field_output: th_text
      format: "{instruction}<translate>"
  - path: ./work/scb-mt-en-th-2020/assorted_government.csv
    type:
      system_prompt: ""
      field_system: system
      field_instruction: en_text
      field_output: th_text
      format: "{instruction}<translate>"
  - path: ./work/scb-mt-en-th-2020/generated_reviews_crowd.csv
    type:
      system_prompt: ""
      field_system: system
      field_instruction: en_text
      field_output: th_text
      format: "{instruction}<translate>"
  - path: ./work/scb-mt-en-th-2020/generated_reviews_translator.csv
    type:
      system_prompt: ""
      field_system: system
      field_instruction: en_text
      field_output: th_text
      format: "{instruction}<translate>"
  - path: ./work/scb-mt-en-th-2020/generated_reviews_yn.csv
    type:
      system_prompt: ""
      field_system: system
      field_instruction: en_text
      field_output: th_text
      format: "{instruction}<translate>"
  - path: ./work/scb-mt-en-th-2020/mozilla_common_voice.csv
    type:
      system_prompt: ""
      field_system: system
      field_instruction: en_text
      field_output: th_text
      format: "{instruction}<translate>"
  - path: ./work/scb-mt-en-th-2020/msr_paraphrase.csv
    type:
      system_prompt: ""
      field_system: system
      field_instruction: en_text
      field_output: th_text
      format: "{instruction}<translate>"
  - path: ./work/scb-mt-en-th-2020/nus_sms.csv
    type:
      system_prompt: ""
      field_system: system
      field_instruction: en_text
      field_output: th_text
      format: "{instruction}<translate>"
  - path: ./work/scb-mt-en-th-2020/paracrawl.csv
    type:
      system_prompt: ""
      field_system: system
      field_instruction: en_text
      field_output: th_text
      format: "{instruction}<translate>"
  - path: ./work/scb-mt-en-th-2020/task_master_1.csv
    type:
      system_prompt: ""
      field_system: system
      field_instruction: en_text
      field_output: th_text
      format: "{instruction}<translate>"
  - path: ./work/scb-mt-en-th-2020/thai_websites.csv
    type:
      system_prompt: ""
      field_system: system
      field_instruction: en_text
      field_output: th_text
      format: "{instruction}<translate>"
  - path: ./work/scb-mt-en-th-2020/wikipedia.csv
    type:
      system_prompt: ""
      field_system: system
      field_instruction: en_text
      field_output: th_text
      format: "{instruction}<translate>"

dataset_prepared_path: ./work/last_run_prepared
val_set_size: 0.02
output_dir: ./work/out


adapter: qlora
lora_model_dir:

sequence_len: 4096
sample_packing: true
pad_to_sequence_len: true

gpu_memory_limit: 20

lora_r: 32
lora_alpha: 16
lora_dropout: 0.05
lora_target_linear: true
lora_fan_in_fan_out:


wandb_project: typhoon-7b
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 8
micro_batch_size: 2
num_epochs: 1
optimizer: paged_adamw_8bit
lr_scheduler: cosine
learning_rate: 0.0004

train_on_inputs: false
group_by_length: false
bf16: true
fp16: false
tf32: false

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


warmup_ratio: 0.01
eval_steps: 10
eval_table_size:
eval_table_max_new_tokens: 128
save_steps: 10
save_total_limit: 10
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:

</details><br>

ping98k/typhoon-7b-en-to-th-lora

This model was qlora finetuned on the scb_mt_enth_2020 dataset. It achieves the following results on the evaluation set:

• Loss: 0.8657

Model description

prompt

Why can camels survive for long without water?<translate>

output

ทำไมอูฐสามารถอยู่รอดได้นานโดยไม่มีน้ำ

known issue

model not train with end translate token correctly. some time model will output <translate> or </translate>

Why can camels survive for long without water?<translate>ทำไมอูฐสามารถอยู่รอดได้นานโดยไม่มีน้ำ<translate>

Why can camels survive for long without water?<translate>ทำไมอูฐสามารถอยู่รอดได้นานโดยไม่มีน้ำ</translate>

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: 0.0004
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 90
• num_epochs: 1

Training results

Framework versions

• PEFT 0.7.1
• Transformers 4.37.0
• Pytorch 2.0.1+cu118
• Datasets 2.16.1
• Tokenizers 0.15.0