Hugging Face's logo

ping98k
/
typhoon-7b-rag-instruct-th

Text Generation
Transformers
Safetensors
Thai
mistral
Generated from Trainer
text-generation-inference
Inference Endpoints
Model card Files Community
2
typhoon-7b-rag-instruct-th / README.md
ping98k's picture
ping98k
Upload 11 files
1885b87 verified
preview code
|
raw
history blame
3.84 kB
metadata
license: apache-2.0
base_model: scb10x/typhoon-7b
tags:
  - generated_from_trainer
model-index:
  - name: out
    results: []

Built with Axolotl

See axolotl config

axolotl version: 0.3.0

base_model: scb10x/typhoon-7b
model_type: MistralForCausalLM
tokenizer_type: LlamaTokenizer
is_mistral_derived_model: true

load_in_8bit: false
load_in_4bit: false
strict: false

datasets:
  - path: finetune-data.jsonl
    type: alpaca
dataset_prepared_path: last_run_prepared
val_set_size: 0.05
output_dir: ./out

sequence_len: 8192
sample_packing: false
pad_to_sequence_len: true
eval_sample_packing: false

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

gradient_accumulation_steps: 8
micro_batch_size: 1
num_epochs: 4
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.000005

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

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

warmup_ratio: 0.05
evals_per_epoch: 5
eval_table_size:
eval_table_max_new_tokens: 128
saves_per_epoch: 5
save_total_limit: 10
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:
  bos_token: "<s>"
  eos_token: "</s>"
  unk_token: "<unk>"

out

This model is a fine-tuned version of scb10x/typhoon-7b on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 0.7682

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: 5e-06
  • train_batch_size: 1
  • eval_batch_size: 1
  • seed: 42
  • distributed_type: multi-GPU
  • num_devices: 2
  • gradient_accumulation_steps: 8
  • total_train_batch_size: 16
  • total_eval_batch_size: 2
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_steps: 23
  • num_epochs: 4

Training results

Training Loss Epoch Step Validation Loss
3.4821 0.0 1 4.2554
0.7752 0.2 48 0.7134
0.7287 0.41 96 0.6403
0.6135 0.61 144 0.6305
0.7828 0.81 192 0.6020
0.3375 1.02 240 0.5951
0.471 1.22 288 0.6191
0.2798 1.42 336 0.6249
0.5071 1.63 384 0.6213
0.2792 1.83 432 0.6176
0.069 2.03 480 0.6393
0.0742 2.23 528 0.6877
0.1309 2.44 576 0.6892
0.0349 2.64 624 0.6701
0.0639 2.84 672 0.6657
0.0273 3.05 720 0.6895
0.0311 3.25 768 0.7606
0.0307 3.45 816 0.7636
0.0791 3.66 864 0.7664
0.0747 3.86 912 0.7682

Framework versions

  • Transformers 4.37.0.dev0
  • Pytorch 2.0.1+cu118
  • Datasets 2.16.1
  • Tokenizers 0.15.0