minpeter / tiny-ko-187m-sft-250725

huggingface.co
Total runs: 15.4K
24-hour runs: 0
7-day runs: 0
30-day runs: 0
Model's Last Updated: 2025年7月30日
text-generation

Introduction of tiny-ko-187m-sft-250725

Model Details of tiny-ko-187m-sft-250725

Built with Axolotl

See axolotl config

axolotl version: 0.12.0.dev0

base_model: minpeter/tiny-ko-187m-base-250725

hub_model_id: minpeter/tiny-ko-187m-sft-250725
output_dir: ./outputs/tiny-ko-187m-sft-250725
wandb_project: "axolotl"
wandb_entity: "kasfiekfs-e"

model_type: LlamaForCausalLM
tokenizer_type: AutoTokenizer

strict: false

chat_template: chatml
datasets:
  - path: HuggingFaceTB/smol-smoltalk
    type: chat_template
    split: train
    field_messages: messages
    message_property_mappings:
      role: role
      content: content

  - path: trillionlabs/multisystem-curated
    type: chat_template
    split: train
    field_messages: messages
    message_property_mappings:
      role: role
      content: content

  - path: allenai/tulu-3-sft-personas-instruction-following
    type: chat_template
    split: train
    field_messages: messages
    message_property_mappings:
      role: role
      content: content

  - path: lemon-mint/smol-koreantalk
    type: chat_template
    split: train
    field_messages: messages
    message_property_mappings:
      role: role
      content: content

  - path: lemon-mint/Korean-FineTome-100k
    type: chat_template
    split: train
    field_messages: messages
    message_property_mappings:
      role: role
      content: content

  - path: heegyu/open-korean-instructions-v20231020
    type: chat_template
    split: train
    field_messages: conversations
    message_property_mappings:
      role: from
      content: value
    roles:
      user: ["human", "user"]
      assistant: ["gpt", "assistant", "bot"]
      system: ["system", "input"]

  - path: coastral/korean-writing-style-instruct
    type: chat_template
    split: train
    field_messages: conversations
    message_property_mappings:
      role: from
      content: value

  - path: devngho/korean-instruction-mix
    type: chat_template
    split: train
    field_messages: messages
    message_property_mappings:
      role: from
      content: value

dataset_prepared_path: last_run_prepared
val_set_size: 0.001
save_safetensors: true
sequence_len: 8192
sample_packing: false
pad_to_sequence_len: false
use_pose: true
pose_max_context_len: 65536

overrides_of_model_config:
  rope_theta: 1000000.0
  max_position_embeddings: 65536

gradient_accumulation_steps: 8
micro_batch_size: 16
num_epochs: 1
optimizer: muon
lr_scheduler: cosine
learning_rate: 3e-4

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

gradient_checkpointing: false
gradient_checkpointing_kwargs:
  use_reentrant: true
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true
sdp_attention:
s2_attention:

save_steps: 200
warmup_steps: 20
eval_steps: 200
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:

special_tokens:
  eos_token: '<|im_end|>'

plugins:
  - axolotl.integrations.cut_cross_entropy.CutCrossEntropyPlugin
  - axolotl.integrations.liger.LigerPlugin
#   - axolotl.integrations.lm_eval.LMEvalPlugin

# lm_eval_tasks:
#   - gsm8k
#   - hellaswag
#   - arc_easy
#   - arc_challenge
#   - piqa
#   - winogrande
#   - openbookqa
#   - wsc
#   - boolq

liger_rope: true
liger_rms_norm: true
liger_glu_activation: true
liger_layer_norm: true
liger_fused_linear_cross_entropy: true

tiny-ko-187m-sft-250725

This model is a fine-tuned version of minpeter/tiny-ko-187m-base-250725 on the HuggingFaceTB/smol-smoltalk, the trillionlabs/multisystem-curated, the allenai/tulu-3-sft-personas-instruction-following, the lemon-mint/smol-koreantalk, the lemon-mint/Korean-FineTome-100k, the heegyu/open-korean-instructions-v20231020, the coastral/korean-writing-style-instruct and the devngho/korean-instruction-mix datasets. It achieves the following results on the evaluation set:

  • Loss: 1.6810
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: 0.0003
  • train_batch_size: 16
  • eval_batch_size: 16
  • seed: 42
  • gradient_accumulation_steps: 8
  • total_train_batch_size: 128
  • optimizer: Use OptimizerNames.ADAMW_TORCH 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: 20
  • training_steps: 13880
Training results
Training Loss Epoch Step Validation Loss
No log 0 0 2.1583
2.0492 0.0144 200 1.9095
1.9084 0.0288 400 1.8642
1.8437 0.0432 600 1.8363
1.8601 0.0576 800 1.8160
1.8143 0.0720 1000 1.8001
1.7259 0.0865 1200 1.7871
1.8008 0.1009 1400 1.7764
1.7179 0.1153 1600 1.7668
1.8134 0.1297 1800 1.7587
1.7683 0.1441 2000 1.7518
1.8099 0.1585 2200 1.7458
1.7953 0.1729 2400 1.7397
1.7673 0.1873 2600 1.7339
1.721 0.2017 2800 1.7297
1.8727 0.2161 3000 1.7252
1.7334 0.2306 3200 1.7209
1.7078 0.2450 3400 1.7172
1.7751 0.2594 3600 1.7139
1.8221 0.2738 3800 1.7106
1.7385 0.2882 4000 1.7077
1.8214 0.3026 4200 1.7049
1.7042 0.3170 4400 1.7026
1.735 0.3314 4600 1.7005
1.647 0.3458 4800 1.6985
1.7222 0.3602 5000 1.6972
1.6963 0.3746 5200 1.6955
1.8047 0.3891 5400 1.6937
1.6151 0.4035 5600 1.6923
1.8008 0.4179 5800 1.6918
1.7152 0.4323 6000 1.6904
1.7522 0.4467 6200 1.6894
1.7645 0.4611 6400 1.6887
1.6721 0.4755 6600 1.6876
1.7343 0.4899 6800 1.6867
1.6748 0.5043 7000 1.6861
1.7417 0.5187 7200 1.6853
1.6245 0.5332 7400 1.6849
1.6081 0.5476 7600 1.6844
1.6696 0.5620 7800 1.6840
1.714 0.5764 8000 1.6836
1.8028 0.5908 8200 1.6832
1.6714 0.6052 8400 1.6828
1.7276 0.6196 8600 1.6827
1.6248 0.6340 8800 1.6823
1.7279 0.6484 9000 1.6823
1.7176 0.6628 9200 1.6820
1.8175 0.6773 9400 1.6819
1.7082 0.6917 9600 1.6817
1.7886 0.7061 9800 1.6815
1.7777 0.7205 10000 1.6814
1.7943 0.7349 10200 1.6814
1.79 0.7493 10400 1.6813
1.6757 0.7637 10600 1.6813
1.6708 0.7781 10800 1.6813
1.7519 0.7925 11000 1.6811
1.7789 0.8069 11200 1.6812
1.7562 0.8213 11400 1.6811
1.7137 0.8358 11600 1.6811
1.7407 0.8502 11800 1.6812
1.6515 0.8646 12000 1.6811
1.6929 0.8790 12200 1.6812
1.7125 0.8934 12400 1.6812
1.6112 0.9078 12600 1.6812
1.7437 0.9222 12800 1.6811
1.7824 0.9366 13000 1.6811
1.6166 0.9510 13200 1.6811
1.743 0.9654 13400 1.6812
1.6377 0.9799 13600 1.6812
1.7345 0.9943 13800 1.6810
Framework versions
  • Transformers 4.53.2
  • Pytorch 2.7.1+cu126
  • Datasets 4.0.0
  • Tokenizers 0.21.2

Runs of minpeter tiny-ko-187m-sft-250725 on huggingface.co

15.4K
Total runs
0
24-hour runs
0
3-day runs
0
7-day runs
0
30-day runs

More Information About tiny-ko-187m-sft-250725 huggingface.co Model

tiny-ko-187m-sft-250725 huggingface.co

tiny-ko-187m-sft-250725 huggingface.co is an AI model on huggingface.co that provides tiny-ko-187m-sft-250725's model effect (), which can be used instantly with this minpeter tiny-ko-187m-sft-250725 model. huggingface.co supports a free trial of the tiny-ko-187m-sft-250725 model, and also provides paid use of the tiny-ko-187m-sft-250725. Support call tiny-ko-187m-sft-250725 model through api, including Node.js, Python, http.

tiny-ko-187m-sft-250725 huggingface.co Url

https://huggingface.co/minpeter/tiny-ko-187m-sft-250725

minpeter tiny-ko-187m-sft-250725 online free

tiny-ko-187m-sft-250725 huggingface.co is an online trial and call api platform, which integrates tiny-ko-187m-sft-250725's modeling effects, including api services, and provides a free online trial of tiny-ko-187m-sft-250725, you can try tiny-ko-187m-sft-250725 online for free by clicking the link below.

minpeter tiny-ko-187m-sft-250725 online free url in huggingface.co:

https://huggingface.co/minpeter/tiny-ko-187m-sft-250725

tiny-ko-187m-sft-250725 install

tiny-ko-187m-sft-250725 is an open source model from GitHub that offers a free installation service, and any user can find tiny-ko-187m-sft-250725 on GitHub to install. At the same time, huggingface.co provides the effect of tiny-ko-187m-sft-250725 install, users can directly use tiny-ko-187m-sft-250725 installed effect in huggingface.co for debugging and trial. It also supports api for free installation.

tiny-ko-187m-sft-250725 install url in huggingface.co:

https://huggingface.co/minpeter/tiny-ko-187m-sft-250725

Url of tiny-ko-187m-sft-250725

tiny-ko-187m-sft-250725 huggingface.co Url

Provider of tiny-ko-187m-sft-250725 huggingface.co

minpeter
ORGANIZATIONS

Other API from minpeter

huggingface.co

Total runs: 253
Run Growth: 0
Growth Rate: 0.00%
Updated:2025年4月7日
huggingface.co

Total runs: 115
Run Growth: 0
Growth Rate: 0.00%
Updated:2025年6月6日
huggingface.co

Total runs: 20
Run Growth: -32
Growth Rate: -160.00%
Updated:2025年7月31日
huggingface.co

Total runs: 0
Run Growth: 0
Growth Rate: 0.00%
Updated:2024年9月16日