t1

<!-- 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: 0.16.0.dev0

seed: 9
auto_resume_from_checkpoints: True
tokenizer_save_jinja_files: True
trust_remote_code: True
tokenizer_use_fast: True
gradient_accumulation: True
load_best_model_at_end: true

base_model: PicoKittens/PicoMistral-23M
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer

datasets: 
  - path: psychopenguin/indian_legal_dataset_qna
    type: alpaca
    split: train[:2%]


sequence_len: 256
fp16: True
adapter: lora
lora_target_linear: True
lora_r: 4
lora_alpha: 8
lora_dropout: 0.05


sdp_attention: True
optimizer: adamw_bnb_8bit
learning_rate: 0.0002
lr_scheduler: cosine
gradient_accumulation_steps: 2
micro_batch_size: 2 #for gpu memory increase
num_epochs: 3
neftune_noise_alpha: 5
early_stopping_patience: 3
save_steps: 100


val_set_size: 0.30
eval_strategy: steps
eval_steps: 100

use_wandb: True
wandb_project: tttt
wandb_name: t1

output_dir: ./final_model
merge_lora: True
hf_use_auth_token: True
hub_model_id: psychopenguin/t1

</details><br>

t1

This model is a fine-tuned version of PicoKittens/PicoMistral-23M on the psychopenguin/indian_legal_dataset_qna dataset. It achieves the following results on the evaluation set:

• Loss: 3.7133
• Ppl: 40.9894
• Memory/max Active (gib): 0.13
• Memory/max Allocated (gib): 0.13
• Memory/device Reserved (gib): 0.19

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: 0.0002
• train_batch_size: 2
• eval_batch_size: 2
• seed: 9
• gradient_accumulation_steps: 2
• total_train_batch_size: 4
• optimizer: Use OptimizerNames.ADAMW_BNB 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: 11
• training_steps: 382
• mixed_precision_training: Native AMP

Training results

Framework versions

• PEFT 0.18.1
• Transformers 5.3.0
• Pytorch 2.9.1+cu128
• Datasets 4.5.0
• Tokenizers 0.22.2