abdulmannan-01/DeepSeek-R1-Distill-Llama-8B-Lora-Finetuned-OS-Dataset

See axolotl config

axolotl version: 0.6.0

adapter: lora
base_model: deepseek-ai/DeepSeek-R1-Distill-Llama-8B
bf16: auto
chat_template: llama3
dataset_prepared_path: last_run_prepared
datasets:
- field_message: messages
  message_field_content: content
  message_field_role: role
  path: OpenSciLM/OS_Train_Data
  drop_system_message: true
  split: train
  type: chat_template
debug: null
deepspeed: null
early_stopping_patience: null
eval_sample_packing: true
eval_table_size: null
evals_per_epoch: 1
flash_attention: false
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 2
gradient_checkpointing: true
group_by_length: false
learning_rate: 2e-5
load_in_4bit: false
load_in_8bit: true
logging_steps: 1
lora_alpha: 32
lora_dropout: 0.05
lora_fan_in_fan_out: null
lora_model_dir: null
lora_modules_to_save:
- embed_tokens
- lm_head
lora_r: 64
lora_target_linear: true
lr_scheduler: cosine
max_steps: 8000
micro_batch_size: 1
model_type: LlamaForCausalLM
num_epochs: 1
optimizer: paged_adamw_8bit
output_dir: ./outputs/deepseek-R1-distill-llama-8B-openscholar-data-lora-8000-1epoch
pad_to_sequence_len: true
resume_from_checkpoint: null
sample_packing: true
saves_per_epoch: 1
sdp_attention: true
sequence_len: 4096 #2048
special_tokens:
  pad_token: <|end_of_text|>
strict: false
tf32: false
tokenizer_type: AutoTokenizer
train_on_inputs: false
val_set_size: 0.05
wandb_entity: null
wandb_log_model: null
wandb_name: null
wandb_project: null
wandb_watch: null
warmup_steps: 1
weight_decay: 0.0
xformers_attention: null

outputs/deepseek-R1-distill-llama-8B-openscholar-data-lora-8000-1epoch

This model is a fine-tuned version of deepseek-ai/DeepSeek-R1-Distill-Llama-8B on the OpenSciLM/OS_Train_Data dataset. It achieves the following results on the evaluation set:

• Loss: 0.3808

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: 2e-05
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 2
• 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: 2
• training_steps: 8000

Training results

Training Loss	Epoch	Step	Validation Loss
0.7391	0.3814	8000	0.3808

Framework versions

• PEFT 0.14.0
• Transformers 4.47.1
• Pytorch 2.5.1+cu124
• Datasets 3.1.0
• Tokenizers 0.21.0