See axolotl config

axolotl version: 0.6.0

base_model: Qwen/Qwen2.5-7B-Instruct
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer
trust_remote_code: true

load_in_8bit: false
load_in_4bit: true
strict: false

datasets:
  - path: aaditya/mimicraw_clinicaltrial_train
    type: alpaca
val_set_size: 0.05
output_dir: ./out

sequence_len: 4096
sample_packing: true
pad_to_sequence_len: true

adapter: qlora
lora_r: 256
lora_alpha: 512
lora_dropout: 0.05
lora_target_linear: true
lora_target_modules:
  - q_proj
  - k_proj
  - v_proj
  - o_proj
  - gate_proj
  - down_proj
  - up_proj

wandb_project: qwen_mimicrawclinicaltrail
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 4
micro_batch_size: 6
num_epochs: 3
optimizer: adamw_torch
lr_scheduler: cosine
learning_rate: 2e-6

train_on_inputs: false
group_by_length: false
bf16: auto
fp16: false
tf32: false

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

warmup_steps: 100
evals_per_epoch: 3
eval_table_size:
saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
save_total_limit: 4

out

This model is a fine-tuned version of Qwen/Qwen2.5-7B-Instruct on the aaditya/mimicraw_clinicaltrial_train dataset. It achieves the following results on the evaluation set:

• Loss: 0.6060

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-06
• train_batch_size: 6
• eval_batch_size: 6
• seed: 42
• distributed_type: multi-GPU
• gradient_accumulation_steps: 4
• total_train_batch_size: 24
• 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: 100
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss
0.8273	0.0008	1	0.8615
0.6312	0.3335	400	0.6677
0.6221	0.6671	800	0.6416
0.1335	1.0	1200	0.6267
0.6062	1.3327	1600	0.6176
0.5861	1.6662	2000	0.6119
0.6194	1.9998	2400	0.6084
0.5953	2.3319	2800	0.6068
0.6394	2.6654	3200	0.6060

Framework versions

• PEFT 0.14.0
• Transformers 4.48.1
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0