shisa-ai/shisa-v1-qwen2-7b

See axolotl config

axolotl version: 0.4.0

base_model: Qwen/Qwen2-7B-Instruct

load_in_8bit: false
load_in_4bit: false
strict: false

chat_template: chatml
datasets:
  # This will be the path used for the data when it is saved to the Volume in the cloud.
  - path: augmxnt/ultra-orca-boros-en-ja-v1
    ds_type: json
    type: sharegpt
dataset_prepared_path: last_run_prepared
val_set_size: 0.05
output_dir: ./out

sequence_len: 8192
sample_packing: true
pad_to_sequence_len: true

neftune_noise_alpha: 5

use_wandb: true
wandb_project: shisa-v2
wandb_entity: augmxnt
wandb_name: shisa-v1-qwen2-7b

gradient_accumulation_steps: 8
micro_batch_size: 1
num_epochs: 3
optimizer: paged_adamw_8bit
lr_scheduler: linear
learning_rate: 8e-6

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_steps: 100
eval_per_epoch: 2 
eval_table_size:
saves_per_epoch: 0
save_steps:
debug:
deepspeed: /workspace/axolotl/deepspeed_configs/zero3_bf16.json
weight_decay: 0.01
fsdp:
fsdp_config:
special_tokens:
  pad_token: <|endoftext|>

out

This model is a fine-tuned version of Qwen/Qwen2-7B-Instruct on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.5239

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: 8e-06
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• total_eval_batch_size: 4
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss
0.8276	1.0196	319	0.5273
0.6577	2.0164	637	0.5103
0.5808	2.9541	936	0.5239

Framework versions

• Transformers 4.40.2
• Pytorch 2.2.2+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1