Hugging Face's logo

arcee-ai
/
raspberry-3B

Text Generation
Transformers
Safetensors
qwen2
Generated from Trainer
conversational
text-generation-inference
Inference Endpoints
Model card Files Community
4
raspberry-3B / README.md
qnguyen3's picture
qnguyen3
Update README.md
242d220 verified
preview code
|
raw
history blame
2.88 kB
metadata
library_name: transformers
license: other
license_name: qwen-research
license_link: https://huggingface.co/Qwen/Qwen2.5-3B-Instruct/blob/main/LICENSE
base_model: Qwen/Qwen2.5-3B
tags:
  - generated_from_trainer
model-index:
  - name: outputs/gelato-3b
    results: []

Built with Axolotl

See axolotl config

axolotl version: 0.4.1

base_model: Qwen/Qwen2.5-3B
load_in_8bit: false
load_in_4bit: false
strict: false

datasets:
  - path: arcee-ai/eval_tome
    type: sharegpt
    conversation: chatml
  - path: arcee-ai/math_code_5k_claude
    type: sharegpt
    conversation: chatml
    split: validation
  - path: Undi95/Capybara-ShareGPT
    type: sharegpt
    conversation: chatml
dataset_prepared_path:
val_set_size: 0.0

sequence_len: 8192
sample_packing: true

lora_fan_in_fan_out:
wandb_project: qwen2.5-3b-gelato
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:
output_dir: ./outputs/gelato-3b
gradient_accumulation_steps: 8
micro_batch_size: 2
num_epochs: 4
optimizer: adamw_bnb_8bit
torchdistx_path:
lr_scheduler: cosine
learning_rate: 0.0002
train_on_inputs: false
group_by_length: false
bf16: auto
fp16: 
tf32: true
gradient_checkpointing: true
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true
gptq_groupsize:
s2_attention:
gptq_model_v1:
warmup_steps: 50
evals_per_epoch:
saves_per_epoch: 1
debug:
deepspeed: deepspeed_configs/zero3_bf16_cpuoffload_params.json
weight_decay: 0.1
fsdp:
fsdp_config:
special_tokens:
  eos_token: "<|im_end|>"
  bos_token: "<|im_start|>"

outputs/gelato-3b

This model is a fine-tuned version of Qwen/Qwen2.5-3B on the None dataset.

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: 42
  • distributed_type: multi-GPU
  • num_devices: 4
  • gradient_accumulation_steps: 8
  • total_train_batch_size: 64
  • total_eval_batch_size: 8
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_steps: 50
  • num_epochs: 4

Training results

Framework versions

  • Transformers 4.45.1
  • Pytorch 2.3.1+cu121
  • Datasets 2.21.0
  • Tokenizers 0.20.0