Hugging Face's logo

Weyaxi
/
Einstein-7B

Text Generation
Transformers
PyTorch
mistral
axolotl
Generated from Trainer
text-generation-inference
Inference Endpoints
Model card Files Community
2
Einstein-7B / README.md
Weyaxi's picture
Weyaxi
Create README.md
36f6450 verified
preview code
|
raw
history blame
3.95 kB
metadata
license: other
tags:
  - axolotl
  - generated_from_trainer
base_model: mistralai/Mistral-7B-v0.1

Built with Axolotl

See axolotl config

axolotl version: 0.3.0

base_model: mistralai/Mistral-7B-v0.1
model_type: MistralForCausalLM
tokenizer_type: LlamaTokenizer
is_mistral_derived_model: true

load_in_8bit: false
load_in_4bit: true
strict: false

datasets:
  - path: sci-datasets/arc_challange_train_alpaca.json
    ds_type: json
    type: alpaca

  - path: sci-datasets/camelai_biology_alpaca.json
    ds_type: json
    type: alpaca

  - path: sci-datasets/camelai_chemistry_alpaca.json
    ds_type: json
    type: alpaca

  - path: sci-datasets/camelai_physics_alpaca.json
    ds_type: json
    type: alpaca

  - path: sci-datasets/openbookqa_alpaca.json
    ds_type: json
    type: alpaca

  - path: sci-datasets/reclor_science_alpaca.json
    ds_type: json
    type: alpaca
    
  - path: sci-datasets/scibench_alpaca.json
    ds_type: json
    type: alpaca

  - path: sci-datasets/scienceqa_alpaca.json
    ds_type: json
    type: alpaca

  - path: sci-datasets/theoremqa_alpaca.json
    ds_type: json
    type: alpaca

  - path: sci-datasets/tiger_scienceeval_alpaca.json
    ds_type: json
    type: alpaca

dataset_prepared_path: last_run_prepared
val_set_size: 0
output_dir: ./science-mistral

adapter: qlora
lora_model_dir:

sequence_len: 8192
sample_packing: true
pad_to_sequence_len: true

lora_r: 128
lora_alpha: 64
lora_dropout: 0.05
lora_target_linear: true
lora_fan_in_fan_out:
lora_target_modules:
  - gate_proj
  - down_proj
  - up_proj
  - q_proj
  - v_proj
  - k_proj
  - o_proj

wandb_project: huggingface
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:
hub_model_id: Weyaxi/science-mistral

# change #
gradient_accumulation_steps: 12
micro_batch_size: 6
num_epochs: 2
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.0002
# change #

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: 10


saves_per_epoch: 3
debug:
deepspeed:
weight_decay: 0.1
fsdp:
fsdp_config:
special_tokens:
  bos_token: "<s>"
  eos_token: "</s>"
  unk_token: "<unk>"

science-mistral

This model is a fine-tuned version of mistralai/Mistral-7B-v0.1 on the None dataset.

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

The following bitsandbytes quantization config was used during training:

  • quant_method: bitsandbytes
  • load_in_8bit: False
  • load_in_4bit: True
  • llm_int8_threshold: 6.0
  • llm_int8_skip_modules: None
  • llm_int8_enable_fp32_cpu_offload: False
  • llm_int8_has_fp16_weight: False
  • bnb_4bit_quant_type: nf4
  • bnb_4bit_use_double_quant: True
  • bnb_4bit_compute_dtype: bfloat16

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 0.0002
  • train_batch_size: 6
  • eval_batch_size: 6
  • seed: 42
  • gradient_accumulation_steps: 12
  • total_train_batch_size: 72
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_steps: 10
  • num_epochs: 2

Training results

Framework versions

  • PEFT 0.7.0
  • Transformers 4.37.0
  • Pytorch 2.0.1+cu118
  • Datasets 2.16.1
  • Tokenizers 0.15.0