metadata
license: mit
base_model: croissantllm/CroissantLLMBase
tags:
- generated_from_trainer
model-index:
- name: out
results: []
See axolotl config
axolotl version: 0.4.0
base_model: croissantllm/CroissantLLMBase
model_type: LlamaForCausalLM
tokenizer_type: LlamaTokenizerFast
is_llama_derived_model: true
load_in_8bit: false
load_in_4bit: false
strict: false
datasets:
- path: manu/mmlu_auxiliary_train_formatted_2
split: train
type: completion
dataset_prepared_path: last_run_prepared
val_set_size: 0.05
output_dir: ./out
sequence_len: 2048
sample_packing: true
pad_to_sequence_len: true
adapter:
lora_model_dir:
lora_r:
lora_alpha:
lora_dropout:
lora_target_linear:
lora_fan_in_fan_out:
wandb_project:
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:
gradient_accumulation_steps: 4
micro_batch_size: 16
num_epochs: 2
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.0002
train_on_inputs: false
group_by_length: false
bf16: auto
fp16:
tf32: false
gradient_checkpointing: true
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true
flash_attn_cross_entropy: false
flash_attn_rms_norm: true
flash_attn_fuse_qkv: false
flash_attn_fuse_mlp: true
warmup_steps: 50
evals_per_epoch: 4
eval_table_size:
saves_per_epoch: 1
debug:
deepspeed: #deepspeed_configs/zero2.json # multi-gpu only
weight_decay: 0.1
fsdp:
fsdp_config:
special_tokens:
out
This model is a fine-tuned version of croissantllm/CroissantLLMBase on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.7378
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: 16
- eval_batch_size: 16
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 64
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 50
- num_epochs: 2
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 2.5429 | 0.0 | 1 | 2.5242 |
| 2.2283 | 0.25 | 69 | 2.2514 |
| 1.9539 | 0.5 | 138 | 2.0381 |
| 1.6608 | 0.75 | 207 | 1.6872 |
| 1.3767 | 1.0 | 276 | 1.3323 |
| 0.7872 | 1.23 | 345 | 1.0583 |
| 0.5873 | 1.48 | 414 | 0.8251 |
| 0.5154 | 1.73 | 483 | 0.7378 |
Framework versions
- Transformers 4.38.0.dev0
- Pytorch 2.1.2+cu121
- Datasets 2.16.1
- Tokenizers 0.15.0