YoungPanda/continued_pretrian

See axolotl config

axolotl version: 0.4.0

base_model: Equall/Saul-Base
model_type: MistralForCausalLM
tokenizer_type: LlamaTokenizer

load_in_8bit: false
load_in_4bit: false
strict: false

datasets:
  - path: Drewskidang/chatlaw
    type: sharegpt
    conversation: chatml
  - path: Drewskidang/tool
    type: sharegpt
    conversation: chatml
  - path: digitalpipelines/samantha-1.1-uncensored
    type: sharegpt
    conversation: chatml
dataset_prepared_path:
val_set_size: 0.05
output_dir: ./out

sequence_len: 8192
sample_packing: true
pad_to_sequence_len: true
eval_sample_packing: false

wandb_project: mistral_chat
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 2
micro_batch_size: 6
num_epochs: 2
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.00005

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
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-4
warmup_steps: 100
evals_per_epoch: 4
eval_table_size:
eval_max_new_tokens: 128
saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:
  bos_token: "<s>"
  eos_token: "</s>"
  unk_token: "<unk>"
tokens: # these are delimiters
  - "<|im_start|>"
  - "<|im_end|>"

out

This model is a fine-tuned version of Equall/Saul-Base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.1824

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: 5e-05
• train_batch_size: 6
• eval_batch_size: 6
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 2
• total_train_batch_size: 96
• total_eval_batch_size: 48
• optimizer: Adam with betas=(0.9,0.999) and epsilon=0.0001
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 100
• num_epochs: 2

Training results

Training Loss	Epoch	Step	Validation Loss
1.5436	0.18	1	1.5013
1.5336	0.36	2	1.4939
1.4731	0.73	4	1.4126
1.3819	1.09	6	1.3104
1.312	1.27	8	1.2592
1.254	1.64	10	1.1824

Framework versions

• Transformers 4.38.2
• Pytorch 2.1.1+cu121
• Datasets 2.18.0
• Tokenizers 0.15.0