See axolotl config

axolotl version: 0.4.0

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

lora_fan_in_fan_out: false
data_seed: 49
seed: 49

datasets:
  - path: _synth_data/alpaca_synth_queries_healed.jsonl
    type: sharegpt
    conversation: alpaca
    shards: 10  # This will divide the dataset into 10 shards
    shards_idx: 2  # This will load only the 3rd shard (indexing starts from 0)
dataset_prepared_path: last_run_prepared
val_set_size: 0.1
output_dir: ./qlora-alpaca-out
hub_model_id: nisargvp/hc-mistral-alpaca-local

adapter: qlora
lora_model_dir:

sequence_len: 512
sample_packing: true
pad_to_sequence_len: true

lora_r: 32
lora_alpha: 16
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: hc-axolotl-mistral-local
wandb_entity: nisargvp

gradient_accumulation_steps: 4
micro_batch_size: 8
eval_batch_size: 16
num_epochs: 1
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.0003
max_grad_norm: 1.0
adam_beta2: 0.95
adam_epsilon: 0.00001
save_total_limit: 12

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

loss_watchdog_threshold: 5.0
loss_watchdog_patience: 3

warmup_steps: 10
evals_per_epoch: 4
eval_table_size:
eval_table_max_new_tokens: 128
saves_per_epoch: 6
debug:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:
  bos_token: "<s>"
  eos_token: "</s>"
  unk_token: "<unk>"
save_safetensors: true

hc-mistral-alpaca-local

This model is a fine-tuned version of mistralai/Mistral-7B-v0.1 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1059

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: 0.0003
• train_batch_size: 8
• eval_batch_size: 16
• seed: 49
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.95) and epsilon=1e-05
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 10
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss
1.1781	0.0061	1	1.2143
0.1752	0.2492	41	0.1504
0.1052	0.4985	82	0.1276
0.1015	0.7477	123	0.1089
0.1205	0.9970	164	0.1059

Framework versions

• PEFT 0.10.0
• Transformers 4.40.2
• Pytorch 2.3.0
• Datasets 2.19.1
• Tokenizers 0.19.1