base_model: mistralai/Mistral-7B-v0.1
# optionally might have model_type or tokenizer_type
model_type: AutoModelForCausalLM
tokenizer_type: LlamaTokenizer
# Automatically upload checkpoint and final model to HF
hub_model_id: AiAF/UFOs-Mistral-7B-v0.1-V4

trust_remote_code: true

load_in_8bit: false
load_in_4bit: true
strict: false

#chat_template: none
datasets:
  - path: AiAF/input_output_master_list.json
    type: input_output
train_on_inputs: True



# datasets:
#   - path: AiAF/master_list_jsonl
#     type: chat_template

#     field_messages: conversations
#     message_field_role: from
#     message_field_content: value


dataset_prepared_path: last_run_prepared
val_set_size: 0.1
output_dir: ./outputs/qlora-out

## You can optionally freeze the entire model and unfreeze a subset of parameters
unfrozen_parameters:
#  - ^lm_head.weight$
#  - ^model.embed_tokens.weight$[:32000]
#  - model.layers.2[0-9]+.block_sparse_moe.gate
#  - model.layers.2[0-9]+.block_sparse_moe.experts
#  - model.layers.3[0-9]+.block_sparse_moe.gate
#  - model.layers.3[0-9]+.block_sparse_moe.experts

model_config:
  output_router_logits: true

adapter: qlora
lora_model_dir:

sequence_len: 2048
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
#  - q_proj
#  - k_proj
#  - v_proj
#  - o_proj
#  - w1
#  - w2
#  - w3

wandb_project:
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 2
micro_batch_size: 1
num_epochs: 1
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.0002

UFOs-Mistral-7B-v0.1-V4

This model is a fine-tuned version of mistralai/Mistral-7B-v0.1 on the AiAF/input_output_master_list.json dataset. It achieves the following results on the evaluation set:

Loss: 2.0053

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: 1
eval_batch_size: 1
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 2
optimizer: Use OptimizerNames.ADAMW_BNB with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: cosine
lr_scheduler_warmup_steps: 14
num_epochs: 1.0

Training results

Training Loss	Epoch	Step	Validation Loss
0.755	1.0	500	2.0053

Framework versions

PEFT 0.14.0
Transformers 4.48.1
Pytorch 2.5.1+cu124
Datasets 3.2.0
Tokenizers 0.21.0