Views
No views yet
theblackcat102/evol-codealpaca-v1 dataset, scored decent on locally run perplexity at 4.36.1base_model: NousResearch/Llama-2-13b-hf
2base_model_config: NousResearch/Llama-2-13b-hf
3model_type: LlamaForCausalLM
4tokenizer_type: LlamaTokenizer
5push_dataset_to_hub:
6hub_model_id:
7
8load_in_8bit: false
9load_in_4bit: true
10strict: false
11
12datasets:
13 - path: theblackcat102/evol-codealpaca-v1
14 type: alpaca
15dataset_prepared_path: last_run_prepared
16val_set_size: 0.01
17output_dir: ./checkpoints/llama-2-13b-qlora
18
19adapter: qlora
20lora_model_dir:
21
22sequence_len: 4096
23max_packed_sequence_len: 4096
24lora_r: 32
25lora_alpha: 16
26lora_dropout: 0.05
27lora_target_modules:
28lora_target_linear: true
29lora_fan_in_fan_out:
30
31wandb_project:
32wandb_watch:
33wandb_run_id:
34wandb_log_model:
35
36gradient_accumulation_steps: 2
37micro_batch_size: 2
38num_epochs: 3
39optimizer: paged_adamw_32bit
40lr_scheduler: cosine
41learning_rate: 0.0001
42
43train_on_inputs: false
44group_by_length: true
45bf16: true
46fp16: false
47tf32: true
48
49gradient_checkpointing: true
50early_stopping_patience:
51resume_from_checkpoint:
52local_rank:
53logging_steps: 1
54xformers_attention: true
55flash_attention:
56
57warmup_steps: 10
58eval_steps: 50
59save_steps:
60debug:
61deepspeed:
62weight_decay: 0.0
63fsdp:
64fsdp_config:
65special_tokens:
66 bos_token: "<s>"
67 eos_token: "</s>"
68 unk_token: "<unk>"accelerate launch scripts/finetune.py configs/your_config.yml --merge_lora --lora_model_dir="./completed-model" --load_in_8bit=False --load_in_4bit=Falsebitsandbytes quantization config was used during training:bitsandbytes quantization config was used during training:bitsandbytes quantization config was used during training:bitsandbytes quantization config was used during training: