Oprava

kody/mistral_sk_llamafactory_train.yaml

@@ -1,89 +0,0 @@
-top.booster: unsloth
-top.checkpoint_path: []
-top.finetuning_type: lora
-top.model_name: Custom
-top.quantization_bit: '4'
-top.quantization_method: bnb
-top.rope_scaling: none
-top.template: alpaca
-train.additional_target: ''
-train.apollo_rank: 16
-train.apollo_scale: 32
-train.apollo_target: all
-train.apollo_update_interval: 200
-train.badam_mode: layer
-train.badam_switch_interval: 50
-train.badam_switch_mode: ascending
-train.badam_update_ratio: 0.05
-train.batch_size: 1
-train.compute_type: fp16
-train.create_new_adapter: false
-train.cutoff_len: 1024
-train.dataset:
-- alpaca_slovak_cleaned
-train.dataset_dir: data
-train.ds_offload: false
-train.ds_stage: none
-train.enable_thinking: false
-train.extra_args: '{"optim": "adamw_8bit", "eval_steps": 1000, "eval_strategy": "steps",
-  "save_total_limit": 2}'
-train.freeze_extra_modules: ''
-train.freeze_language_model: false
-train.freeze_multi_modal_projector: true
-train.freeze_trainable_layers: 2
-train.freeze_trainable_modules: all
-train.freeze_vision_tower: true
-train.galore_rank: 16
-train.galore_scale: 2
-train.galore_target: all
-train.galore_update_interval: 200
-train.gradient_accumulation_steps: 8
-train.hub_private_repo: false
-train.image_max_pixels: 768*768
-train.image_min_pixels: 32*32
-train.learning_rate: 2e-4
-train.logging_steps: 5
-train.lora_alpha: 32
-train.lora_dropout: 0.05
-train.lora_rank: 16
-train.lora_target: q_proj,k_proj,v_proj,o_proj,gate_proj,up_proj,down_proj
-train.loraplus_lr_ratio: 0
-train.lr_scheduler_type: cosine
-train.mask_history: false
-train.max_grad_norm: '1.0'
-train.max_samples: '50000'
-train.neat_packing: false
-train.neftune_alpha: 0
-train.num_train_epochs: '1.0'
-train.packing: false
-train.ppo_score_norm: false
-train.ppo_whiten_rewards: false
-train.pref_beta: 0.1
-train.pref_ftx: 0
-train.pref_loss: sigmoid
-train.project: huggingface
-train.report_to: none
-train.resize_vocab: false
-train.reward_model: []
-train.save_steps: 1000
-train.swanlab_api_key: ''
-train.swanlab_link: null
-train.swanlab_mode: cloud
-train.swanlab_project: llamafactory
-train.swanlab_run_name: ''
-train.swanlab_workspace: ''
-train.trackio_space_id: trackio
-train.train_on_prompt: false
-train.training_stage: Supervised Fine-Tuning
-train.use_apollo: false
-train.use_badam: false
-train.use_dora: false
-train.use_galore: false
-train.use_llama_pro: false
-train.use_pissa: false
-train.use_rslora: false
-train.use_swanlab: false
-train.val_size: 0.025
-train.video_max_pixels: 256*256
-train.video_min_pixels: 16*16
-train.warmup_steps: 150

kody/training_args.yaml

@@ -0,0 +1,43 @@
+cutoff_len: 1024
+dataset: alpaca_slovak_cleaned
+dataset_dir: data
+ddp_timeout: 180000000
+do_train: true
+double_quantization: true
+enable_thinking: false
+eval_steps: 1000
+eval_strategy: steps
+finetuning_type: lora
+flash_attn: auto
+fp16: true
+gradient_accumulation_steps: 8
+include_num_input_tokens_seen: true
+learning_rate: 0.0002
+logging_steps: 5
+lora_alpha: 32
+lora_dropout: 0.05
+lora_rank: 16
+lora_target: q_proj,k_proj,v_proj,o_proj,gate_proj,up_proj,down_proj
+lr_scheduler_type: cosine
+max_grad_norm: 1.0
+max_samples: 50000
+model_name_or_path: slovak-nlp/mistral-sk-7b
+num_train_epochs: 1.0
+optim: adamw_8bit
+output_dir: /home/schwarc/diplomovka/mistral_sk_alpaca/llamafactory-full-lora
+packing: false
+per_device_eval_batch_size: 1
+per_device_train_batch_size: 1
+plot_loss: true
+preprocessing_num_workers: 16
+quantization_bit: 4
+quantization_method: bnb
+report_to: none
+save_steps: 1000
+save_total_limit: 2
+stage: sft
+template: alpaca
+trust_remote_code: true
+use_unsloth: true
+val_size: 0.025
+warmup_steps: 150