Update 4b-ad.yml

4b-ad.yml

@@ -2,7 +2,7 @@ base_model: NewEden/Hamanasu-KTO-4B
 model_type: AutoModelForCausalLM
 tokenizer_type: AutoTokenizer
 
-hub_model_id: NewEden/Hamanasu-4B-Adventure
+hub_model_id: NewEden/Hamanasu-4B-Adventur-new-lr-2
 hub_strategy: "all_checkpoints"
 push_dataset_to_hub:
 hf_use_auth_token: true
@@ -29,12 +29,6 @@ datasets:
     type: dan-chat-advanced
   - path: PocketDoc/Dans-Failuremaxx-Adventure-3
     type: dan-chat-advanced
-  - path: NewEden/Misc-Mang-Sharegpt
-    type: dan-chat-advanced
-  - path: PocketDoc/Dans-Failuremaxx-Adventure
-    type: dan-chat-advanced    
-  - path: PocketDoc/Dans-Failuremaxx-Adventure-2
-    type: dan-chat-advanced  
   - path: PocketDoc/Dans-Prosemaxx-InstructWriter-ZeroShot-2
     type: dan-chat-advanced  
   - path: PocketDoc/Dans-Prosemaxx-InstructWriter-ZeroShot-3
@@ -45,14 +39,14 @@ datasets:
     type: dan-chat-advanced  
   - path: PocketDoc/Dans-Prosemaxx-InstructWriter-Continue
     type: dan-chat-advanced  
-  - path: PocketDoc/Dans-Prosemaxx-InstructWriter-ZeroShot
-    type: dan-chat-advanced
+  - path: PocketDoc/Dans-Personamaxx-VN
+    type: dan-chat-advanced  
 shuffle_merged_datasets: true
 dataset_prepared_path: prepared_data
-val_set_size: 0.0
-output_dir: ./phi4-adventure
+val_set_size: 0.01
+output_dir: ./adventure
 
-sequence_len: 16384
+sequence_len: 32768
 sample_packing: true
 pad_to_sequence_len: true
 
@@ -69,7 +63,7 @@ micro_batch_size: 4
 num_epochs: 2
 optimizer: adamw_bnb_8bit
 lr_scheduler: cosine
-learning_rate: 2e-5
+learning_rate: 5e-6
 
 train_on_inputs: false
 group_by_length: false
@@ -91,7 +85,7 @@ eval_table_size:
 eval_max_new_tokens: 128
 saves_per_epoch: 2
 debug:
-deepspeed: ./deepspeed_configs/zero3_bf16_cpuoffload_params.json
-weight_decay: 0.01
+deepspeed: ./deepspeed_configs/zero3_bf16.json
+weight_decay: 0.02
 fsdp:
 fsdp_config: