Update Instruct-4b.yml

Instruct-4b.yml

@@ -1,8 +1,8 @@
-base_model: N
+base_model: NewEden/4B-PT
 model_type: AutoModelForCausalLM
 tokenizer_type: AutoTokenizer
 
-hub_model_id: NewEden/32b-rp
+hub_model_id: NewEden/4B-Inst
 hub_strategy: "all_checkpoints"
 push_dataset_to_hub:
 hf_use_auth_token: true
@@ -24,8 +24,6 @@ strict: false
 datasets:
   - path: NewEden/Helpsteer-3-Filtered
     type: dan-chat-advanced
-  - path: NewEden/No_Robots-R1-Filtered
-    type: dan-chat-advanced
   - path: NewEden/GSM8K-R1-filtered
     type: dan-chat-advanced
   - path: NewEden/Hydrus-R1-Thinking-Sharegpt
@@ -52,7 +50,7 @@ dataset_prepared_path: prepared_data
 val_set_size: 0.0
 output_dir: ./4b-inst
 
-sequence_len: 32768
+sequence_len: 16384
 sample_packing: true
 pad_to_sequence_len: true
 
@@ -63,10 +61,10 @@ wandb_watch:
 wandb_name: isnt-atmpt-1
 wandb_log_model:
 
-gradient_accumulation_steps: 4
-micro_batch_size: 6
+gradient_accumulation_steps: 2
+micro_batch_size: 1
 num_epochs: 2
-optimizer: adamw_bnb_8bit
+optimizer: paged_adamw_8bit
 lr_scheduler: cosine
 learning_rate: 2e-5
 max_grad_norm: 0.2