multipack for gemma (#1313)

* multipack for gemma

* chore: lint

* handle cache_position kwarg in updated llama modeling

* add position_ids to rotary embed call for updated llama modeling

examples/gemma/qlora.yml

@@ -1,49 +1,49 @@
 # use google/gemma-7b if you have access
-base_model: mhenrichsen/gemma-7b 
+base_model: mhenrichsen/gemma-7b
 model_type: AutoModelForCausalLM
 tokenizer_type: AutoTokenizer
- 
+
 load_in_8bit: false
 load_in_4bit: true
 strict: false
- 
+
 # huggingface repo
 datasets:
   - path: mhenrichsen/alpaca_2k_test
     type: alpaca
 val_set_size: 0.1
 output_dir: ./out
- 
+
 adapter: qlora
 lora_r: 32
 lora_alpha: 16
 lora_dropout: 0.05
 lora_target_linear: true
- 
+
 sequence_len: 4096
 sample_packing: false
 pad_to_sequence_len: false
- 
+
 wandb_project:
 wandb_entity:
 wandb_watch:
 wandb_name:
 wandb_log_model:
- 
- 
+
+
 gradient_accumulation_steps: 3
 micro_batch_size: 2
 num_epochs: 4
 optimizer: adamw_bnb_8bit
 lr_scheduler: cosine
 learning_rate: 0.0002
- 
+
 train_on_inputs: false
 group_by_length: false
 bf16: auto
 fp16:
 tf32: false
- 
+
 gradient_checkpointing: true
 early_stopping_patience:
 resume_from_checkpoint:
@@ -51,7 +51,7 @@ local_rank:
 logging_steps: 1
 xformers_attention:
 flash_attention: true
- 
+
 warmup_ratio: 0.1
 evals_per_epoch: 4
 eval_table_size: