Feat: Add cohere (commandr)

examples/cohere/lora.yml

@@ -0,0 +1,66 @@
+base_model: CohereForAI/c4ai-command-r-v01
+model_type: CohereForCausalLM
+tokenizer_type: CohereTokenizer
+
+load_in_8bit: true
+load_in_4bit: false
+strict: false
+
+datasets:
+  - path: mhenrichsen/alpaca_2k_test
+    type: alpaca
+dataset_prepared_path:
+val_set_size: 0.05
+output_dir: ./lora-out
+
+sequence_len: 4096
+sample_packing: true
+pad_to_sequence_len: true
+
+adapter: lora
+lora_model_dir:
+lora_r: 32
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_linear: true
+lora_fan_in_fan_out:
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+gradient_accumulation_steps: 4
+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:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+s2_attention:
+
+warmup_steps: 10
+evals_per_epoch: 4
+eval_table_size:
+eval_max_new_tokens: 128
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.0
+fsdp:
+fsdp_config:
+special_tokens:

src/axolotl/monkeypatch/multipack.py

@@ -1,4 +1,5 @@
 """multipack patching for v2 of sample packing"""
+
 import importlib
 
 import transformers
@@ -18,6 +19,7 @@
     "gemma",
     "gemmoe",
     "starcoder2",
+    "cohere",
 ]
 
 
@@ -56,6 +58,10 @@ def patch_for_multipack(model_type, model_name=None):
         patch_remote(model_name, ".configuration_gemmoe", ".modeling_gemmoe")
     elif model_type == "jamba":
         patch_remote(model_name, ".configuration_jamba", ".modeling_jamba")
+    elif model_type == "cohere":
+        transformers.models.cohere.modeling_cohere._get_unpad_data = (  # pylint: disable=protected-access
+            get_unpad_data
+        )
 
 
 def patch_remote(model_name, config_name, modeling_name):