fix some of the edge cases for Jamba

examples/jamba/README.md

@@ -1,5 +1,10 @@
 # Jamba
 
-qlora w/ deepspeed needs at least 2x GPUs and 35GiB VRAM per GPU
-
-qlora single-gpu - training will start, but loss is off by an order of magnitude
+- ✅ qlora w/ deepspeed Zero-2 needs at least 2x GPUs and
+  - 35GiB VRAM per GPU w minimal context length
+  - 56GiB VRAM per GPU (w multipack enabled)
+- ✅ qlora w/ deepspeed Zero-3 needs at least 2x GPUs and 67GiB VRAM (wtf?)
+- ✅ qlora single-gpu, ~51GiB VRAM
+- ✅ multipack
+- ❓ FSDP
+- ❓ 8-bit LoRA

examples/jamba/qlora.yaml

@@ -0,0 +1,62 @@
+base_model: ai21labs/Jamba-v0.1
+trust_remote_code: true
+
+load_in_8bit: false
+load_in_4bit: true
+strict: false
+
+datasets:
+  - path: mhenrichsen/alpaca_2k_test
+    type: alpaca
+dataset_prepared_path:
+val_set_size: 0.0
+output_dir: ./out
+
+sequence_len: 4096
+sample_packing: false
+pad_to_sequence_len: false
+eval_sample_packing: false
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+adapter: qlora
+lora_r: 8
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_linear: true
+
+low_cpu_mem_usage: true
+gradient_accumulation_steps: 4
+micro_batch_size: 1
+num_epochs: 2
+optimizer: paged_adamw_8bit
+lr_scheduler: cosine
+learning_rate: 0.00001
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: false
+
+gradient_checkpointing: true
+gradient_checkpointing_kwargs:
+  use_reentrant: false
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+warmup_steps: 10
+evals_per_epoch:
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.0
+special_tokens:

src/axolotl/monkeypatch/multipack.py

@@ -48,14 +48,16 @@ def patch_for_multipack(model_type, model_name=None):
             get_unpad_data
         )
     elif model_type == "gemmoe":
-        model_config = AutoConfig.from_pretrained(model_name, trust_remote_code=True)
-        # we need to load the model here in order for modeling_gemmoe to be available
-        with init_empty_weights():
-            AutoModelForCausalLM.from_pretrained(model_name, trust_remote_code=True)
-        module_name = model_config.__class__.__module__.replace(
-            ".configuration_gemmoe", ".modeling_gemmoe"
-        )
-        modeling_gemmoe = importlib.import_module(module_name)
-        modeling_gemmoe._get_unpad_data = (  # pylint: disable=protected-access
-            get_unpad_data
-        )
+        patch_remote(model_name, ".configuration_gemmoe", ".modeling_gemmoe")
+    elif model_type == "jamba":
+        patch_remote(model_name, ".configuration_jamba", ".modeling_jamba")
+
+
+def patch_remote(model_name, config_name, modeling_name):
+    model_config = AutoConfig.from_pretrained(model_name, trust_remote_code=True)
+    # we need to load the model here in order for modeling_* to be available
+    with init_empty_weights():
+        AutoModelForCausalLM.from_pretrained(model_name, trust_remote_code=True)
+    module_name = model_config.__class__.__module__.replace(config_name, modeling_name)
+    modeling_arch = importlib.import_module(module_name)
+    modeling_arch._get_unpad_data = get_unpad_data  # pylint: disable=protected-access

src/axolotl/utils/models.py

@@ -456,6 +456,10 @@ def load_model(
             "bnb_4bit_quant_type": "nf4",
             "bnb_4bit_quant_storage": torch.bfloat16,
         }
+        if cfg.model_config_type == "jamba" and not cfg.deepspeed:
+            # for some reason, this causes the loss to be off by an order of magnitude
+            # but deepspeed needs this still in bfloat16
+            bnb_config["bnb_4bit_quant_storage"] = torch.float32
 
         if cfg.bnb_config_kwargs:
             bnb_config.update(cfg.bnb_config_kwargs)