Llama 3.3 70B

docs/source/api_ref_models.rst

@@ -6,6 +6,31 @@ torchtune.models
 
 .. currentmodule:: torchtune.models
 
+llama3.3
+--------
+
+Text-only models from the 3.3 version of `Llama3 family <https://llama.meta.com/llama3/>`_.
+
+Important: You need to request access on `Hugging Face <https://huggingface.co/meta-llama/Llama-3.3-70B-Instruct>`__ before downloading it.
+
+To download the Llama-3.3-70B-Instruct model:
+
+.. code-block:: bash
+
+    tune download meta-llama/Llama-3.3-70B-Instruct --ignore-patterns "original/consolidated.00.pth" --hf-token <HF_TOKEN>
+
+.. autosummary::
+    :toctree: generated/
+    :nosignatures:
+
+    llama3_3.llama3_3_70b
+    llama3_3.lora_llama3_3_70b
+    llama3_3.qlora_llama3_3_70b
+
+.. note::
+
+    The Llama3.3 tokenizer reuses the :class:`~torchtune.models.llama3.llama3_tokenizer` class.
+
 llama3.2
 --------
 

recipes/configs/llama3_3/70B_full.yaml

@@ -0,0 +1,138 @@
+# Config for multi-device full finetuning in full_finetune_distributed.py
+# using a Llama3.3 70B Instruct model
+#
+# This config assumes that you've run the following command before launching
+# this run:
+#   tune download meta-llama/Llama-3.3-70B-Instruct --ignore-patterns "original/consolidated*"
+#
+# To launch on 8 devices, run the following command from root:
+#   tune run --nproc_per_node 8 full_finetune_distributed --config llama3_3/70B_full
+#
+# You can add specific overrides through the command line. For example
+# to override the checkpointer directory while launching training
+# you can run:
+#   tune run --nproc_per_node 8 full_finetune_distributed --config llama3_3/70B_full checkpointer.checkpoint_dir=<YOUR_CHECKPOINT_DIR>
+#
+# This config is only tested on an 8xA100 machine.
+#
+
+# Tokenizer
+tokenizer:
+  _component_: torchtune.models.llama3.llama3_tokenizer
+  path: /tmp/Llama-3.3-70B-Instruct/original/tokenizer.model
+  max_seq_len: null
+
+# Dataset
+dataset:
+  _component_: torchtune.datasets.alpaca_dataset
+  packed: False  # True increases speed
+seed: null
+shuffle: True
+
+# Model Arguments
+model:
+  _component_: torchtune.models.llama3_3.llama3_3_70b
+
+checkpointer:
+  _component_: torchtune.training.FullModelHFCheckpointer
+  checkpoint_dir: /tmp/Llama-3.3-70B-Instruct/
+  checkpoint_files: [
+    model-00001-of-00030.safetensors,
+    model-00002-of-00030.safetensors,
+    model-00003-of-00030.safetensors,
+    model-00004-of-00030.safetensors,
+    model-00005-of-00030.safetensors,
+    model-00006-of-00030.safetensors,
+    model-00007-of-00030.safetensors,
+    model-00008-of-00030.safetensors,
+    model-00009-of-00030.safetensors,
+    model-00010-of-00030.safetensors,
+    model-00011-of-00030.safetensors,
+    model-00012-of-00030.safetensors,
+    model-00013-of-00030.safetensors,
+    model-00014-of-00030.safetensors,
+    model-00015-of-00030.safetensors,
+    model-00016-of-00030.safetensors,
+    model-00017-of-00030.safetensors,
+    model-00018-of-00030.safetensors,
+    model-00019-of-00030.safetensors,
+    model-00020-of-00030.safetensors,
+    model-00021-of-00030.safetensors,
+    model-00022-of-00030.safetensors,
+    model-00023-of-00030.safetensors,
+    model-00024-of-00030.safetensors,
+    model-00025-of-00030.safetensors,
+    model-00026-of-00030.safetensors,
+    model-00027-of-00030.safetensors,
+    model-00028-of-00030.safetensors,
+    model-00029-of-00030.safetensors,
+    model-00030-of-00030.safetensors,
+  ]
+  recipe_checkpoint: null
+  output_dir: /tmp/Llama-3.3-70B-Instruct/
+  model_type: LLAMA3
+resume_from_checkpoint: False
+
+# Fine-tuning arguments
+batch_size: 2
+epochs: 1
+
+optimizer:
+  _component_: torch.optim.AdamW
+  lr: 2e-5
+  # Note: highly recommended to use fused=True optimizer flag
+  # with CPU offload for faster optimizer step.
+  fused: True
+
+loss:
+  _component_: torchtune.modules.loss.CEWithChunkedOutputLoss
+max_steps_per_epoch: null
+gradient_accumulation_steps: 1  # Use to increase virtual batch size
+
+
+# Training env
+device: cuda
+
+# Memory management
+enable_activation_checkpointing: True  # True reduces memory
+enable_activation_offloading: False  # True reduces memory
+custom_sharded_layers: ['tok_embeddings', 'output']  # Layers to shard separately (useful for large vocab size models). Lower Memory, but lower speed.
+fsdp_cpu_offload: True
+compile: False  # pytorch compile, set to true for better perf/memory
+optimizer_in_bwd: False  # True saves memory. Requires gradient_accumulation_steps=1
+
+# Reduced precision
+dtype: bf16
+
+# Logging
+metric_logger:
+  _component_: torchtune.training.metric_logging.DiskLogger
+  log_dir: ${output_dir}
+output_dir: /tmp/full-llama3_3-finetune
+log_every_n_steps: 1
+log_peak_memory_stats: True
+
+# Profiler (disabled)
+profiler:
+  _component_: torchtune.training.setup_torch_profiler
+  enabled: False
+
+  #Output directory of trace artifacts
+  output_dir: ${output_dir}/profiling_outputs
+
+  #`torch.profiler.ProfilerActivity` types to trace
+  cpu: True
+  cuda: True
+
+  #trace options passed to `torch.profiler.profile`
+  profile_memory: False
+  with_stack: False
+  record_shapes: True
+  with_flops: False
+
+  # `torch.profiler.schedule` options:
+  # wait_steps -> wait, warmup_steps -> warmup, active_steps -> active, num_cycles -> repeat
+  wait_steps: 5
+  warmup_steps: 3
+  active_steps: 2
+  num_cycles: 1

recipes/configs/llama3_3/70B_lora.yaml

@@ -0,0 +1,132 @@
+# Config for multi-device LoRA in lora_finetune_distributed.py
+# using a Llama3.3 70B model
+#
+# This config assumes that you've run the following command before launching
+# this run:
+#   tune download meta-llama/Llama-3.3-70B-Instruct --ignore-patterns "original/consolidated*"
+#
+# This config needs 8 GPUs to run
+#   tune run --nproc_per_node 8 lora_finetune_distributed --config llama3_3/70B_lora
+
+# Model Arguments
+model:
+  _component_: torchtune.models.llama3_3.lora_llama3_3_70b
+  lora_attn_modules: ['q_proj', 'v_proj', 'output_proj']
+  apply_lora_to_mlp: True
+  apply_lora_to_output: False
+  lora_rank: 16  # higher increases accuracy and memory
+  lora_alpha: 32  # usually alpha=2*rank
+  lora_dropout: 0.0
+
+tokenizer:
+  _component_: torchtune.models.llama3.llama3_tokenizer
+  path: /tmp/Llama-3.3-70B-Instruct/original/tokenizer.model
+  max_seq_len: null
+
+checkpointer:
+  _component_: torchtune.training.FullModelHFCheckpointer
+  checkpoint_dir: /tmp/Llama-3.3-70B-Instruct/
+  checkpoint_files: [
+    model-00001-of-00030.safetensors,
+    model-00002-of-00030.safetensors,
+    model-00003-of-00030.safetensors,
+    model-00004-of-00030.safetensors,
+    model-00005-of-00030.safetensors,
+    model-00006-of-00030.safetensors,
+    model-00007-of-00030.safetensors,
+    model-00008-of-00030.safetensors,
+    model-00009-of-00030.safetensors,
+    model-00010-of-00030.safetensors,
+    model-00011-of-00030.safetensors,
+    model-00012-of-00030.safetensors,
+    model-00013-of-00030.safetensors,
+    model-00014-of-00030.safetensors,
+    model-00015-of-00030.safetensors,
+    model-00016-of-00030.safetensors,
+    model-00017-of-00030.safetensors,
+    model-00018-of-00030.safetensors,
+    model-00019-of-00030.safetensors,
+    model-00020-of-00030.safetensors,
+    model-00021-of-00030.safetensors,
+    model-00022-of-00030.safetensors,
+    model-00023-of-00030.safetensors,
+    model-00024-of-00030.safetensors,
+    model-00025-of-00030.safetensors,
+    model-00026-of-00030.safetensors,
+    model-00027-of-00030.safetensors,
+    model-00028-of-00030.safetensors,
+    model-00029-of-00030.safetensors,
+    model-00030-of-00030.safetensors,
+  ]
+  recipe_checkpoint: null
+  output_dir: /tmp/Llama-3.3-70B-Instruct/
+  model_type: LLAMA3
+resume_from_checkpoint: False
+save_adapter_weights_only: True # Set to false to save the whole model + adapter merged
+
+# Dataset and Sampler
+dataset:
+  _component_: torchtune.datasets.alpaca_dataset
+  packed: False  # True increases speed
+seed: null
+shuffle: True
+batch_size: 2
+
+# Optimizer and Scheduler
+optimizer:
+  _component_: torch.optim.AdamW
+  fused: True
+  weight_decay: 0.01
+  lr: 3e-4
+lr_scheduler:
+  _component_: torchtune.training.lr_schedulers.get_cosine_schedule_with_warmup
+  num_warmup_steps: 100
+
+loss:
+  _component_: torchtune.modules.loss.CEWithChunkedOutputLoss
+
+# Training
+epochs: 1
+max_steps_per_epoch: null
+gradient_accumulation_steps: 1  # Use to increase virtual batch size
+compile: False  # pytorch compile, set to true for better perf/memory
+
+# Logging
+output_dir: /tmp/lora-llama3_3-finetune-output
+metric_logger:
+  _component_: torchtune.training.metric_logging.DiskLogger
+  log_dir: ${output_dir}
+log_every_n_steps: 1
+log_peak_memory_stats: True
+
+# Environment
+device: cuda
+dtype: bf16
+enable_activation_checkpointing: True  # True reduces memory
+enable_activation_offloading: False  # True reduces memory
+# custom_sharded_layers: ['tok_embeddings', 'output']  # Layers to shard separately (useful for large vocab size models). Lower Memory, but lower speed.
+
+# Profiler (disabled)
+profiler:
+  _component_: torchtune.training.setup_torch_profiler
+  enabled: False
+
+  #Output directory of trace artifacts
+  output_dir: ${output_dir}/profiling_outputs
+
+  #`torch.profiler.ProfilerActivity` types to trace
+  cpu: True
+  cuda: True
+
+  #trace options passed to `torch.profiler.profile`
+  profile_memory: False
+  with_stack: False
+  record_shapes: True
+  with_flops: False
+
+  # `torch.profiler.schedule` options:
+  # wait_steps -> wait, warmup_steps -> warmup, active_steps -> active, num_cycles -> repeat
+  wait_steps: 5
+  warmup_steps: 3
+  active_steps: 2
+  num_cycles: 1