Add QAT support for distributed finetuning (pytorch#980)

recipes/configs/llama2/7B_qat_full.yaml

@@ -0,0 +1,76 @@
+# Config for multi-device QAT finetuning in qat_distributed.py
+# using a Llama2 7B model
+#
+# This config assumes that you've run the following command before launching
+# this run:
+#   tune download meta-llama/Llama-2-7b-hf --output-dir /tmp/Llama-2-7b-hf --hf-token <HF_TOKEN>
+#
+# To launch on 4 devices, run the following command from root:
+#   tune run --nnodes 1 --nproc_per_node 4 qat_distributed --config llama2/7B_qat_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 --nnodes 1 --nproc_per_node 4 qat_distributed --config llama2/7B_qat_full checkpointer.checkpoint_dir=<YOUR_CHECKPOINT_DIR>
+
+
+# Tokenizer
+tokenizer:
+  _component_: torchtune.models.llama2.llama2_tokenizer
+  path: /tmp/Llama-2-7b-hf/tokenizer.model
+
+# Dataset
+dataset:
+  _component_: torchtune.datasets.alpaca_dataset
+seed: null
+shuffle: True
+
+# Model Arguments
+model:
+  _component_: torchtune.models.llama2.llama2_7b
+
+checkpointer:
+  _component_: torchtune.utils.FullModelHFCheckpointer
+  checkpoint_dir: /tmp/Llama-2-7b-hf
+  checkpoint_files: [
+    pytorch_model-00001-of-00002.bin,
+    pytorch_model-00002-of-00002.bin
+  ]
+  recipe_checkpoint: null
+  output_dir: /tmp/Llama-2-7b-hf
+  model_type: LLAMA2
+resume_from_checkpoint: False
+
+# Fine-tuning arguments
+batch_size: 2
+epochs: 3
+optimizer:
+  _component_: torch.optim.AdamW
+  lr: 2e-5
+loss:
+  _component_: torch.nn.CrossEntropyLoss
+max_steps_per_epoch: null
+gradient_accumulation_steps: 1
+
+# QAT arguments
+quantizer:
+  _component_: torchtune.utils.quantization.Int8DynActInt4WeightQATQuantizer
+  groupsize: 256
+
+# Training env
+device: cuda
+
+# Memory management
+enable_activation_checkpointing: True
+memory_efficient_fsdp_wrap: False
+
+# Reduced precision
+dtype: bf16
+
+# Logging
+metric_logger:
+  _component_: torchtune.utils.metric_logging.DiskLogger
+  log_dir: ${output_dir}
+output_dir: /tmp/alpaca-llama2-finetune
+log_every_n_steps: 1
+log_peak_memory_stats: False

recipes/configs/llama3/8B_qat_full.yaml

@@ -0,0 +1,77 @@
+# Config for multi-device QAT finetuning in qat_distributed.py
+# using a Llama3 8B Instruct model
+#
+# This config assumes that you've run the following command before launching
+# this run:
+#   tune download meta-llama/Meta-Llama-3-8B-Instruct --output-dir /tmp/Meta-Llama-3-8B-Instruct --hf-token <HF_TOKEN>
+#
+# To launch on 4 devices, run the following command from root:
+#   tune run --nproc_per_node 4 qat_distributed --config llama3/8B_qat_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 4 qat_distributed --config llama3/8B_qat_full checkpointer.checkpoint_dir=<YOUR_CHECKPOINT_DIR>
+
+# Tokenizer
+tokenizer:
+  _component_: torchtune.models.llama3.llama3_tokenizer
+  path: /tmp/Meta-Llama-3-8B-Instruct/original/tokenizer.model
+
+# Dataset
+dataset:
+  _component_: torchtune.datasets.alpaca_dataset
+seed: null
+shuffle: True
+
+# Model Arguments
+model:
+  _component_: torchtune.models.llama3.llama3_8b
+
+checkpointer:
+  _component_: torchtune.utils.FullModelMetaCheckpointer
+  checkpoint_dir: /tmp/Meta-Llama-3-8B-Instruct/original/
+  checkpoint_files: [
+    consolidated.00.pth
+  ]
+  recipe_checkpoint: null
+  output_dir: /tmp/Meta-Llama-3-8B-Instruct/
+  model_type: LLAMA3
+resume_from_checkpoint: False
+
+# Fine-tuning arguments
+batch_size: 2
+epochs: 3
+
+# QAT arguments
+quantizer:
+  _component_: torchtune.utils.quantization.Int8DynActInt4WeightQATQuantizer
+  groupsize: 256
+
+optimizer:
+  _component_: torch.optim.AdamW
+  lr: 2e-5
+  foreach: False
+
+loss:
+  _component_: torch.nn.CrossEntropyLoss
+max_steps_per_epoch: null
+gradient_accumulation_steps: 1
+
+# Training env
+device: cuda
+
+# Memory management
+enable_activation_checkpointing: True
+memory_efficient_fsdp_wrap: True
+
+# Reduced precision
+dtype: bf16
+
+# Logging
+metric_logger:
+  _component_: torchtune.utils.metric_logging.DiskLogger
+  log_dir: ${output_dir}
+output_dir: /tmp/alpaca-llama3-finetune
+log_every_n_steps: 1
+log_peak_memory_stats: False