Support multiple KV Cache layouts based on dimension of K and V tensors

PiperOrigin-RevId: 719386925

Author: talumbau

ai_edge_torch/generative/examples/experimental/gemma/convert_gemma2_gpu_to_tflite.py

@@ -0,0 +1,114 @@
+# Copyright 2024 The AI Edge Torch Authors.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ==============================================================================
+
+"""Example of converting a Gemma2 model to multi-signature tflite model."""
+
+import os
+import pathlib
+
+from absl import app
+from absl import flags
+from ai_edge_torch.generative.examples.experimental.gemma import gemma2_gpu
+from ai_edge_torch.generative.layers.experimental import kv_cache
+from ai_edge_torch.generative.utilities import converter
+from ai_edge_torch.generative.utilities.model_builder import ExportConfig
+import torch
+
+_CHECKPOINT_PATH = flags.DEFINE_string(
+    'checkpoint_path',
+    os.path.join(pathlib.Path.home(), 'Downloads/llm_data/gemma2-2b'),
+    'The path to the model checkpoint, or directory holding the checkpoint.',
+)
+_OUTPUT_PATH = flags.DEFINE_string(
+    'output_path',
+    '/tmp/',
+    'The path to export the tflite model.',
+)
+_OUTPUT_NAME_PREFIX = flags.DEFINE_string(
+    'output_name_prefix',
+    'gemma2',
+    'The prefix of the output tflite model name.',
+)
+_PREFILL_SEQ_LENS = flags.DEFINE_multi_integer(
+    'prefill_seq_lens',
+    (8, 64, 128, 256, 512, 1024),
+    'List of the maximum sizes of prefill input tensors.',
+)
+_KV_CACHE_MAX_LEN = flags.DEFINE_integer(
+    'kv_cache_max_len',
+    1280,
+    'The maximum size of KV cache buffer, including both prefill and decode.',
+)
+_QUANTIZE = flags.DEFINE_bool(
+    'quantize',
+    True,
+    'Whether the model should be quantized.',
+)
+_LORA_RANKS = flags.DEFINE_multi_integer(
+    'lora_ranks',
+    None,
+    'If set, the model will be converted with the provided list of LoRA ranks.',
+)
+
+
+def _create_mask(mask_len, kv_cache_max_len):
+  mask = torch.full(
+      (mask_len, kv_cache_max_len), float('-inf'), dtype=torch.float32
+  )
+  mask = torch.triu(mask, diagonal=1).unsqueeze(0).unsqueeze(0)
+  return mask
+
+
+def _create_export_config(
+    prefill_seq_lens: list[int], kv_cache_max_len: int
+) -> ExportConfig:
+  """Creates the export config for the model."""
+  export_config = ExportConfig()
+  if isinstance(prefill_seq_lens, list):
+    prefill_mask = [_create_mask(i, kv_cache_max_len) for i in prefill_seq_lens]
+  else:
+    prefill_mask = _create_mask(prefill_seq_lens, kv_cach_max_len)
+
+  export_config.prefill_mask = prefill_mask
+
+  decode_mask = torch.full(
+      (1, kv_cache_max_len), float('-inf'), dtype=torch.float32
+  )
+  decode_mask = torch.triu(decode_mask, diagonal=1).unsqueeze(0).unsqueeze(0)
+  export_config.decode_mask = decode_mask
+  export_config.kvcache_cls = kv_cache.KVCacheTransposed
+  return export_config
+
+
+def main(_):
+  pytorch_model = gemma2_gpu.build_2b_model(
+      _CHECKPOINT_PATH.value, kv_cache_max_len=_KV_CACHE_MAX_LEN.value
+  )
+
+  converter.convert_to_tflite(
+      pytorch_model,
+      output_path=_OUTPUT_PATH.value,
+      output_name_prefix=_OUTPUT_NAME_PREFIX.value,
+      prefill_seq_len=_PREFILL_SEQ_LENS.value,
+      quantize=_QUANTIZE.value,
+      lora_ranks=_LORA_RANKS.value,
+      export_config=_create_export_config(
+          _PREFILL_SEQ_LENS.value, _KV_CACHE_MAX_LEN.value
+      ),
+  )
+
+
+if __name__ == '__main__':
+  app.run(main)