Support pytorch checkpoints for Qwen3/Phi4 mini (#13984)

Author: metascroy

examples/models/checkpoint.py

@@ -7,6 +7,8 @@
 
 # pyre-unsafe
 
+import json
+import os
 from pathlib import Path
 from typing import Any, Dict, Optional
 
@@ -74,3 +76,39 @@ def get_checkpoint_dtype(checkpoint: Dict[str, Any]) -> Optional[torch.dtype]:
                 f"Mixed dtype model. Dtype of {first_key}: {first.dtype}. Mismatches in the checkpoint: {mismatched_dtypes}"
             )
     return dtype
+
+
+def load_checkpoint_from_pytorch_model(input_dir: str) -> Dict:
+    index_path = os.path.join(input_dir, "pytorch_model.bin.index.json")
+    if os.path.exists(index_path):
+        # Sharded checkpoint.
+        with open(index_path, "r") as f:
+            index = json.load(f)
+        weight_map = index["weight_map"]
+        checkpoint_shards = sorted(set(weight_map.values()))
+
+        # Load all the shards into memory
+        shard_to_weights = {}
+        for shard in checkpoint_shards:
+            shard_to_weights[shard] = torch.load(
+                os.path.join(input_dir, shard),
+                weights_only=True,
+                map_location=torch.device("cpu"),
+            )
+
+        # Merge tensors into consolidated state dict.
+        merged_state_dict = {}
+        for weight_name, shard in weight_map.items():
+            tensor = shard_to_weights[shard][weight_name]
+            merged_state_dict[weight_name] = tensor
+        return merged_state_dict
+
+    # Single checkpoint
+    model_path = os.path.join(input_dir, "pytorch_model.bin")
+    if os.path.exists(model_path):
+        state_dict = torch.load(
+            model_path, weights_only=True, map_location=torch.device("cpu")
+        )
+        return state_dict
+
+    raise FileNotFoundError(f"Could not find pytorch_model checkpoint in {input_dir}")

examples/models/phi_4_mini/convert_weights.py

@@ -1,8 +1,8 @@
 import argparse
-import os
 from typing import Dict
 
 import torch
+from executorch.examples.models.checkpoint import load_checkpoint_from_pytorch_model
 
 from torchtune.models.convert_weights import get_mapped_key
 
@@ -87,10 +87,8 @@ def phi_4_tune_to_meta(state_dict: Dict[str, torch.Tensor]) -> Dict[str, torch.T
     Convert a state dict from torchtune's format to Meta's format. This function
     doesn't handle any sharding or splitting of state dicts. It follows the
     state_dict IN -> state_dict OUT pattern.
-
     Args:
         state_dict (Dict[str, torch.Tensor]): State dict in torchtune's format.
-
     Returns:
         Dict[str, torch.Tensor]: State dict in Meta's format.
     """
@@ -105,14 +103,15 @@ def phi_4_tune_to_meta(state_dict: Dict[str, torch.Tensor]) -> Dict[str, torch.T
     converted_state_dict["output.weight"] = converted_state_dict[
         "tok_embeddings.weight"
     ]
-
     return converted_state_dict
 
 
 def convert_weights(input_dir_or_checkpoint: str, output_file: str) -> None:
-    # If input_dir_or_checkpoint is a directory downloaded from HF, FullModelHFCheckpointer is used to extract the state dict
-    # If input_dir_or_checkpoint is a checkpoint (from eager model model), it is loaded directly
-    if os.path.isdir(input_dir_or_checkpoint):
+    try:
+        sd = load_checkpoint_from_pytorch_model(input_dir_or_checkpoint)
+        print("Converting checkpoint...")
+        sd = phi_4_hf_to_meta(sd)
+    except FileNotFoundError:
         checkpointer = FullModelHFCheckpointer(
             checkpoint_dir=input_dir_or_checkpoint,
             checkpoint_files=[
@@ -127,11 +126,6 @@ def convert_weights(input_dir_or_checkpoint: str, output_file: str) -> None:
         sd = sd["model"]
         print("Converting checkpoint...")
         sd = phi_4_tune_to_meta(sd)
-    else:
-        print("Loading checkpoint from file...")
-        sd = torch.load(input_dir_or_checkpoint, map_location="cpu", weights_only=True)
-        print("Converting checkpoint...")
-        sd = phi_4_hf_to_meta(sd)
 
     print("Saving checkpoint...")
     torch.save(sd, output_file)

examples/models/qwen3/convert_weights.py

@@ -5,6 +5,7 @@
 from typing import Dict
 
 import torch
+from executorch.examples.models.checkpoint import load_checkpoint_from_pytorch_model
 from safetensors.torch import load_file
 
 from torchtune.models.convert_weights import get_mapped_key
@@ -80,19 +81,34 @@ def load_checkpoint_from_safetensors(input_dir: str) -> Dict:
             tensor = shard_to_weights[shard][weight_name]
             merged_state_dict[weight_name] = tensor
         return merged_state_dict
-    else:
-        # Single checkpoint.
-        state_dict = load_file(os.path.join(input_dir, "model.safetensors"))
-        return state_dict
+
+    # Single checkpoint.
+    model_path = os.path.join(input_dir, "model.safetensors")
+    if os.path.exists(model_path):
+        return load_file(os.path.join(input_dir, "model.safetensors"))
+
+    raise FileNotFoundError(f"Could not find safetensors checkpoint in {input_dir}")
 
 
 def load_checkpoint(input_dir: str) -> Dict:
-    pytorch_path = os.path.join(input_dir, "pytorch_model.bin")
-    if os.path.exists(pytorch_path):
-        print("Loading checkpoint from PyTorch .bin file")
-        return torch.load(pytorch_path, map_location="cpu", weights_only=True)
-    print("Loading checkpoint from safetensors directory")
-    return load_checkpoint_from_safetensors(input_dir)
+    try:
+        print("Loading checkpoint from pytorch_model directory")
+        state_dict = load_checkpoint_from_pytorch_model(input_dir)
+        return state_dict
+    except FileNotFoundError:
+        print(
+            "Could not find pytorch_model checkpoints in directory, trying safetensors"
+        )
+        pass
+
+    try:
+        print("Loading checkpoint from safetensors directory")
+        state_dict = load_checkpoint_from_safetensors(input_dir)
+        return state_dict
+    except FileNotFoundError:
+        pass
+
+    raise FileNotFoundError(f"Could not find checkpoint in {input_dir}")
 
 
 def convert_weights(input_dir: str, output_file: str) -> None: