Add LLaMa as an example. (#722)

* Add LLaMa as an example.

* Bugfix.

* Switch to using safetensors.

* Properly switch to half-precision.

* Get the text sampling to work.

.gitignore

@@ -16,3 +16,4 @@ __pycache__
 *.safetensors
 *.so
 *.dylib
+llama-tokenizer.json

Cargo.toml

@@ -30,6 +30,8 @@ safetensors = "0.3.0"
 cpython = { version = "0.7.1", optional = true }
 regex = { version = "1.6.0", optional = true }
 image = { version = "0.24.5", optional = true }
+clap = { version = "4.2.4", features = ["derive"], optional = true }
+serde_json = { version = "1.0.96", optional = true }
 
 [dev-dependencies]
 anyhow = "1"
@@ -54,3 +56,7 @@ required-features = ["rl-python"]
 [[example]]
 name = "stable-diffusion"
 required-features = ["regex"]
+
+[[example]]
+name = "llama"
+required-features = ["regex", "clap", "serde_json"]

examples/llama/convert_checkpoint.py

@@ -0,0 +1,63 @@
+# Adapted from https://github.com/Lightning-AI/lit-llama/blob/main/scripts/convert_checkpoint.py
+import sys
+import torch
+from typing import Dict
+from pathlib import Path
+from safetensors.torch import save_file
+
+def convert_state_dict(state_dict: Dict[str, torch.Tensor], dtype: torch.dtype = torch.float16) -> Dict[str, torch.Tensor]:
+    print("start conv")
+
+    def get_and_remove(key):
+        v = state_dict[key].to(dtype)
+        del state_dict[key]
+        return v
+
+    converted = {}
+    converted["transformer.wte.weight"] = get_and_remove("tok_embeddings.weight")
+    converted["lm_head.weight"] = get_and_remove("output.weight")
+    converted["transformer.ln_f.scale"] = get_and_remove("norm.weight")
+
+    for layer_idx in sorted(set([k.split(".")[1] for k in state_dict if k.startswith("layers")])):
+        print(layer_idx)
+
+        # attention
+        # the wq, wk, wv from the FB model are stacked in our model as c_attn
+        converted[f"transformer.h.{layer_idx}.attn.c_attn.weight"] = torch.cat(
+            (
+                get_and_remove(f"layers.{layer_idx}.attention.wq.weight"),
+                get_and_remove(f"layers.{layer_idx}.attention.wk.weight"),
+                get_and_remove(f"layers.{layer_idx}.attention.wv.weight"),
+                )
+        )
+        converted[f"transformer.h.{layer_idx}.attn.c_proj.weight"] = get_and_remove(
+            f"layers.{layer_idx}.attention.wo.weight"
+            )
+        # mlp
+        converted[f"transformer.h.{layer_idx}.mlp.c_fc1.weight"] = get_and_remove(
+            f"layers.{layer_idx}.feed_forward.w1.weight"
+            )
+        converted[f"transformer.h.{layer_idx}.mlp.c_proj.weight"] = get_and_remove(
+            f"layers.{layer_idx}.feed_forward.w2.weight"
+            )
+        converted[f"transformer.h.{layer_idx}.mlp.c_fc2.weight"] = get_and_remove(
+            f"layers.{layer_idx}.feed_forward.w3.weight"
+            )
+        # rms norm
+        converted[f"transformer.h.{layer_idx}.rms_1.scale"] = get_and_remove(f"layers.{layer_idx}.attention_norm.weight")
+        converted[f"transformer.h.{layer_idx}.rms_2.scale"] = get_and_remove(f"layers.{layer_idx}.ffn_norm.weight")
+    return converted
+
+def convert_weights(llama_ckpt, *, output_st: Path = Path("llama.safetensors"), dtype: str = "float16") -> None:
+    dt = getattr(torch, dtype, None)
+    if not isinstance(dt, torch.dtype):
+        raise ValueError(f"{dtype} is not a valid dtype.")
+    checkpoint = torch.load(llama_ckpt, map_location="cpu")
+    converted = convert_state_dict(checkpoint, dtype=dt)
+    del checkpoint
+    save_file(converted, output_st)
+
+if __name__ == "__main__":
+    if len(sys.argv) != 2:
+        raise ValueError(f"usage: convert_checkpoint.py ..../LLaMA/7B/consolidated.00.pth")
+    convert_weights(sys.argv[1])