convert : allow conversion of Mistral HF models (ggerganov#6144)

* Allow conversion of Mistral HF models

* Homogenize Llama, Mistral, Mixtral under the same entry.

* Fix tokenizer, permute tensors

* Use sentencepiece tokenizer, or fall back to hfft.

* convert-hf : small fix for mypy

* convert-hf : fix duplicated block_count

* convert-hf : add vocab size to metadata

---------

Co-authored-by: Jared Van Bortel <[email protected]>

convert-hf-to-gguf.py

@@ -23,7 +23,7 @@
  sys.path.insert(1, str(Path(__file__).parent / 'gguf-py'))
 import gguf
 
-from convert import LlamaHfVocab
+from convert import LlamaHfVocab, permute
 
 
 ###### MODEL DEFINITIONS ######
@@ -1052,12 +1052,72 @@ def set_gguf_parameters(self):
  self.gguf_writer.add_layer_norm_eps(self.find_hparam(["layer_norm_eps", "norm_eps"]))
 
 
-@Model.register("MixtralForCausalLM")
-class MixtralModel(Model):
+@Model.register("LlamaForCausalLM", "MistralForCausalLM", "MixtralForCausalLM")
+class LlamaModel(Model):
  model_arch = gguf.MODEL_ARCH.LLAMA
 
  def set_vocab(self):
- self._set_vocab_sentencepiece()
+ try:
+ self. _set_vocab_sentencepiece()
+ except FileNotFoundError:
+ self._set_vocab_llama_hf()
+
+ def set_gguf_parameters(self):
+ super().set_gguf_parameters()
+ hparams = self.hparams
+ self.gguf_writer.add_vocab_size(hparams["vocab_size"])
+ self.gguf_writer.add_rope_dimension_count(hparams["hidden_size"] // hparams["num_attention_heads"])
+
+ # Same as super class, but permuting q_proj, k_proj
+ def write_tensors(self):
+ block_count = self.hparams.get("n_layers", self.hparams.get("num_hidden_layers", self.hparams.get("n_layer")))
+ tensor_map = gguf.get_tensor_name_map(self.model_arch, block_count)
+ n_head = self.hparams.get("num_attention_heads")
+ n_kv_head = self.hparams.get("num_key_value_heads")
+ for name, data_torch in self.get_tensors():
+ # we don't need these
+ if name.endswith((".attention.masked_bias", ".attention.bias", ".attention.rotary_emb.inv_freq")):
+ continue
+
+ old_dtype = data_torch.dtype
+
+ # convert any unsupported data types to float32
+ if data_torch.dtype not in (torch.float16, torch.float32):
+ data_torch = data_torch.to(torch.float32)
+
+ data = data_torch.numpy()
+
+ if name.endswith("q_proj.weight"):
+ data = permute(data, n_head, n_head)
+ if name.endswith("k_proj.weight"):
+ data = permute(data, n_head, n_kv_head)
+
+ data = data.squeeze()
+
+ # map tensor names
+ new_name = tensor_map.get_name(name, try_suffixes=(".weight", ".bias"))
+ if new_name is None:
+ print(f"Can not map tensor {name!r}")
+ sys.exit()
+
+ n_dims = len(data.shape)
+ data_dtype = data.dtype
+
+ # if f32 desired, convert any float16 to float32
+ if self.ftype == 0 and data_dtype == np.float16:
+ data = data.astype(np.float32)
+
+ # TODO: Why cant we use these float16 as-is? There should be not reason to store float16 as float32
+ if self.ftype == 1 and data_dtype == np.float16 and n_dims == 1:
+ data = data.astype(np.float32)
+
+ # if f16 desired, convert any float32 2-dim weight tensors to float16
+ if self.ftype == 1 and data_dtype == np.float32 and name.endswith(".weight") and n_dims == 2:
+ data = data.astype(np.float16)
+
+ print(f"{new_name}, n_dims = {n_dims}, {old_dtype} --> {data.dtype}")
+
+ self.gguf_writer.add_tensor(new_name, data)
 
 
 @Model.register("GrokForCausalLM")