diff options
Diffstat (limited to 'convert-hf-to-gguf.py')
| -rwxr-xr-x | convert-hf-to-gguf.py | 81 |
1 files changed, 81 insertions, 0 deletions
diff --git a/convert-hf-to-gguf.py b/convert-hf-to-gguf.py index 7a0a8c3d..6ab7f486 100755 --- a/convert-hf-to-gguf.py +++ b/convert-hf-to-gguf.py @@ -201,6 +201,8 @@ class Model: return PlamoModel if model_architecture == "CodeShellForCausalLM": return CodeShellModel + if model_architecture == "OrionForCausalLM": + return OrionModel return Model def _is_model_safetensors(self) -> bool: @@ -250,6 +252,8 @@ class Model: return gguf.MODEL_ARCH.PLAMO if arch == "CodeShellForCausalLM": return gguf.MODEL_ARCH.CODESHELL + if arch == "OrionForCausalLM": + return gguf.MODEL_ARCH.ORION raise NotImplementedError(f'Architecture "{arch}" not supported!') @@ -572,6 +576,83 @@ class MPTModel(Model): self.gguf_writer.add_tensor("output.weight", data) +class OrionModel(Model): + def set_vocab(self): + self._set_vocab_sentencepiece() + + def set_gguf_parameters(self): + block_count = self.hparams["num_hidden_layers"] + head_count = self.hparams["num_attention_heads"] + head_count_kv = self.hparams.get("num_key_value_heads", head_count) + hf_repo = self.hparams.get("_name_or_path", "") + + ctx_length = 0 + if "max_sequence_length" in self.hparams: + ctx_length = self.hparams["max_sequence_length"] + elif "max_position_embeddings" in self.hparams: + ctx_length = self.hparams["max_position_embeddings"] + elif "model_max_length" in self.hparams: + ctx_length = self.hparams["model_max_length"] + else: + print("gguf: can not find ctx length parameter.") + sys.exit() + + self.gguf_writer.add_file_type(self.ftype) + self.gguf_writer.add_name(self.dir_model.name) + self.gguf_writer.add_source_hf_repo(hf_repo) + self.gguf_writer.add_tensor_data_layout("Meta AI original pth") + self.gguf_writer.add_context_length(ctx_length) + self.gguf_writer.add_embedding_length(self.hparams["hidden_size"]) + self.gguf_writer.add_block_count(block_count) + self.gguf_writer.add_feed_forward_length(self.hparams["intermediate_size"]) + self.gguf_writer.add_head_count(head_count) + self.gguf_writer.add_head_count_kv(head_count_kv) + self.gguf_writer.add_layer_norm_eps(self.hparams["rms_norm_eps"]) + + def write_tensors(self): + # Collect tensors from generator object + model_kv = dict(self.get_tensors()) + block_count = self.hparams["num_hidden_layers"] + tensor_map = gguf.get_tensor_name_map(self.model_arch, block_count) + + for name, data_torch in model_kv.items(): + # we don't need these + if name.endswith(".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.squeeze().numpy() + + # 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"{name} -> {new_name}, n_dims = {n_dims}, {old_dtype} --> {data.dtype}") + self.gguf_writer.add_tensor(new_name, data) + + class BaichuanModel(Model): def set_vocab(self): self._set_vocab_sentencepiece() |