oops, actually fix gguf_writer placement

convert-hf-to-gguf-update.py

@@ -81,7 +81,14 @@ models = [
     {"name": "jina-v2-en",     "tokt": TOKENIZER_TYPE.WPM, "repo": "https://huggingface.co/jinaai/jina-embeddings-v2-base-en", }, # WPM!
     {"name": "jina-v2-es",     "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/jinaai/jina-embeddings-v2-base-es", },
     {"name": "jina-v2-de",     "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/jinaai/jina-embeddings-v2-base-de", },
+<<<<<<< Updated upstream
     {"name": "smaug-bpe",      "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/abacusai/Smaug-Llama-3-70B-Instruct", },
+=======
+    {"name": "bloom",         "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/bigscience/bloom-7b1", },
+    {"name": "gptbigcode",   "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/bigcode/gpt_bigcode-santacoder", },
+    {"name": "phi2",        "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/microsoft/phi-2", },
+    {"name": "codeshell",  "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/WisdomShell/CodeShell-7B-Chat", },
+>>>>>>> Stashed changes
 ]
 
 

convert-hf-to-gguf.py

@@ -56,11 +56,11 @@ class Model:
     part_names: list[str]
     is_safetensors: bool
     hparams: dict[str, Any]
-    gguf_writer: gguf.GGUFManager
     block_count: int
     tensor_map: gguf.TensorNameMap
     tensor_names: set[str] | None
     fname_out: Path
+    gguf_writer: gguf.GGUFManager
 
     # subclasses should define this!
     model_arch: gguf.MODEL_ARCH
@@ -82,8 +82,6 @@ class Model:
             self.part_names = Model.get_model_part_names(self.dir_model, ".bin")
             
         self.hparams = Model.load_hparams(self.dir_model)
-        self.gguf_writer = gguf.GGUFManager(fname_out, gguf.MODEL_ARCH_NAMES[self.model_arch], split_arguments,
-                                            endianess=self.endianess, use_temp_file=self.use_temp_file)
         self.block_count = self.find_hparam(["n_layers", "num_hidden_layers", "n_layer"])
         self.tensor_map = gguf.get_tensor_name_map(self.model_arch, self.block_count)
         self.tensor_names = None
@@ -100,6 +98,8 @@ class Model:
         ftype_lw: str = ftype_up.lower()
         # allow templating the file name with the output ftype, useful with the "auto" ftype
         self.fname_out = fname_out.parent / fname_out.name.format(ftype_lw, outtype=ftype_lw, ftype=ftype_lw, OUTTYPE=ftype_up, FTYPE=ftype_up)
+        self.gguf_writer = gguf.GGUFManager(fname_out, gguf.MODEL_ARCH_NAMES[self.model_arch], split_arguments,
+                                            endianess=self.endianess, use_temp_file=self.use_temp_file)
 
     @classmethod
     def __init_subclass__(cls):

gguf-py/gguf/gguf_manager.py

@@ -305,20 +305,10 @@ class GGUFManager:
             tensor.byteswap(inplace=True)
 
         # TODO reimplement temp file
-        #if self.use_temp_file and self.temp_file is None:
-        #    fp = tempfile.SpooledTemporaryFile(mode="w+b", max_size=256 * 1024 * 1024)
-        #    fp.seek(0)
-        #    self.temp_file = fp
+        # I'm pretty sure it gets handled per shard?
 
         self.tensors.append((name, tensor, raw_dtype))
 
-        #if self.temp_file is None:
-        #    self.tensors.append(tensor)
-        #    return
-
-        #tensor.tofile(self.temp_file)
-        #self.write_padding(self.temp_file, tensor.nbytes)
-
     def close(self) -> None:
         for _, _, writer in self.split_strategy:
             writer.close()