Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[pull] master from ggerganov:master #25

Closed
wants to merge 8 commits into from
Closed
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
72 changes: 72 additions & 0 deletions convert-hf-to-gguf.py
Original file line number Diff line number Diff line change
Expand Up @@ -189,6 +189,8 @@ def from_model_architecture(model_architecture):
return StableLMModel
if model_architecture == "QWenLMHeadModel":
return QwenModel
if model_architecture == "Qwen2ForCausalLM":
return Model
if model_architecture == "MixtralForCausalLM":
return MixtralModel
if model_architecture == "GPT2LMHeadModel":
Expand All @@ -197,6 +199,8 @@ def from_model_architecture(model_architecture):
return Phi2Model
if model_architecture == "PlamoForCausalLM":
return PlamoModel
if model_architecture == "CodeShellForCausalLM":
return CodeShellModel
return Model

def _is_model_safetensors(self) -> bool:
Expand Down Expand Up @@ -234,6 +238,8 @@ def _get_model_architecture(self) -> gguf.MODEL_ARCH:
return gguf.MODEL_ARCH.STABLELM
if arch == "QWenLMHeadModel":
return gguf.MODEL_ARCH.QWEN
if arch == "Qwen2ForCausalLM":
return gguf.MODEL_ARCH.QWEN2
if arch == "MixtralForCausalLM":
return gguf.MODEL_ARCH.LLAMA
if arch == "GPT2LMHeadModel":
Expand All @@ -242,6 +248,8 @@ def _get_model_architecture(self) -> gguf.MODEL_ARCH:
return gguf.MODEL_ARCH.PHI2
if arch == "PlamoForCausalLM":
return gguf.MODEL_ARCH.PLAMO
if arch == "CodeShellForCausalLM":
return gguf.MODEL_ARCH.CODESHELL

raise NotImplementedError(f'Architecture "{arch}" not supported!')

Expand Down Expand Up @@ -1176,6 +1184,70 @@ def write_tensors(self):
self.gguf_writer.add_tensor(new_name, data)


class CodeShellModel(Model):
def set_gguf_parameters(self):
block_count = self.hparams["n_layer"]

self.gguf_writer.add_name("CodeShell")
self.gguf_writer.add_context_length(self.hparams["n_positions"])
self.gguf_writer.add_embedding_length(self.hparams["n_embd"])
self.gguf_writer.add_feed_forward_length(4 * self.hparams["n_embd"])
self.gguf_writer.add_block_count(block_count)
self.gguf_writer.add_head_count(self.hparams["n_head"])
self.gguf_writer.add_head_count_kv(self.hparams["num_query_groups"])
self.gguf_writer.add_layer_norm_eps(self.hparams["layer_norm_epsilon"])
self.gguf_writer.add_file_type(self.ftype)
self.gguf_writer.add_rope_freq_base(10000.0)
self.gguf_writer.add_rope_scaling_type(gguf.RopeScalingType.LINEAR)
self.gguf_writer.add_rope_scaling_factor(1.0)

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)
tensors = dict(self.get_tensors())
has_lm_head = "lm_head.weight" in tensors.keys() or "output.weight" in tensors.keys()
for name, data_torch in tensors.items():
# we don't need these
if name.endswith((".attn.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"{new_name}, n_dims = {n_dims}, {old_dtype} --> {data.dtype}")

self.gguf_writer.add_tensor(new_name, data)

if not has_lm_head and name == "transformer.wte.weight":
self.gguf_writer.add_tensor("output.weight", data)
print(name, f"=> output.weight, shape = {data.shape}, {old_dtype} --> {data.dtype}")

###### CONVERSION LOGIC ######


Expand Down
32 changes: 32 additions & 0 deletions examples/imatrix/README.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,32 @@
# llama.cpp/examples/imatrix

Compute an importance matrix for a model and given text dataset. Can be used during quantization to enchance the quality of the quantum models.
More information is available here: https://github.com/ggerganov/llama.cpp/pull/4861

## Usage

```
./imatrix -m <some_fp_model> -f <some_training_data> [-o <output_file>] [--verbosity <verbosity_level>]
[-ofreq num_chunks] [-ow <0 or 1>] [other common params]
```

Here `-m` with a model name and `-f` with a file containing training data (such as e.g. `wiki.train.raw`) are mandatory.
The parameters in square brackets are optional and have the following meaning:
* `-o` (or `--output-file`) specifies the name of the file where the computed data will be stored. If missing `imatrix.dat` is used.
* `--verbosity` specifies the verbosity level. If set to `0`, no output other than the perplexity of the processed chunks will be generated. If set to `1`, each time the results are saved a message is written to `stderr`. If `>=2`, a message is output each time data is collected for any tensor. Default verbosity level is `1`.
* `-ofreq` (or `--output-frequency`) specifies how often the so far computed result is saved to disk. Default is 10 (i.e., every 10 chunks)
* `-ow` (or `--output-weight`) specifies if data will be collected for the `output.weight` tensor. My experience is that it is better to not utilize the importance matrix when quantizing `output.weight`, so this is set to `false` by default.

For faster computation, make sure to use GPU offloading via the `-ngl` argument

## Example

```bash
LLAMA_CUBLAS=1 make -j

# generate importance matrix (imatrix.dat)
./imatrix -m ggml-model-f16.gguf -f train-data.txt -ngl 99

# use the imatrix to perform a Q4_K_M quantization
./quantize --imatrix imatrix.dat ggml-model-f16.gguf ./ggml-model-q4_k_m.gguf q4_k_m
```
Loading
Loading