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.

Sign up for GitHub

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

Jump to bottom

Is there a possible memory leak in llama_cpp.llama_decode()? #924

Closed
3 of 4 tasks
littlebai3618 opened this issue Nov 17, 2023 · 3 comments
Closed
3 of 4 tasks

Is there a possible memory leak in llama_cpp.llama_decode()? #924

littlebai3618 opened this issue Nov 17, 2023 · 3 comments

Comments

@littlebai3618
Copy link

Prerequisites

Please answer the following questions for yourself before submitting an issue.

  • I am running the latest code. Development is very rapid so there are no tagged versions as of now.
  • I carefully followed the README.md.
  • I searched using keywords relevant to my issue to make sure that I am creating a new issue that is not already open (or closed).
  • I reviewed the Discussions, and have a new bug or useful enhancement to share.

Expected Behavior

Is there a possible memory leak in llama_cpp.llama_decode()? If this is a normal behavior, please let me know.

Background: I'm using the low-level API provided by llama_cpp.py to implement a Python version of continuous batch processing based on parallel.cpp in llama.cpp. During runtime, the program's memory and GPU memory usage keep increasing slowly, and eventually, the program crashes.

I used memory_profiler to observe the memory usage of each line in the self.eval method of the higher-level llama API. I found that the memory usage significantly increases in the following line and is not released in subsequent runs

14415.5 MiB    125.1 MiB           1               self._ctx.decode(self._batch)

My understanding is that any operations performed by decode should operate within pre-allocated memory instead of occupying new memory.

Please note that not every call to this line increases memory usage, but once it increases, the memory is not released.

Current Behavior

As mentioned above, the memory appears to be increasing abnormally.

Environment and Context

  • Physical (or virtual) hardware you are using, e.g. for Linux:

$ lscpu

Architecture:                    x86_64
CPU op-mode(s):                  32-bit, 64-bit
Byte Order:                      Little Endian
Address sizes:                   46 bits physical, 48 bits virtual
CPU(s):                          64
On-line CPU(s) list:             0-63
Thread(s) per core:              2
Core(s) per socket:              16
Socket(s):                       2
NUMA node(s):                    2
Vendor ID:                       GenuineIntel
CPU family:                      6
Model:                           85
Model name:                      Intel(R) Xeon(R) Gold 5218 CPU @ 2.30GHz
Stepping:                        7
CPU MHz:                         1027.243
CPU max MHz:                     3900.0000
CPU min MHz:                     1000.0000
BogoMIPS:                        4600.00
Virtualization:                  VT-x
L1d cache:                       1 MiB
L1i cache:                       1 MiB
L2 cache:                        32 MiB
L3 cache:                        44 MiB

nvidia-smi

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 525.85.05    Driver Version: 525.85.05    CUDA Version: 12.0     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  Quadro RTX 8000     Off  | 00000000:3B:00.0 Off |                  N/A |
| 33%   29C    P8    23W / 260W |      0MiB / 49152MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+
  • Operating System, e.g. for Linux:

$ uname -a

Linux aistudio-31572-prod-0 4.19.96 #1 SMP Tue Mar 10 10:34:01 CST 2020 x86_64 x86_64 x86_64 GNU/Linux
  • SDK version, e.g. for Linux:
pip list
llama_cpp_python         0.2.18
memory-profiler          0.61.0

python                   3.11.5
make  GNU Make 4.2.1
g++   g++ (Ubuntu 9.4.0-1ubuntu1~20.04.1) 9.4.0
  • model
    It appears that you are using the "CodeLlama-7B-HF" model from the repository you mentioned (https://huggingface.co/codellama/CodeLlama-7b-hf). You mentioned that you performed the conversion using the "convert.py" script included in the repository.

Steps to Reproduce

  1. CMAKE_ARGS="-DLLAMA_CUBLAS=on -DLLAMA_CUDA_DMMV_X=128 -DLLAMA_CUDA_MMV_Y=32" FORCE_CMAKE=1 pip install llama-cpp-python==0.2.18 --no-cache-dir --force-reinstall --upgrade
  2. add memcheck code to llama.py
from memory_profiler import profile
class Llama():
    ......
    @profile
    def eval(self, tokens: Sequence[int]):
        """Evaluate a list of tokens.

        Args:
            tokens: The list of tokens to evaluate.
        """
        assert self._ctx.ctx is not None
        assert self._batch.batch is not None
        self._ctx.kv_cache_seq_rm(-1, self.n_tokens, -1)
        for i in range(0, len(tokens), self.n_batch):
            batch = tokens[i: min(len(tokens), i + self.n_batch)]
            n_past = self.n_tokens
            n_tokens = len(batch)
            self._batch.set_batch(
                batch=batch, n_past=n_past, logits_all=self.context_params.logits_all
            )
            self._ctx.decode(self._batch)
            # Save tokens
            self.input_ids[n_past: n_past + n_tokens] = batch
            # Save logits
            rows = n_tokens
            cols = self._n_vocab
            offset = (
                0 if self.context_params.logits_all else n_tokens - 1
            )  # NOTE: Only save the last token logits if logits_all is False
            self.scores[n_past + offset: n_past + n_tokens, :].reshape(
                -1
            )[:] = self._ctx.get_logits()[offset * cols: rows * cols]
            # Update n_tokens
            self.n_tokens += n_tokens
  1. Start the provided API server.:
python3 -m llama_cpp.server --model ./CodeLlama-7B/ggml-model-q8_0.gguf
  1. Observe the memory usage.
  2. Call the API endpoint /v1/completions to perform a completion operation.

Failure Logs

Please ignore the line numbers. I have inherited the Llama class to add memory monitoring code.

ggml_init_cublas: GGML_CUDA_FORCE_MMQ:   no
ggml_init_cublas: CUDA_USE_TENSOR_CORES: yes
ggml_init_cublas: found 1 CUDA devices:
  Device 0: Quadro RTX 8000, compute capability 7.5
llama_model_loader: loaded meta data with 18 key-value pairs and 291 tensors from ./CodeLlama-7B/ggml-model-q8_0.gguf (version GGUF V3 (latest))
llama_model_loader: - tensor    0:                token_embd.weight q8_0     [  4096, 32017,     1,     1 ]
llama_model_loader: - tensor    1:              blk.0.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor    2:              blk.0.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor    3:              blk.0.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor    4:         blk.0.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor    5:            blk.0.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor    6:              blk.0.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor    7:            blk.0.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor    8:           blk.0.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor    9:            blk.0.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor   10:              blk.1.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   11:              blk.1.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   12:              blk.1.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   13:         blk.1.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   14:            blk.1.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor   15:              blk.1.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor   16:            blk.1.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor   17:           blk.1.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor   18:            blk.1.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor   19:              blk.2.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   20:              blk.2.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   21:              blk.2.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   22:         blk.2.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   23:            blk.2.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor   24:              blk.2.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor   25:            blk.2.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor   26:           blk.2.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor   27:            blk.2.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor   28:              blk.3.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   29:              blk.3.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   30:              blk.3.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   31:         blk.3.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   32:            blk.3.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor   33:              blk.3.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor   34:            blk.3.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor   35:           blk.3.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor   36:            blk.3.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor   37:              blk.4.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   38:              blk.4.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   39:              blk.4.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   40:         blk.4.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   41:            blk.4.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor   42:              blk.4.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor   43:            blk.4.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor   44:           blk.4.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor   45:            blk.4.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor   46:              blk.5.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   47:              blk.5.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   48:              blk.5.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   49:         blk.5.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   50:            blk.5.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor   51:              blk.5.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor   52:            blk.5.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor   53:           blk.5.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor   54:            blk.5.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor   55:              blk.6.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   56:              blk.6.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   57:              blk.6.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   58:         blk.6.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   59:            blk.6.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor   60:              blk.6.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor   61:            blk.6.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor   62:           blk.6.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor   63:            blk.6.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor   64:              blk.7.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   65:              blk.7.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   66:              blk.7.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   67:         blk.7.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   68:            blk.7.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor   69:              blk.7.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor   70:            blk.7.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor   71:           blk.7.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor   72:            blk.7.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor   73:              blk.8.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   74:              blk.8.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   75:              blk.8.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   76:         blk.8.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   77:            blk.8.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor   78:              blk.8.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor   79:            blk.8.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor   80:           blk.8.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor   81:            blk.8.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor   82:              blk.9.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   83:              blk.9.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   84:              blk.9.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   85:         blk.9.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   86:            blk.9.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor   87:              blk.9.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor   88:            blk.9.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor   89:           blk.9.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor   90:            blk.9.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor   91:             blk.10.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   92:             blk.10.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   93:             blk.10.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   94:        blk.10.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor   95:           blk.10.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor   96:             blk.10.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor   97:           blk.10.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor   98:          blk.10.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor   99:           blk.10.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  100:             blk.11.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  101:             blk.11.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  102:             blk.11.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  103:        blk.11.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  104:           blk.11.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  105:             blk.11.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  106:           blk.11.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor  107:          blk.11.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  108:           blk.11.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  109:             blk.12.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  110:             blk.12.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  111:             blk.12.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  112:        blk.12.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  113:           blk.12.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  114:             blk.12.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  115:           blk.12.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor  116:          blk.12.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  117:           blk.12.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  118:             blk.13.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  119:             blk.13.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  120:             blk.13.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  121:        blk.13.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  122:           blk.13.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  123:             blk.13.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  124:           blk.13.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor  125:          blk.13.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  126:           blk.13.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  127:             blk.14.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  128:             blk.14.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  129:             blk.14.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  130:        blk.14.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  131:           blk.14.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  132:             blk.14.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  133:           blk.14.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor  134:          blk.14.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  135:           blk.14.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  136:             blk.15.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  137:             blk.15.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  138:             blk.15.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  139:        blk.15.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  140:           blk.15.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  141:             blk.15.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  142:           blk.15.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor  143:          blk.15.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  144:           blk.15.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  145:             blk.16.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  146:             blk.16.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  147:             blk.16.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  148:        blk.16.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  149:           blk.16.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  150:             blk.16.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  151:           blk.16.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor  152:          blk.16.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  153:           blk.16.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  154:             blk.17.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  155:             blk.17.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  156:             blk.17.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  157:        blk.17.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  158:           blk.17.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  159:             blk.17.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  160:           blk.17.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor  161:          blk.17.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  162:           blk.17.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  163:             blk.18.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  164:             blk.18.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  165:             blk.18.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  166:        blk.18.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  167:           blk.18.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  168:             blk.18.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  169:           blk.18.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor  170:          blk.18.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  171:           blk.18.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  172:             blk.19.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  173:             blk.19.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  174:             blk.19.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  175:        blk.19.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  176:           blk.19.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  177:             blk.19.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  178:           blk.19.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor  179:          blk.19.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  180:           blk.19.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  181:             blk.20.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  182:             blk.20.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  183:             blk.20.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  184:        blk.20.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  185:           blk.20.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  186:             blk.20.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  187:           blk.20.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor  188:          blk.20.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  189:           blk.20.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  190:             blk.21.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  191:             blk.21.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  192:             blk.21.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  193:        blk.21.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  194:           blk.21.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  195:             blk.21.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  196:           blk.21.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor  197:          blk.21.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  198:           blk.21.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  199:             blk.22.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  200:             blk.22.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  201:             blk.22.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  202:        blk.22.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  203:           blk.22.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  204:             blk.22.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  205:           blk.22.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor  206:          blk.22.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  207:           blk.22.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  208:             blk.23.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  209:             blk.23.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  210:             blk.23.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  211:        blk.23.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  212:           blk.23.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  213:             blk.23.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  214:           blk.23.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor  215:          blk.23.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  216:           blk.23.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  217:             blk.24.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  218:             blk.24.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  219:             blk.24.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  220:        blk.24.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  221:           blk.24.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  222:             blk.24.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  223:           blk.24.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor  224:          blk.24.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  225:           blk.24.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  226:             blk.25.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  227:             blk.25.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  228:             blk.25.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  229:        blk.25.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  230:           blk.25.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  231:             blk.25.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  232:           blk.25.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor  233:          blk.25.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  234:           blk.25.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  235:             blk.26.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  236:             blk.26.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  237:             blk.26.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  238:        blk.26.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  239:           blk.26.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  240:             blk.26.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  241:           blk.26.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor  242:          blk.26.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  243:           blk.26.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  244:             blk.27.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  245:             blk.27.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  246:             blk.27.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  247:        blk.27.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  248:           blk.27.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  249:             blk.27.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  250:           blk.27.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor  251:          blk.27.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  252:           blk.27.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  253:             blk.28.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  254:             blk.28.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  255:             blk.28.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  256:        blk.28.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  257:           blk.28.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  258:             blk.28.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  259:           blk.28.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor  260:          blk.28.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  261:           blk.28.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  262:             blk.29.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  263:             blk.29.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  264:             blk.29.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  265:        blk.29.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  266:           blk.29.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  267:             blk.29.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  268:           blk.29.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor  269:          blk.29.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  270:           blk.29.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  271:             blk.30.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  272:             blk.30.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  273:             blk.30.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  274:        blk.30.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  275:           blk.30.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  276:             blk.30.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  277:           blk.30.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor  278:          blk.30.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  279:           blk.30.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  280:             blk.31.attn_q.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  281:             blk.31.attn_k.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  282:             blk.31.attn_v.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  283:        blk.31.attn_output.weight q8_0     [  4096,  4096,     1,     1 ]
llama_model_loader: - tensor  284:           blk.31.ffn_gate.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  285:             blk.31.ffn_up.weight q8_0     [  4096, 11008,     1,     1 ]
llama_model_loader: - tensor  286:           blk.31.ffn_down.weight q8_0     [ 11008,  4096,     1,     1 ]
llama_model_loader: - tensor  287:          blk.31.attn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  288:           blk.31.ffn_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  289:               output_norm.weight f32      [  4096,     1,     1,     1 ]
llama_model_loader: - tensor  290:                    output.weight q8_0     [  4096, 32017,     1,     1 ]
llama_model_loader: - kv   0:                       general.architecture str     
llama_model_loader: - kv   1:                               general.name str     
llama_model_loader: - kv   2:                       llama.context_length u32     
llama_model_loader: - kv   3:                     llama.embedding_length u32     
llama_model_loader: - kv   4:                          llama.block_count u32     
llama_model_loader: - kv   5:                  llama.feed_forward_length u32     
llama_model_loader: - kv   6:                 llama.rope.dimension_count u32     
llama_model_loader: - kv   7:                 llama.attention.head_count u32     
llama_model_loader: - kv   8:              llama.attention.head_count_kv u32     
llama_model_loader: - kv   9:     llama.attention.layer_norm_rms_epsilon f32     
llama_model_loader: - kv  10:                       llama.rope.freq_base f32     
llama_model_loader: - kv  11:                          general.file_type u32     
llama_model_loader: - kv  12:                       tokenizer.ggml.model str     
llama_model_loader: - kv  13:                      tokenizer.ggml.tokens arr     
llama_model_loader: - kv  14:                      tokenizer.ggml.scores arr     
llama_model_loader: - kv  15:                  tokenizer.ggml.token_type arr     
llama_model_loader: - kv  16:                tokenizer.ggml.bos_token_id u32     
llama_model_loader: - kv  17:                tokenizer.ggml.eos_token_id u32     
llama_model_loader: - type  f32:   65 tensors
llama_model_loader: - type q8_0:  226 tensors
llm_load_vocab: mismatch in special tokens definition ( 265/32017 vs 260/32017 ).
llm_load_print_meta: format           = GGUF V3 (latest)
llm_load_print_meta: arch             = llama
llm_load_print_meta: vocab type       = SPM
llm_load_print_meta: n_vocab          = 32017
llm_load_print_meta: n_merges         = 0
llm_load_print_meta: n_ctx_train      = 16384
llm_load_print_meta: n_embd           = 4096
llm_load_print_meta: n_head           = 32
llm_load_print_meta: n_head_kv        = 32
llm_load_print_meta: n_layer          = 32
llm_load_print_meta: n_rot            = 128
llm_load_print_meta: n_gqa            = 1
llm_load_print_meta: f_norm_eps       = 0.0e+00
llm_load_print_meta: f_norm_rms_eps   = 1.0e-05
llm_load_print_meta: f_clamp_kqv      = 0.0e+00
llm_load_print_meta: f_max_alibi_bias = 0.0e+00
llm_load_print_meta: n_ff             = 11008
llm_load_print_meta: rope scaling     = linear
llm_load_print_meta: freq_base_train  = 1000000.0
llm_load_print_meta: freq_scale_train = 1
llm_load_print_meta: n_yarn_orig_ctx  = 16384
llm_load_print_meta: rope_finetuned   = unknown
llm_load_print_meta: model type       = 7B
llm_load_print_meta: model ftype      = mostly Q8_0
llm_load_print_meta: model params     = 6.74 B
llm_load_print_meta: model size       = 6.67 GiB (8.50 BPW) 
llm_load_print_meta: general.name   = models
llm_load_print_meta: BOS token = 1 '<s>'
llm_load_print_meta: EOS token = 2 '</s>'
llm_load_print_meta: UNK token = 0 '<unk>'
llm_load_print_meta: LF token  = 13 '<0x0A>'
llm_load_tensors: ggml ctx size =    0.11 MB
llm_load_tensors: using CUDA for GPU acceleration
llm_load_tensors: mem required  =  132.99 MB
llm_load_tensors: offloading 32 repeating layers to GPU
llm_load_tensors: offloading non-repeating layers to GPU
llm_load_tensors: offloaded 35/35 layers to GPU
llm_load_tensors: VRAM used: 6695.90 MB
...................................................................................................
llama_new_context_with_model: n_ctx      = 4096
llama_new_context_with_model: freq_base  = 1000000.0
llama_new_context_with_model: freq_scale = 1
llama_kv_cache_init: offloading v cache to GPU
llama_kv_cache_init: offloading k cache to GPU
llama_kv_cache_init: VRAM kv self = 2048.00 MB
llama_new_context_with_model: kv self size  = 2048.00 MB
llama_build_graph: non-view tensors processed: 740/740
llama_new_context_with_model: compute buffer total size = 2305.59 MB
llama_new_context_with_model: VRAM scratch buffer: 2304.03 MB
llama_new_context_with_model: total VRAM used: 11047.93 MB (model: 6695.90 MB, context: 4352.03 MB)
AVX = 1 | AVX2 = 1 | AVX512 = 1 | AVX512_VBMI = 0 | AVX512_VNNI = 1 | FMA = 1 | NEON = 0 | ARM_FMA = 0 | F16C = 1 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 1 | SSE3 = 1 | SSSE3 = 1 | VSX = 0 |

Here are some of the detection results:

Line #    Mem usage    Increment  Occurrences   Line Contents
=============================================================
   115  14290.4 MiB  14290.4 MiB           1       @profile
   116                                             def eval(self, tokens: Sequence[int]):
   117                                                 """Evaluate a list of tokens.
   118                                         
   119                                                 Args:
   120                                                     tokens: The list of tokens to evaluate.
   121                                                 """
   122  14290.4 MiB      0.0 MiB           1           assert self._ctx.ctx is not None
   123  14290.4 MiB      0.0 MiB           1           assert self._batch.batch is not None
   124  14290.4 MiB      0.0 MiB           1           self._ctx.kv_cache_seq_rm(-1, self.n_tokens, -1)
   125  14417.7 MiB      0.0 MiB           2           for i in range(0, len(tokens), self.n_batch):
   126  14290.4 MiB      0.0 MiB           1               batch = tokens[i: min(len(tokens), i + self.n_batch)]
   127  14290.4 MiB      0.0 MiB           1               n_past = self.n_tokens
   128  14290.4 MiB      0.0 MiB           1               n_tokens = len(batch)
   129  14290.4 MiB      0.0 MiB           2               self._batch.set_batch(
   130  14290.4 MiB      0.0 MiB           1                   batch=batch, n_past=n_past, logits_all=self.context_params.logits_all
   131                                                     )
   132  14415.5 MiB    125.1 MiB           1               self._ctx.decode(self._batch)
   133                                                     # Save tokens
   134  14415.5 MiB      0.0 MiB           1               self.input_ids[n_past: n_past + n_tokens] = batch
   135                                                     # Save logits
   136  14415.5 MiB      0.0 MiB           1               rows = n_tokens
   137  14415.5 MiB      0.0 MiB           1               cols = self._n_vocab
   138  14415.5 MiB      0.0 MiB           1               offset = (
   139  14415.5 MiB      0.0 MiB           1                   0 if self.context_params.logits_all else n_tokens - 1
   140                                                     )  # NOTE: Only save the last token logits if logits_all is False
   141  14417.7 MiB      2.0 MiB           3               self.scores[n_past + offset: n_past + n_tokens, :].reshape(
   142  14415.7 MiB      0.0 MiB           1                   -1
   143  14415.7 MiB      0.2 MiB           2               )[:] = self._ctx.get_logits()[offset * cols: rows * cols]
   144                                                     # Update n_tokens
   145  14417.7 MiB      0.0 MiB           1               self.n_tokens += n_tokens



Line #    Mem usage    Increment  Occurrences   Line Contents
=============================================================
   115  14417.8 MiB  14417.8 MiB           1       @profile
   116                                             def eval(self, tokens: Sequence[int]):
   117                                                 """Evaluate a list of tokens.
   118                                         
   119                                                 Args:
   120                                                     tokens: The list of tokens to evaluate.
   121                                                 """
   122  14417.8 MiB      0.0 MiB           1           assert self._ctx.ctx is not None
   123  14417.8 MiB      0.0 MiB           1           assert self._batch.batch is not None
   124  14417.8 MiB      0.0 MiB           1           self._ctx.kv_cache_seq_rm(-1, self.n_tokens, -1)
   125  14418.8 MiB      0.0 MiB           2           for i in range(0, len(tokens), self.n_batch):
   126  14417.8 MiB      0.0 MiB           1               batch = tokens[i: min(len(tokens), i + self.n_batch)]
   127  14417.8 MiB      0.0 MiB           1               n_past = self.n_tokens
   128  14417.8 MiB      0.0 MiB           1               n_tokens = len(batch)
   129  14417.8 MiB      0.0 MiB           2               self._batch.set_batch(
   130  14417.8 MiB      0.0 MiB           1                   batch=batch, n_past=n_past, logits_all=self.context_params.logits_all
   131                                                     )
   132  14418.8 MiB      1.0 MiB           1               self._ctx.decode(self._batch)
   133                                                     # Save tokens
   134  14418.8 MiB      0.0 MiB           1               self.input_ids[n_past: n_past + n_tokens] = batch
   135                                                     # Save logits
   136  14418.8 MiB      0.0 MiB           1               rows = n_tokens
   137  14418.8 MiB      0.0 MiB           1               cols = self._n_vocab
   138  14418.8 MiB      0.0 MiB           1               offset = (
   139  14418.8 MiB      0.0 MiB           1                   0 if self.context_params.logits_all else n_tokens - 1
   140                                                     )  # NOTE: Only save the last token logits if logits_all is False
   141  14418.8 MiB      0.0 MiB           3               self.scores[n_past + offset: n_past + n_tokens, :].reshape(
   142  14418.8 MiB      0.0 MiB           1                   -1
   143  14418.8 MiB      0.0 MiB           2               )[:] = self._ctx.get_logits()[offset * cols: rows * cols]
   144                                                     # Update n_tokens
   145  14418.8 MiB      0.0 MiB           1               self.n_tokens += n_tokens



Line #    Mem usage    Increment  Occurrences   Line Contents
=============================================================
   115  14418.8 MiB  14418.8 MiB           1       @profile
   116                                             def eval(self, tokens: Sequence[int]):
   117                                                 """Evaluate a list of tokens.
   118                                         
   119                                                 Args:
   120                                                     tokens: The list of tokens to evaluate.
   121                                                 """
   122  14418.8 MiB      0.0 MiB           1           assert self._ctx.ctx is not None
   123  14418.8 MiB      0.0 MiB           1           assert self._batch.batch is not None
   124  14418.8 MiB      0.0 MiB           1           self._ctx.kv_cache_seq_rm(-1, self.n_tokens, -1)
   125  14418.8 MiB      0.0 MiB           2           for i in range(0, len(tokens), self.n_batch):
   126  14418.8 MiB      0.0 MiB           1               batch = tokens[i: min(len(tokens), i + self.n_batch)]
   127  14418.8 MiB      0.0 MiB           1               n_past = self.n_tokens
   128  14418.8 MiB      0.0 MiB           1               n_tokens = len(batch)
   129  14418.8 MiB      0.0 MiB           2               self._batch.set_batch(
   130  14418.8 MiB      0.0 MiB           1                   batch=batch, n_past=n_past, logits_all=self.context_params.logits_all
   131                                                     )
   132  14418.8 MiB      0.0 MiB           1               self._ctx.decode(self._batch)
   133                                                     # Save tokens
   134  14418.8 MiB      0.0 MiB           1               self.input_ids[n_past: n_past + n_tokens] = batch
   135                                                     # Save logits
   136  14418.8 MiB      0.0 MiB           1               rows = n_tokens
   137  14418.8 MiB      0.0 MiB           1               cols = self._n_vocab
   138  14418.8 MiB      0.0 MiB           1               offset = (
   139  14418.8 MiB      0.0 MiB           1                   0 if self.context_params.logits_all else n_tokens - 1
   140                                                     )  # NOTE: Only save the last token logits if logits_all is False
   141  14418.8 MiB      0.0 MiB           3               self.scores[n_past + offset: n_past + n_tokens, :].reshape(
   142  14418.8 MiB      0.0 MiB           1                   -1
   143  14418.8 MiB      0.0 MiB           2               )[:] = self._ctx.get_logits()[offset * cols: rows * cols]
   144                                                     # Update n_tokens
   145  14418.8 MiB      0.0 MiB           1               self.n_tokens += n_tokens



Line #    Mem usage    Increment  Occurrences   Line Contents
=============================================================
   115  14418.8 MiB  14418.8 MiB           1       @profile
   116                                             def eval(self, tokens: Sequence[int]):
   117                                                 """Evaluate a list of tokens.
   118                                         
   119                                                 Args:
   120                                                     tokens: The list of tokens to evaluate.
   121                                                 """
   122  14418.8 MiB      0.0 MiB           1           assert self._ctx.ctx is not None
   123  14418.8 MiB      0.0 MiB           1           assert self._batch.batch is not None
   124  14418.8 MiB      0.0 MiB           1           self._ctx.kv_cache_seq_rm(-1, self.n_tokens, -1)
   125  14418.8 MiB      0.0 MiB           2           for i in range(0, len(tokens), self.n_batch):
   126  14418.8 MiB      0.0 MiB           1               batch = tokens[i: min(len(tokens), i + self.n_batch)]
   127  14418.8 MiB      0.0 MiB           1               n_past = self.n_tokens
   128  14418.8 MiB      0.0 MiB           1               n_tokens = len(batch)
   129  14418.8 MiB      0.0 MiB           2               self._batch.set_batch(
   130  14418.8 MiB      0.0 MiB           1                   batch=batch, n_past=n_past, logits_all=self.context_params.logits_all
   131                                                     )
   132  14418.8 MiB      0.0 MiB           1               self._ctx.decode(self._batch)
   133                                                     # Save tokens
   134  14418.8 MiB      0.0 MiB           1               self.input_ids[n_past: n_past + n_tokens] = batch
   135                                                     # Save logits
   136  14418.8 MiB      0.0 MiB           1               rows = n_tokens
   137  14418.8 MiB      0.0 MiB           1               cols = self._n_vocab
   138  14418.8 MiB      0.0 MiB           1               offset = (
   139  14418.8 MiB      0.0 MiB           1                   0 if self.context_params.logits_all else n_tokens - 1
   140                                                     )  # NOTE: Only save the last token logits if logits_all is False
   141  14418.8 MiB      0.0 MiB           3               self.scores[n_past + offset: n_past + n_tokens, :].reshape(
   142  14418.8 MiB      0.0 MiB           1                   -1
   143  14418.8 MiB      0.0 MiB           2               )[:] = self._ctx.get_logits()[offset * cols: rows * cols]
   144                                                     # Update n_tokens
   145  14418.8 MiB      0.0 MiB           1               self.n_tokens += n_tokens
@littlebai3618 littlebai3618 mentioned this issue Nov 17, 2023
Closed
4 tasks
@abetlen
Copy link
Owner

abetlen commented Nov 21, 2023 •
edited
Loading

Thanks @littlebai3618, I'm actually working on the same thing for #771 so will look into this. My guess is that llama.cpp is simply not shrinking the kv cache and repeated calls lead to excess fragmentation and cache bloat, do you always OOM or does it hit ever hit a ceiling?

@littlebai3618 littlebai3618 mentioned this issue Nov 21, 2023
Closed
@littlebai3618
Copy link
Author

Here are the latest developments:

  1. The maintainer of llama.cpp has stated that the kv cache is not the issue.
  2. My latest test results show that there is a problem with memory usage and process termination when parallel processing long prompts (1500-2000).
  3. I noticed that llama.cpp is not thread-safe. I have tried executing all the logic within the same thread, but the memory usage issue still persists.

I'm not aware of the progress on continuous batch processing, so I've submitted a draft of my simplified version. This draft code can handle 1000 sentences in parallel with 10 instances. However, it crashes when the prompt is too long. Please review it for any potential bugs.

If this issue can be resolved, I'd be happy to assist you with implementing support for continuous batch processing.

@littlebai3618
Copy link
Author

After investigation, it was discovered that I misunderstood the batch parameter. The issue of fragmented kv cache has been resolved, and the latest llama.cpp has fixed this problem.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
2 participants
@abetlen @littlebai3618