Skip to content

convert-pth-to-ggml.py failed with RuntimeError #35

Closed
Closed
convert-pth-to-ggml.py failed with RuntimeError#35
Labels
modelModel specific

Description

@KevinXuxuxu
KevinXuxuxu
opened on Mar 12, 2023

Hi there, I downloaded my LLaMa weights through bit-torrent, and tried to convert the 7B model to ggml FP16 format:

$python convert-pth-to-ggml.py models/7B/ 1 
normalizer.cc(51) LOG(INFO) precompiled_charsmap is empty. use identity normalization.
{'dim': 4096, 'multiple_of': 256, 'n_heads': 32, 'n_layers': 32, 'norm_eps': 1e-06, 'vocab_size': 32000}
n_parts =  1
Processing part  0
Traceback (most recent call last):
  File "/Users/fzxu/Documents/code/llama.cpp/convert-pth-to-ggml.py", line 89, in <module>
    model = torch.load(fname_model, map_location="cpu")
  File "/opt/anaconda3/envs/llama.cpp/lib/python3.10/site-packages/torch/serialization.py", line 712, in load
    return _load(opened_zipfile, map_location, pickle_module, **pickle_load_args)
  File "/opt/anaconda3/envs/llama.cpp/lib/python3.10/site-packages/torch/serialization.py", line 1049, in _load
    result = unpickler.load()
  File "/opt/anaconda3/envs/llama.cpp/lib/python3.10/site-packages/torch/serialization.py", line 1019, in persistent_load
    load_tensor(dtype, nbytes, key, _maybe_decode_ascii(location))
  File "/opt/anaconda3/envs/llama.cpp/lib/python3.10/site-packages/torch/serialization.py", line 997, in load_tensor
    storage = zip_file.get_storage_from_record(name, numel, torch._UntypedStorage).storage()._untyped()
RuntimeError: PytorchStreamReader failed reading file data/27: invalid header or archive is corrupted

Does this mean my downloaded version of model weights is corrupted? Or am I missing something?
I have filed request to Meta and hopefully I can try again with data from official download source.

Activity

  1. G2G2G2G commented on Mar 12, 2023

    @G2G2G2G
    G2G2G2G
    on Mar 12, 2023

    what is "data/27" file, that is within your models/7B folder? you downloaded the wrong thing

  2. KevinXuxuxu commented on Mar 12, 2023

    @KevinXuxuxu
    KevinXuxuxu
    on Mar 12, 2023
    Author

    Here's the file structure of my downloaded model:

    $ ls ./models 
7B                      tokenizer.model         tokenizer_checklist.chk
$ ls ./models/7B
checklist.chk       consolidated.00.pth params.json

    There isn't a directory called data and this looks normal to me.
    As for the data/27 file, it seems to be some file structure within the pth file which seems to be zipped (making some guess by checking the pytorch serialization code: https://github.com/pytorch/pytorch/blob/master/torch/serialization.py#L1112)

  3. prusnak commented on Mar 12, 2023

    @prusnak
    prusnak
    on Mar 12, 2023
    Contributor

    @KevinXuxuxu Can you post the hashes of the downloaded files?

    on Linux:

    sha256sum ./models/7B/*

    on macOS:

    shasum -a 256 ./models/7B/*

    My hashes are:

    7935c843a25ae265d60bf4543b90bfd91c4911b728412b5c1d5cff42a3cd5645  ./models/7B/checklist.chk
700df0d3013b703a806d2ae7f1bfb8e59814e3d06ae78be0c66368a50059f33d  ./models/7B/consolidated.00.pth
7e89e242ddc0dd6f060b43ca219ce8b3e8f08959a72cb3c0855df8bb04d46265  ./models/7B/params.json

  4. KevinXuxuxu commented on Mar 12, 2023

    @KevinXuxuxu
    KevinXuxuxu
    on Mar 12, 2023
    Author

    @prusnak Thanks for providing the shasum for my validation!

    $ shasum -a 256 ./models/7B/*
7935c843a25ae265d60bf4543b90bfd91c4911b728412b5c1d5cff42a3cd5645  ./models/7B/checklist.chk
008cfbd68936367b15a311494c8c8259c4902dbb461896ae767084372cdfa3fc  ./models/7B/consolidated.00.pth
7e89e242ddc0dd6f060b43ca219ce8b3e8f08959a72cb3c0855df8bb04d46265  ./models/7B/params.json

    Indeed my consolidated.00.pth file is somewhat corrupted. May I ask how you get the data? From official Meta download or bit-torrent?
    Closing this comment while I try to get a correct version of the model weights.

  5. KevinXuxuxu
    closed this as completedon Mar 12, 2023

  6. prettydeep commented on Mar 14, 2023

    @prettydeep
    prettydeep
    on Mar 14, 2023

    @prusnak Can you provide hashes for the 13B files?

  7. KevinXuxuxu commented on Mar 14, 2023

    @KevinXuxuxu
    KevinXuxuxu
    on Mar 14, 2023
    Author

    For anyone who has doubt about their data, try using https://github.com/cocktailpeanut/dalai which has the weights downloaded for you, and they seem to come from reliable source.

  8. gjmulder
    added
    modelModel specific
    on Mar 15, 2023

  9. tanishhshahh commented on Mar 17, 2023

    @tanishhshahh
    tanishhshahh
    on Mar 17, 2023

    Here's the file structure of my downloaded model:

    $ ls ./models 
7B                      tokenizer.model         tokenizer_checklist.chk
$ ls ./models/7B
checklist.chk       consolidated.00.pth params.json

    There isn't a directory called data and this looks normal to me. As for the data/27 file, it seems to be some file structure within the pth file which seems to be zipped (making some guess by checking the pytorch serialization code: https://github.com/pytorch/pytorch/blob/master/torch/serialization.py#L1112)

    Can you please provide a link to download the LLaMA files

  10. ggml-org
    deleted a comment from G2G2G2G on Mar 23, 2023
  11. 44670
    added a commit that references this issue on Aug 2, 2023

    Merge pull request ggml-org#35 from cmp-nct/tokenenizer-fix-1

    5e80f7f
  12. rururush
    added a commit that references this issue on Mar 16, 2026

    [feat] add more op (ggml-org#35)

    a1ab674
  13. phuongncn
    added a commit that references this issue on Apr 28, 2026

    Fix Zen4 Flash Attention (ggml-org#35)

    724854e
  14. wmeddie
    added a commit that references this issue on May 26, 2026

    ggml-ve : Qwen3.5/3.6 hybrid SSM+attention kernels (WIP, layer-0 corr…

    c834796
  15. Soket1
    added 2 commits that reference this issue on Jun 11, 2026

    feat(xdna): per-dispatch submit-latency breakdown probe (ggml-org#35 …

    2ee8c08

    feat(xdna): per-shape wait attribution in submit breakdown probe (ggm…

    990e163
  16. Soket1
    added 5 commits that reference this issue on Jul 1, 2026

    xdna: ggml-org#35 f3best cold-dispatch root-cause probes (measured)

    62c67fc

    xdna: ggml-org#35 f3best LOOPPROBE — confirm warm across layers in on…

    3a9b34c

    xdna: ggml-org#35 f3best UNIFIED LOOP — 16 layers in one backend call…

    9ee3ed2

    xdna: ggml-org#35 f3best loop e2e = 21.5 t/s (2x CPU); decompose non-…

    439b62a

    xdna: ggml-org#35 add npu_f3best_loop preset to correctness_test harness

    b3f3e81
  17. zhsy12345689
    added 2 commits that reference this issue on Jul 30, 2026

    Merge pull request ggml-org#35 from tc-mb/feat/omni-duplex-round1

    f0ae680

    Merge pull request ggml-org#35 from tc-mb/feat/omni-duplex-round1

    35cf946
  18. SimonTeixidor
    added a commit that references this issue on Sep 13, 2026

    Merge pull request ggml-org#35 from arc-uri-el/contribution/checkpoin…

    0007bc6
  19. SimonTeixidor
    added a commit that references this issue on Oct 12, 2026

    quantize : choose quant types to hit a bpw or file size target (upstr…

    85fc11b
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    modelModel specific

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions