“<method 'append' of 'list' objects> returned a result with an exception set”

[GluttonousCat]

I encountered this error when using the embedding function "llama.create_embedding", under this code. abstract is a text about 700 tokens, when i use the previous version 0.1.83, it can run. now the version is 0.2.22

llama = Llama(model_path='./llama-2-7b.Q4_K_M.gguf', embedding=True, n_ctx=2048, n_gpu_layers=30)
abstract_embedding = llama.create_embedding(input=abstract).get('data')[0].get('embedding')

this is the report

Traceback (most recent call last):
  File "C:\Users\AI\PycharmProjects\roberta-gat\demo.py", line 107, in <module>
    get_abstract_embedding(path='data/validation.csv', start=0)
  File "C:\Users\AI\PycharmProjects\roberta-gat\demo.py", line 50, in get_abstract_embedding
    abstract_embedding = llama.create_embedding(input=abstract).get('data')[0].get('embedding')
                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\AI\.conda\envs\roberta-gat\Lib\site-packages\llama_cpp\llama.py", line 1309, in create_embedding
    data.append(
SystemError: <method 'append' of 'list' objects> returned a result with an exception set

This is the source code , and I find the value of llama_cpp.llama_get_embeddings(self._ctx.ctx) maybe is null?

for index, input in enumerate(inputs):
    tokens = self.tokenize(input.encode("utf-8"), special=True)
    self.reset()
    self.eval(tokens)
    n_tokens = len(tokens)
    total_tokens += n_tokens
    embedding = llama_cpp.llama_get_embeddings(self._ctx.ctx)[
        : llama_cpp.llama_n_embd(self._model.model)
    ]

    data.append(
        {
            "object": "embedding",
            "embedding": embedding,
            "index": index,
        }
    )

[CharlieChenDSS]

got the same error with llama-cpp-python v0.2.22 :

  File "/Users/xxxx/opt/anaconda3/envs/langchain_python_v3.9/lib/python3.9/site-packages/llama_cpp/llama.py", line 1309, in create_embedding
    data.append(
SystemError: <method 'append' of 'list' objects> returned a result with an error set

[CharlieChenDSS]

@abetlen could you please help look into this issue when you get a chance? Thanks.
It could be raised from below code block, function create_embedding of llama.py
embedding = llama_cpp.llama_get_embeddings(self._ctx.ctx)[ : llama_cpp.llama_n_embd(self._model.model) ]

[GaryZhen]

I meet that problem as well. I am using Chroma, which asks me to specify the embedding function.
I have to use llama-cpp-embedding from langchain. However, it always shows this problem.

[brandonrobertz]

I was able to replicate this and was able to fix on my end. Does this PR also fix for y'all? #1019

[zakki]

Thank you! I had same problem and 1019 fixed that.

[GluttonousCat]

I was able to replicate this and was able to fix on my end. Does this PR also fix for y'all? #1019

Thank you! i fixed that error by #1019