search

withcatai / node-llama-cpp

Run AI models locally on your machine with node.js bindings for llama.cpp. Force a JSON schema on the model output on the generation level
https://withcatai.github.io/node-llama-cpp/
MIT License
736 stars 63 forks source link

kv slot none #190

Closed kelvinwop closed 3 months ago

kelvinwop commented 3 months ago

Issue description

[Error: could not find a KV slot for the batch (try reducing the size of the batch or increase the context)]

Expected Behavior

to give model response

Actual Behavior

[Error: could not find a KV slot for the batch (try reducing the size of the batch or increase the context)] crash

Steps to reproduce

uses cuda

My Environment

llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output. llama_model_loader: - kv 0: general.architecture str = llama llama_model_loader: - kv 1: general.name str = LLaMA v2 llama_model_loader: - kv 2: llama.context_length u32 = 4096 llama_model_loader: - kv 3: llama.embedding_length u32 = 4096 llama_model_loader: - kv 4: llama.block_count u32 = 32 llama_model_loader: - kv 5: llama.feed_forward_length u32 = 11008 llama_model_loader: - kv 6: llama.rope.dimension_count u32 = 128 llama_model_loader: - kv 7: llama.attention.head_count u32 = 32 llama_model_loader: - kv 8: llama.attention.head_count_kv u32 = 32 llama_model_loader: - kv 9: llama.attention.layer_norm_rms_epsilon f32 = 0.000001 llama_model_loader: - kv 10: general.file_type u32 = 8 llama_model_loader: - kv 11: tokenizer.ggml.model str = llama llama_model_loader: - kv 12: tokenizer.ggml.tokens arr[str,32000] = ["", "", "", "<0x00>", "<... llama_model_loader: - kv 13: tokenizer.ggml.scores arr[f32,32000] = [0.000000, 0.000000, 0.000000, 0.0000... llama_model_loader: - kv 14: tokenizer.ggml.token_type arr[i32,32000] = [2, 3, 3, 6, 6, 6, 6, 6, 6, 6, 6, 6, ... llama_model_loader: - kv 15: tokenizer.ggml.bos_token_id u32 = 1 llama_model_loader: - kv 16: tokenizer.ggml.eos_token_id u32 = 2 llama_model_loader: - kv 17: tokenizer.ggml.unknown_token_id u32 = 0 llama_model_loader: - kv 18: general.quantization_version u32 = 2 llama_model_loader: - type f32: 65 tensors llama_model_loader: - type q5_0: 225 tensors llama_model_loader: - type q6_K: 1 tensors llm_load_vocab: special tokens definition check successful ( 259/32000 ). llm_load_print_meta: format = GGUF V2 llm_load_print_meta: arch = llama llm_load_print_meta: vocab type = SPM llm_load_print_meta: n_vocab = 32000 llm_load_print_meta: n_merges = 0 llm_load_print_meta: n_ctx_train = 4096 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_embd_head_k = 128 llm_load_print_meta: n_embd_head_v = 128 llm_load_print_meta: n_gqa = 1 llm_load_print_meta: n_embd_k_gqa = 4096 llm_load_print_meta: n_embd_v_gqa = 4096 llm_load_print_meta: f_norm_eps = 0.0e+00 llm_load_print_meta: f_norm_rms_eps = 1.0e-06 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: f_logit_scale = 0.0e+00 llm_load_print_meta: n_ff = 11008 llm_load_print_meta: n_expert = 0 llm_load_print_meta: n_expert_used = 0 llm_load_print_meta: causal attn = 1 llm_load_print_meta: pooling type = 0 llm_load_print_meta: rope type = 0 llm_load_print_meta: rope scaling = linear llm_load_print_meta: freq_base_train = 10000.0 llm_load_print_meta: freq_scale_train = 1 llm_load_print_meta: n_yarn_orig_ctx = 4096 llm_load_print_meta: rope_finetuned = unknown llm_load_print_meta: ssm_d_conv = 0 llm_load_print_meta: ssm_d_inner = 0 llm_load_print_meta: ssm_d_state = 0 llm_load_print_meta: ssm_dt_rank = 0 llm_load_print_meta: model type = 7B llm_load_print_meta: model ftype = Q5_0 llm_load_print_meta: model params = 6.74 B llm_load_print_meta: model size = 4.33 GiB (5.52 BPW) llm_load_print_meta: general.name = LLaMA v2 llm_load_print_meta: BOS token = 1 '' llm_load_print_meta: EOS token = 2 '' llm_load_print_meta: UNK token = 0 '' llm_load_print_meta: LF token = 13 '<0x0A>' ggml_cuda_init: GGML_CUDA_FORCE_MMQ: no ggml_cuda_init: CUDA_USE_TENSOR_CORES: yes ggml_cuda_init: found 1 CUDA devices: Device 0: NVIDIA GeForce RTX 3080, compute capability 8.6, VMM: yes llm_load_tensors: ggml ctx size = 0.22 MiB llm_load_tensors: offloading 32 repeating layers to GPU llm_load_tensors: offloading non-repeating layers to GPU llm_load_tensors: offloaded 33/33 layers to GPU llm_load_tensors: CPU buffer size = 85.94 MiB llm_load_tensors: CUDA0 buffer size = 4349.56 MiB .................................................................................................. llama_new_context_with_model: n_ctx = 4096 llama_new_context_with_model: n_batch = 2048 llama_new_context_with_model: n_ubatch = 512 llama_new_context_with_model: freq_base = 10000.0 llama_new_context_with_model: freq_scale = 1 llama_kv_cache_init: CUDA0 KV buffer size = 2048.00 MiB llama_new_context_with_model: KV self size = 2048.00 MiB, K (f16): 1024.00 MiB, V (f16): 1024.00 MiB llama_new_context_with_model: CUDA_Host output buffer size = 250.00 MiB llama_new_context_with_model: CUDA0 compute buffer size = 296.00 MiB llama_new_context_with_model: CUDA_Host compute buffer size = 16.00 MiB llama_new_context_with_model: graph nodes = 1060 llama_new_context_with_model: graph splits = 2

Additional Context

No response

Relevant Features Used

Are you willing to resolve this issue by submitting a Pull Request?

Yes, I have the time, but I don't know how to start. I would need guidance.

kelvinwop commented 3 months ago

seems to be fixed by reloading the model periodically:

context = new LlamaContext({model});
session = new LlamaChatSession({context});```

not sure if other more efficient fixes exist

for (let i=0; i<100; i++) {
  console.log(`--------batch ${i} -----------`)
  let resp = await prompt(session, 'testing testing 123!');
  session = resp.session;
}

causes the error every 11 prompts

giladgd commented 3 months ago

This is already solved in the version 3 beta