search

oobabooga / text-generation-webui

A Gradio web UI for Large Language Models.
GNU Affero General Public License v3.0
40.15k stars 5.27k forks source link

The same model, with the same prompt has different performance in webui and LM Studio #5966

Open AndreyRGW opened 5 months ago

AndreyRGW commented 5 months ago

Describe the bug

I checked IlyaGusev/saiga_llama3_8b_gguf, in LM Studio I get around 45-49 tokens, while in webui I get only 21 tokens.

image LM Studio ^

image Webui ^

Is there an existing issue for this?

Reproduction

Test same model in LM Studio and then in Webui

Screenshot

No response

Logs

20:08:32-676508 INFO     Loading "saiga_llama3_8b_gguf"
20:08:33-058723 INFO     llama.cpp weights detected: "models\saiga_llama3_8b_gguf\model-q4_K.gguf"
llama_model_loader: loaded meta data with 23 key-value pairs and 291 tensors from models\saiga_llama3_8b_gguf\model-q4_K.gguf (version GGUF V3 (latest))
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_model_loader: - kv   2:                           llama.vocab_size u32              = 128256
llama_model_loader: - kv   3:                       llama.context_length u32              = 8192
llama_model_loader: - kv   4:                     llama.embedding_length u32              = 4096
llama_model_loader: - kv   5:                          llama.block_count u32              = 32
llama_model_loader: - kv   6:                  llama.feed_forward_length u32              = 14336
llama_model_loader: - kv   7:                 llama.rope.dimension_count u32              = 128
llama_model_loader: - kv   8:                 llama.attention.head_count u32              = 32
llama_model_loader: - kv   9:              llama.attention.head_count_kv u32              = 8
llama_model_loader: - kv  10:     llama.attention.layer_norm_rms_epsilon f32              = 0.000010
llama_model_loader: - kv  11:                       llama.rope.freq_base f32              = 500000.000000
llama_model_loader: - kv  12:                          general.file_type u32              = 15
llama_model_loader: - kv  13:                       tokenizer.ggml.model str              = gpt2
llama_model_loader: - kv  14:                      tokenizer.ggml.tokens arr[str,128256]  = ["!", "\"", "#", "$", "%", "&", "'", ...
llama_model_loader: - kv  15:                      tokenizer.ggml.scores arr[f32,128256]  = [0.000000, 0.000000, 0.000000, 0.0000...
llama_model_loader: - kv  16:                  tokenizer.ggml.token_type arr[i32,128256]  = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
llama_model_loader: - kv  17:                      tokenizer.ggml.merges arr[str,280147]  = ["Ġ Ġ", "Ġ ĠĠĠ", "ĠĠ ĠĠ", "...
llama_model_loader: - kv  18:                tokenizer.ggml.bos_token_id u32              = 128000
llama_model_loader: - kv  19:                tokenizer.ggml.eos_token_id u32              = 128009
llama_model_loader: - kv  20:            tokenizer.ggml.padding_token_id u32              = 128000
llama_model_loader: - kv  21:                    tokenizer.chat_template str              = {% set loop_messages = messages %}{% ...
llama_model_loader: - kv  22:               general.quantization_version u32              = 2
llama_model_loader: - type  f32:   65 tensors
llama_model_loader: - type q4_K:  193 tensors
llama_model_loader: - type q6_K:   33 tensors
llm_load_vocab: special tokens definition check successful ( 256/128256 ).
llm_load_print_meta: format           = GGUF V3 (latest)
llm_load_print_meta: arch             = llama
llm_load_print_meta: vocab type       = BPE
llm_load_print_meta: n_vocab          = 128256
llm_load_print_meta: n_merges         = 280147
llm_load_print_meta: n_ctx_train      = 8192
llm_load_print_meta: n_embd           = 4096
llm_load_print_meta: n_head           = 32
llm_load_print_meta: n_head_kv        = 8
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            = 4
llm_load_print_meta: n_embd_k_gqa     = 1024
llm_load_print_meta: n_embd_v_gqa     = 1024
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: f_logit_scale    = 0.0e+00
llm_load_print_meta: n_ff             = 14336
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  = 500000.0
llm_load_print_meta: freq_scale_train = 1
llm_load_print_meta: n_yarn_orig_ctx  = 8192
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       = 8B
llm_load_print_meta: model ftype      = Q4_K - Medium
llm_load_print_meta: model params     = 8.03 B
llm_load_print_meta: model size       = 4.58 GiB (4.89 BPW)
llm_load_print_meta: general.name     = ..
llm_load_print_meta: BOS token        = 128000 '<|begin_of_text|>'
llm_load_print_meta: EOS token        = 128009 '<|eot_id|>'
llm_load_print_meta: PAD token        = 128000 '<|begin_of_text|>'
llm_load_print_meta: LF token         = 128 'Ä'
llm_load_print_meta: EOT token        = 128009 '<|eot_id|>'
ggml_cuda_init: GGML_CUDA_FORCE_MMQ:   yes
ggml_cuda_init: CUDA_USE_TENSOR_CORES: no
ggml_cuda_init: found 1 CUDA devices:
  Device 0: NVIDIA GeForce RTX 2060, compute capability 7.5, VMM: yes
llm_load_tensors: ggml ctx size =    0.30 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 =   281.81 MiB
llm_load_tensors:      CUDA0 buffer size =  4403.49 MiB
........................................................................................
llama_new_context_with_model: n_ctx      = 8192
llama_new_context_with_model: n_batch    = 512
llama_new_context_with_model: n_ubatch   = 512
llama_new_context_with_model: freq_base  = 500000.0
llama_new_context_with_model: freq_scale = 1
llama_kv_cache_init:      CUDA0 KV buffer size =  1024.00 MiB
llama_new_context_with_model: KV self size  = 1024.00 MiB, K (f16):  512.00 MiB, V (f16):  512.00 MiB
llama_new_context_with_model:  CUDA_Host  output buffer size =     0.49 MiB
llama_new_context_with_model:      CUDA0 compute buffer size =   560.00 MiB
llama_new_context_with_model:  CUDA_Host compute buffer size =    24.01 MiB
llama_new_context_with_model: graph nodes  = 1030
llama_new_context_with_model: graph splits = 2
AVX = 1 | AVX_VNNI = 0 | AVX2 = 1 | AVX512 = 0 | AVX512_VBMI = 0 | AVX512_VNNI = 0 | FMA = 1 | NEON = 0 | ARM_FMA = 0 | F16C = 1 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 1 | SSE3 = 1 | SSSE3 = 0 | VSX = 0 | MATMUL_INT8 = 0 |
Model metadata: {'general.name': '..', 'general.architecture': 'llama', 'llama.block_count': '32', 'llama.vocab_size': '128256', 'llama.context_length': '8192', 'llama.rope.dimension_count': '128', 'llama.embedding_length': '4096', 'llama.feed_forward_length': '14336', 'llama.attention.head_count': '32', 'tokenizer.ggml.eos_token_id': '128009', 'general.file_type': '15', 'llama.attention.head_count_kv': '8', 'llama.attention.layer_norm_rms_epsilon': '0.000010', 'llama.rope.freq_base': '500000.000000', 'tokenizer.ggml.model': 'gpt2', 'general.quantization_version': '2', 'tokenizer.ggml.bos_token_id': '128000', 'tokenizer.ggml.padding_token_id': '128000', 'tokenizer.chat_template': "{% set loop_messages = messages %}{% for message in loop_messages %}{% set content = '<|start_header_id|>' + message['role'] + '<|end_header_id|>\n\n'+ message['content'] | trim + '<|eot_id|>' %}{% if loop.index0 == 0 %}{% set content = bos_token + content %}{% endif %}{{ content }}{% endfor %}{% if add_generation_prompt %}{{ '<|start_header_id|>assistant<|end_header_id|>\n\n' }}{% endif %}"}
Guessed chat format: llama-3
20:08:44-026782 INFO     LOADER: "llama.cpp"
20:08:44-029291 INFO     TRUNCATION LENGTH: 8192
20:08:44-030042 INFO     INSTRUCTION TEMPLATE: "Custom (obtained from model metadata)"
20:08:44-031543 INFO     Loaded the model in 11.35 seconds.
Output generated in 6.25 seconds (20.92 tokens/s, 138 tokens, context 468, seed 246950652)

System Info

OS - Win 11 23H2
GPU - RTX 2060 12G
EsterLoken commented 5 months ago

Not much to add other than that I've noticed the same while playing around with llama3. LMStudio generates at around 70-80 tok/s while the same model (Meta-Llama-3-8B-Instruct-Q8_0.gguf) in webui tops out at roughly half of that, around 45 tok/s.

RTX 4090, Win 10 22H2

kokoabear commented 1 week ago

I logged in merely to see if anyone else had seen this. Ooba runs around 9tok/s and LMStudio runs around 41tok/s. Tried with two different models too and both do better on LMStudio (Q4 quants of Llama and a random model, LemonadeRP)