dinonovak
commented
7 months ago I am trying to serve model with
python -O -u -m vllm.entrypoints.openai.api_server \ --host=127.0.0.1 \ --port=8000 \ --model=$MODEL_DIR \ --tokenizer=hf-internal-testing/llama-tokenizer
but error returned is:
INFO 12-31 11:44:10 api_server.py:719] args: Namespace(host='127.0.0.1', port=8000, allow_credentials=False, allowed_origins=[''], allowed_methods=[''], allowed_headers=[''], served_model_name=None, chat_template=None, response_role='assistant', model='/mistral/mistralai-Instruct-mssql', tokenizer='hf-internal-testing/llama-tokenizer', revision=None, tokenizer_revision=None, tokenizer_mode='auto', trust_remote_code=False, download_dir=None, load_format='auto', dtype='auto', max_model_len=None, worker_use_ray=False, pipeline_parallel_size=1, tensor_parallel_size=1, max_parallel_loading_workers=None, block_size=16, seed=0, swap_space=4, gpu_memory_utilization=0.9, max_num_batched_tokens=None, max_num_seqs=256, max_paddings=256, disable_log_stats=False, quantization=None, enforce_eager=False, max_context_len_to_capture=8192, engine_use_ray=False, disable_log_requests=False, max_log_len=None)
INFO 12-31 11:44:10 llm_engine.py:73] Initializing an LLM engine with config: model='/mistral/mistralai-Instruct-mssql', tokenizer='hf-internal-testing/llama-tokenizer', tokenizer_mode=auto, revision=None, tokenizer_revision=None, trust_remote_code=False, dtype=torch.float16, max_seq_len=32768, download_dir=None, load_format=auto, tensor_parallel_size=1, quantization=None, enforce_eager=False, seed=0)
tokenizer_config.json: 100%|█████████████████████████████████████████████████████████████████████████████| 700/700 [00:00<00:00, 2.81MB/s]
tokenizer.model: 100%|█████████████████████████████████████████████████████████████████████████████████| 500k/500k [00:00<00:00, 14.6MB/s]
tokenizer.json: 100%|████████████████████████████████████████████████████████████████████████████████| 1.84M/1.84M [00:00<00:00, 3.59MB/s]
special_tokens_map.json: 100%|███████████████████████████████████████████████████████████████████████████| 411/411 [00:00<00:00, 2.13MB/s]
Traceback (most recent call last):
File "/opt/conda/lib/python3.10/runpy.py", line 196, in _run_module_as_main
return _run_code(code, main_globals, None,
File "/opt/conda/lib/python3.10/runpy.py", line 86, in _run_code
exec(code, run_globals)
File "/opt/conda/lib/python3.10/site-packages/vllm/entrypoints/openai/api_server.py", line 729, in
Deanozk
I am training model based on "mistralai/Mistral-7B-Instruct-v0.1"
but I am unable to serve it using docker image vllm/vllm-openai:latest
I am exe using python3 -m vllm.entrypoints.openai.api_server --model --gpu-memory-utilization 0.90
I tried rebooting the instant and now I am getting only following error constantly repeating: ../aten/src/ATen/native/cuda/ScatterGatherKernel.cu:144: operator(): block: [8970,0,0], thread: [36,0,0] Assertion
idx_dim >= 0 && idx_dim < index_size && "index out of bounds"failed. ../aten/src/ATen/native/cuda/ScatterGatherKernel.cu:144: operator(): block: [6527,0,0], thread: [34,0,0] Assertionidx_dim >= 0 && idx_dim < index_size && "index out of bounds"failed. ...training parameters below:
################################################################################
QLoRA parameters
################################################################################
LoRA attention dimension
lora_r = 64
Alpha parameter for LoRA scaling
lora_alpha = 16
Dropout probability for LoRA layers
lora_dropout = 0.1
################################################################################
bitsandbytes parameters
################################################################################
Activate 4-bit precision base model loading
use_4bit = True
Compute dtype for 4-bit base models
bnb_4bit_compute_dtype = "float16"
Quantization type (fp4 or nf4)
bnb_4bit_quant_type = "nf4"
Activate nested quantization for 4-bit base models (double quantization)
use_nested_quant = False
################################################################################
TrainingArguments parameters
################################################################################
Output directory where the model predictions and checkpoints will be stored
output_dir = "./results"
Number of training epochs
num_train_epochs = 1
Enable fp16/bf16 training (set bf16 to True with an A100)
fp16 = False bf16 = False
Batch size per GPU for training
per_device_train_batch_size = 4
Batch size per GPU for evaluation
per_device_eval_batch_size = 4
Number of update steps to accumulate the gradients for
gradient_accumulation_steps = 1
Enable gradient checkpointing
gradient_checkpointing = True
Maximum gradient normal (gradient clipping)
max_grad_norm = 0.3
Initial learning rate (AdamW optimizer)
learning_rate = 2e-4
Weight decay to apply to all layers except bias/LayerNorm weights
weight_decay = 0.001
Optimizer to use
optim = "paged_adamw_32bit"
Learning rate schedule (constant a bit better than cosine)
lr_scheduler_type = "constant"
Number of training steps (overrides num_train_epochs)
max_steps = -1
Ratio of steps for a linear warmup (from 0 to learning rate)
warmup_ratio = 0.03
Group sequences into batches with same length
Saves memory and speeds up training considerably
group_by_length = True
Save checkpoint every X updates steps
save_steps = 25
Log every X updates steps
logging_steps = 25
################################################################################
SFT parameters
################################################################################
Maximum sequence length to use
max_seq_length = None
Pack multiple short examples in the same input sequence to increase efficiency
packing = False
Load the entire model on the GPU 0
device_map = {"": 0}