70 lines
2.0 KiB
JSON
70 lines
2.0 KiB
JSON
{
|
|
"model_config": {
|
|
"name": "QwenVL",
|
|
"model_path": "Qwen/Qwen2.5-VL-3B-Instruct",
|
|
"trust_remote_code": true,
|
|
"torch_dtype": "auto",
|
|
"device_map": "cpu",
|
|
"low_cpu_mem_usage": true,
|
|
"use_cache": false,
|
|
"cache_dir": null
|
|
},
|
|
"compression_config": {
|
|
"name": "PTQ",
|
|
"quantization": {
|
|
"name": "int4_gptq",
|
|
"bits": 4,
|
|
"quant_method": {
|
|
"weight": "per-group",
|
|
"group_size": 128
|
|
},
|
|
"quant_helpers": [],
|
|
"smooth_alpha": 0.5,
|
|
"low_memory": false,
|
|
"modules_to_quantize": [],
|
|
"zero_point": true,
|
|
"mse_range": false,
|
|
"ignore_layers": [
|
|
"lm_head",
|
|
"model.language_model.embed_tokens"
|
|
],
|
|
"quant_analyse": false
|
|
},
|
|
"cache": null
|
|
},
|
|
"dataset_config": {
|
|
"name": "MultiModalDataset",
|
|
"data_path": "HuggingFaceM4/ChartQA",
|
|
"max_seq_length": 4096,
|
|
"num_samples": 256,
|
|
"batch_size": 1,
|
|
"shuffle": false
|
|
},
|
|
"global_config": {
|
|
"save_path": "./qwen2_5_vl-3b_int4_gptq",
|
|
"max_seq_length": 4096,
|
|
"hidden_size": 2048,
|
|
"model_arch_type": "qwen2_5_vl",
|
|
"deploy_backend": "vllm"
|
|
},
|
|
"infer_config": null,
|
|
"debug_info": {
|
|
"python": "3.12.11 | packaged by Anaconda, Inc. | (main, Jun 5 2025, 13:09:17) [GCC 11.2.0]",
|
|
"angelslim": {
|
|
"name": "angelslim",
|
|
"version": "ac47e5364f7f0140d538944a079de569d364fc37",
|
|
"source": "git"
|
|
},
|
|
"torch": {
|
|
"name": "torch",
|
|
"version": "2.6.0+cu124",
|
|
"source": "pip"
|
|
},
|
|
"transformers": {
|
|
"name": "transformers",
|
|
"version": "4.55.0.dev0",
|
|
"source": "pip"
|
|
},
|
|
"torch_cuda_version": "12.4"
|
|
}
|
|
} |