Upload config.json with huggingface_hub
Browse files- config.json +88 -0
config.json
ADDED
|
@@ -0,0 +1,88 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "vinai/PhoGPT-4B-Chat",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"MPTForCausalLM"
|
| 5 |
+
],
|
| 6 |
+
"attn_config": {
|
| 7 |
+
"alibi": true,
|
| 8 |
+
"alibi_bias_max": 8,
|
| 9 |
+
"attn_impl": "torch",
|
| 10 |
+
"attn_pdrop": 0.0,
|
| 11 |
+
"attn_type": "multihead_attention",
|
| 12 |
+
"attn_uses_sequence_id": false,
|
| 13 |
+
"clip_qkv": null,
|
| 14 |
+
"prefix_lm": false,
|
| 15 |
+
"qk_gn": false,
|
| 16 |
+
"qk_ln": false,
|
| 17 |
+
"rope": false,
|
| 18 |
+
"rope_dail_config": {
|
| 19 |
+
"pos_idx_in_fp32": true,
|
| 20 |
+
"type": "original",
|
| 21 |
+
"xpos_scale_base": 512
|
| 22 |
+
},
|
| 23 |
+
"rope_hf_config": {
|
| 24 |
+
"factor": 1.0,
|
| 25 |
+
"type": "no_scaling"
|
| 26 |
+
},
|
| 27 |
+
"rope_impl": "dail",
|
| 28 |
+
"rope_theta": 10000,
|
| 29 |
+
"sliding_window_size": -1,
|
| 30 |
+
"softmax_scale": null
|
| 31 |
+
},
|
| 32 |
+
"auto_map": {
|
| 33 |
+
"AutoConfig": "vinai/PhoGPT-4B-Chat--configuration_mpt.MPTConfig",
|
| 34 |
+
"AutoModelForCausalLM": "vinai/PhoGPT-4B-Chat--modeling_mpt.MPTForCausalLM"
|
| 35 |
+
},
|
| 36 |
+
"d_model": 3072,
|
| 37 |
+
"emb_pdrop": 0.0,
|
| 38 |
+
"embedding_fraction": 1.0,
|
| 39 |
+
"expansion_ratio": 4,
|
| 40 |
+
"fc_type": "torch",
|
| 41 |
+
"ffn_config": {
|
| 42 |
+
"fc_type": "torch",
|
| 43 |
+
"ffn_type": "mptmlp"
|
| 44 |
+
},
|
| 45 |
+
"init_config": {
|
| 46 |
+
"emb_init_std": null,
|
| 47 |
+
"emb_init_uniform_lim": null,
|
| 48 |
+
"fan_mode": "fan_in",
|
| 49 |
+
"init_div_is_residual": true,
|
| 50 |
+
"init_gain": 0.0,
|
| 51 |
+
"init_nonlinearity": "relu",
|
| 52 |
+
"init_std": null,
|
| 53 |
+
"name": "kaiming_normal_",
|
| 54 |
+
"verbose": 0
|
| 55 |
+
},
|
| 56 |
+
"init_device": "cpu",
|
| 57 |
+
"learned_pos_emb": false,
|
| 58 |
+
"logit_scale": null,
|
| 59 |
+
"max_seq_len": 8192,
|
| 60 |
+
"model_type": "mpt",
|
| 61 |
+
"n_heads": 24,
|
| 62 |
+
"n_layers": 32,
|
| 63 |
+
"no_bias": false,
|
| 64 |
+
"norm_type": "low_precision_layernorm",
|
| 65 |
+
"pretraining_tp": 1,
|
| 66 |
+
"quantization_config": {
|
| 67 |
+
"_load_in_4bit": true,
|
| 68 |
+
"_load_in_8bit": false,
|
| 69 |
+
"bnb_4bit_compute_dtype": "float16",
|
| 70 |
+
"bnb_4bit_quant_storage": "uint8",
|
| 71 |
+
"bnb_4bit_quant_type": "nf4",
|
| 72 |
+
"bnb_4bit_use_double_quant": true,
|
| 73 |
+
"llm_int8_enable_fp32_cpu_offload": false,
|
| 74 |
+
"llm_int8_has_fp16_weight": false,
|
| 75 |
+
"llm_int8_skip_modules": null,
|
| 76 |
+
"llm_int8_threshold": 6.0,
|
| 77 |
+
"load_in_4bit": true,
|
| 78 |
+
"load_in_8bit": false,
|
| 79 |
+
"quant_method": "bitsandbytes"
|
| 80 |
+
},
|
| 81 |
+
"resid_pdrop": 0.0,
|
| 82 |
+
"torch_dtype": "float32",
|
| 83 |
+
"transformers_version": "4.42.4",
|
| 84 |
+
"use_cache": false,
|
| 85 |
+
"use_pad_tok_in_ffn": true,
|
| 86 |
+
"verbose": 0,
|
| 87 |
+
"vocab_size": 20480
|
| 88 |
+
}
|