Upload folder using huggingface_hub (#1)
Browse files- 5c04bb3363cf56ba5ee5d84f27b3c818433baef844daba81336944b9d24ff782 (523ab896477255e6846c76c4e337354836194ee6)
- 38d7d68a49c36a103d86e9528d836206755966b7cd408e545932061821a976ce (2d3788763148c5ab882ee6e5767ef24e378697d3)
- README.md +85 -0
- config.json +67 -0
- configuration_chatglm.py +58 -0
- generation_config.json +10 -0
- model-00001-of-00002.safetensors +3 -0
- model-00002-of-00002.safetensors +3 -0
- model.safetensors.index.json +774 -0
- modeling_chatglm.py +898 -0
- smash_config.json +31 -0
- special_tokens_map.json +9 -0
- tokenizer.model +3 -0
- tokenizer_config.json +27 -0
README.md
ADDED
|
@@ -0,0 +1,85 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
|
| 3 |
+
base_model: THUDM/LongWriter-glm4-9b
|
| 4 |
+
metrics:
|
| 5 |
+
- memory_disk
|
| 6 |
+
- memory_inference
|
| 7 |
+
- inference_latency
|
| 8 |
+
- inference_throughput
|
| 9 |
+
- inference_CO2_emissions
|
| 10 |
+
- inference_energy_consumption
|
| 11 |
+
tags:
|
| 12 |
+
- pruna-ai
|
| 13 |
+
---
|
| 14 |
+
<!-- header start -->
|
| 15 |
+
<!-- 200823 -->
|
| 16 |
+
<div style="width: auto; margin-left: auto; margin-right: auto">
|
| 17 |
+
<a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer">
|
| 18 |
+
<img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
|
| 19 |
+
</a>
|
| 20 |
+
</div>
|
| 21 |
+
<!-- header end -->
|
| 22 |
+
|
| 23 |
+
[](https://twitter.com/PrunaAI)
|
| 24 |
+
[](https://github.com/PrunaAI)
|
| 25 |
+
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
|
| 26 |
+
[](https://discord.gg/rskEr4BZJx)
|
| 27 |
+
|
| 28 |
+
# Simply make AI models cheaper, smaller, faster, and greener!
|
| 29 |
+
|
| 30 |
+
- Give a thumbs up if you like this model!
|
| 31 |
+
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
|
| 32 |
+
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
|
| 33 |
+
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
|
| 34 |
+
- Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help.
|
| 35 |
+
|
| 36 |
+
## Results
|
| 37 |
+
|
| 38 |
+
|
| 39 |
+
|
| 40 |
+
**Frequently Asked Questions**
|
| 41 |
+
- ***How does the compression work?*** The model is compressed with llm-int8.
|
| 42 |
+
- ***How does the model quality change?*** The quality of the model output might vary compared to the base model.
|
| 43 |
+
- ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
|
| 44 |
+
- ***What is the model format?*** We use safetensors.
|
| 45 |
+
- ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data.
|
| 46 |
+
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
|
| 47 |
+
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
|
| 48 |
+
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
|
| 49 |
+
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
|
| 50 |
+
|
| 51 |
+
## Setup
|
| 52 |
+
|
| 53 |
+
You can run the smashed model with these steps:
|
| 54 |
+
|
| 55 |
+
0. Check requirements from the original repo THUDM/LongWriter-glm4-9b installed. In particular, check python, cuda, and transformers versions.
|
| 56 |
+
1. Make sure that you have installed quantization related packages.
|
| 57 |
+
```bash
|
| 58 |
+
pip install transformers accelerate bitsandbytes>0.37.0
|
| 59 |
+
```
|
| 60 |
+
2. Load & run the model.
|
| 61 |
+
```python
|
| 62 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
| 63 |
+
|
| 64 |
+
|
| 65 |
+
model = AutoModelForCausalLM.from_pretrained("PrunaAI/THUDM-LongWriter-glm4-9b-bnb-4bit-smashed", trust_remote_code=True, device_map='auto')
|
| 66 |
+
tokenizer = AutoTokenizer.from_pretrained("THUDM/LongWriter-glm4-9b")
|
| 67 |
+
|
| 68 |
+
input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"]
|
| 69 |
+
|
| 70 |
+
outputs = model.generate(input_ids, max_new_tokens=216)
|
| 71 |
+
tokenizer.decode(outputs[0])
|
| 72 |
+
```
|
| 73 |
+
|
| 74 |
+
## Configurations
|
| 75 |
+
|
| 76 |
+
The configuration info are in `smash_config.json`.
|
| 77 |
+
|
| 78 |
+
## Credits & License
|
| 79 |
+
|
| 80 |
+
The license of the smashed model follows the license of the original model. Please check the license of the original model THUDM/LongWriter-glm4-9b before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
|
| 81 |
+
|
| 82 |
+
## Want to compress other models?
|
| 83 |
+
|
| 84 |
+
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
|
| 85 |
+
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
|
config.json
ADDED
|
@@ -0,0 +1,67 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "/ceph/hdd/staff/charpent/.cache/modelsdhkyhu_nhop2bfo6",
|
| 3 |
+
"add_bias_linear": false,
|
| 4 |
+
"add_qkv_bias": true,
|
| 5 |
+
"apply_query_key_layer_scaling": true,
|
| 6 |
+
"apply_residual_connection_post_layernorm": false,
|
| 7 |
+
"architectures": [
|
| 8 |
+
"ChatGLMForConditionalGeneration"
|
| 9 |
+
],
|
| 10 |
+
"attention_dropout": 0.0,
|
| 11 |
+
"attention_softmax_in_fp32": true,
|
| 12 |
+
"auto_map": {
|
| 13 |
+
"AutoConfig": "configuration_chatglm.ChatGLMConfig",
|
| 14 |
+
"AutoModel": "THUDM/LongWriter-glm4-9b--modeling_chatglm.ChatGLMForConditionalGeneration",
|
| 15 |
+
"AutoModelForCausalLM": "modeling_chatglm.ChatGLMForConditionalGeneration",
|
| 16 |
+
"AutoModelForSeq2SeqLM": "THUDM/LongWriter-glm4-9b--modeling_chatglm.ChatGLMForConditionalGeneration",
|
| 17 |
+
"AutoModelForSequenceClassification": "THUDM/LongWriter-glm4-9b--modeling_chatglm.ChatGLMForSequenceClassification"
|
| 18 |
+
},
|
| 19 |
+
"bias_dropout_fusion": true,
|
| 20 |
+
"classifier_dropout": null,
|
| 21 |
+
"eos_token_id": [
|
| 22 |
+
151329,
|
| 23 |
+
151336,
|
| 24 |
+
151338
|
| 25 |
+
],
|
| 26 |
+
"ffn_hidden_size": 13696,
|
| 27 |
+
"fp32_residual_connection": false,
|
| 28 |
+
"hidden_dropout": 0.0,
|
| 29 |
+
"hidden_size": 4096,
|
| 30 |
+
"kv_channels": 128,
|
| 31 |
+
"layernorm_epsilon": 1.5625e-07,
|
| 32 |
+
"model_type": "chatglm",
|
| 33 |
+
"multi_query_attention": true,
|
| 34 |
+
"multi_query_group_num": 2,
|
| 35 |
+
"num_attention_heads": 32,
|
| 36 |
+
"num_hidden_layers": 40,
|
| 37 |
+
"num_layers": 40,
|
| 38 |
+
"original_rope": true,
|
| 39 |
+
"pad_token_id": 151329,
|
| 40 |
+
"padded_vocab_size": 151552,
|
| 41 |
+
"post_layer_norm": true,
|
| 42 |
+
"quantization_config": {
|
| 43 |
+
"_load_in_4bit": true,
|
| 44 |
+
"_load_in_8bit": false,
|
| 45 |
+
"bnb_4bit_compute_dtype": "bfloat16",
|
| 46 |
+
"bnb_4bit_quant_storage": "uint8",
|
| 47 |
+
"bnb_4bit_quant_type": "fp4",
|
| 48 |
+
"bnb_4bit_use_double_quant": false,
|
| 49 |
+
"llm_int8_enable_fp32_cpu_offload": false,
|
| 50 |
+
"llm_int8_has_fp16_weight": false,
|
| 51 |
+
"llm_int8_skip_modules": [
|
| 52 |
+
"lm_head"
|
| 53 |
+
],
|
| 54 |
+
"llm_int8_threshold": 6.0,
|
| 55 |
+
"load_in_4bit": true,
|
| 56 |
+
"load_in_8bit": false,
|
| 57 |
+
"quant_method": "bitsandbytes"
|
| 58 |
+
},
|
| 59 |
+
"rmsnorm": true,
|
| 60 |
+
"rope_ratio": 500,
|
| 61 |
+
"seq_length": 1048576,
|
| 62 |
+
"tie_word_embeddings": false,
|
| 63 |
+
"torch_dtype": "bfloat16",
|
| 64 |
+
"transformers_version": "4.42.4",
|
| 65 |
+
"use_cache": true,
|
| 66 |
+
"vocab_size": 151552
|
| 67 |
+
}
|
configuration_chatglm.py
ADDED
|
@@ -0,0 +1,58 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
from transformers import PretrainedConfig
|
| 2 |
+
|
| 3 |
+
|
| 4 |
+
class ChatGLMConfig(PretrainedConfig):
|
| 5 |
+
model_type = "chatglm"
|
| 6 |
+
|
| 7 |
+
def __init__(
|
| 8 |
+
self,
|
| 9 |
+
num_layers=28,
|
| 10 |
+
padded_vocab_size=65024,
|
| 11 |
+
hidden_size=4096,
|
| 12 |
+
ffn_hidden_size=13696,
|
| 13 |
+
kv_channels=128,
|
| 14 |
+
num_attention_heads=32,
|
| 15 |
+
seq_length=2048,
|
| 16 |
+
hidden_dropout=0.0,
|
| 17 |
+
classifier_dropout=None,
|
| 18 |
+
attention_dropout=0.0,
|
| 19 |
+
layernorm_epsilon=1e-5,
|
| 20 |
+
rmsnorm=True,
|
| 21 |
+
apply_residual_connection_post_layernorm=False,
|
| 22 |
+
post_layer_norm=True,
|
| 23 |
+
add_bias_linear=False,
|
| 24 |
+
add_qkv_bias=False,
|
| 25 |
+
bias_dropout_fusion=True,
|
| 26 |
+
multi_query_attention=False,
|
| 27 |
+
multi_query_group_num=1,
|
| 28 |
+
rope_ratio=1,
|
| 29 |
+
apply_query_key_layer_scaling=True,
|
| 30 |
+
attention_softmax_in_fp32=True,
|
| 31 |
+
fp32_residual_connection=False,
|
| 32 |
+
**kwargs
|
| 33 |
+
):
|
| 34 |
+
self.num_layers = num_layers
|
| 35 |
+
self.vocab_size = padded_vocab_size
|
| 36 |
+
self.padded_vocab_size = padded_vocab_size
|
| 37 |
+
self.hidden_size = hidden_size
|
| 38 |
+
self.ffn_hidden_size = ffn_hidden_size
|
| 39 |
+
self.kv_channels = kv_channels
|
| 40 |
+
self.num_attention_heads = num_attention_heads
|
| 41 |
+
self.seq_length = seq_length
|
| 42 |
+
self.hidden_dropout = hidden_dropout
|
| 43 |
+
self.classifier_dropout = classifier_dropout
|
| 44 |
+
self.attention_dropout = attention_dropout
|
| 45 |
+
self.layernorm_epsilon = layernorm_epsilon
|
| 46 |
+
self.rmsnorm = rmsnorm
|
| 47 |
+
self.apply_residual_connection_post_layernorm = apply_residual_connection_post_layernorm
|
| 48 |
+
self.post_layer_norm = post_layer_norm
|
| 49 |
+
self.add_bias_linear = add_bias_linear
|
| 50 |
+
self.add_qkv_bias = add_qkv_bias
|
| 51 |
+
self.bias_dropout_fusion = bias_dropout_fusion
|
| 52 |
+
self.multi_query_attention = multi_query_attention
|
| 53 |
+
self.multi_query_group_num = multi_query_group_num
|
| 54 |
+
self.rope_ratio = rope_ratio
|
| 55 |
+
self.apply_query_key_layer_scaling = apply_query_key_layer_scaling
|
| 56 |
+
self.attention_softmax_in_fp32 = attention_softmax_in_fp32
|
| 57 |
+
self.fp32_residual_connection = fp32_residual_connection
|
| 58 |
+
super().__init__(**kwargs)
|
generation_config.json
ADDED
|
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"eos_token_id": [
|
| 4 |
+
151329,
|
| 5 |
+
151336,
|
| 6 |
+
151338
|
| 7 |
+
],
|
| 8 |
+
"pad_token_id": 151329,
|
| 9 |
+
"transformers_version": "4.42.4"
|
| 10 |
+
}
|
model-00001-of-00002.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1e1d3c6fbd5642c24b57b9205f739fb036a1fe46e0e51826223fc0a8eceb5605
|
| 3 |
+
size 4996695308
|
model-00002-of-00002.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6b27bdda7afa891460dde7825493ceb8bddef2412a7e55213fbebe8b95856c24
|
| 3 |
+
size 1183988127
|
model.safetensors.index.json
ADDED
|
@@ -0,0 +1,774 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"metadata": {
|
| 3 |
+
"total_size": 6180576195
|
| 4 |
+
},
|
| 5 |
+
"weight_map": {
|
| 6 |
+
"transformer.embedding.word_embeddings.weight": "model-00001-of-00002.safetensors",
|
| 7 |
+
"transformer.encoder.final_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 8 |
+
"transformer.encoder.layers.0.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 9 |
+
"transformer.encoder.layers.0.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 10 |
+
"transformer.encoder.layers.0.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 11 |
+
"transformer.encoder.layers.0.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 12 |
+
"transformer.encoder.layers.0.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 13 |
+
"transformer.encoder.layers.0.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 14 |
+
"transformer.encoder.layers.0.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 15 |
+
"transformer.encoder.layers.0.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 16 |
+
"transformer.encoder.layers.0.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 17 |
+
"transformer.encoder.layers.0.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 18 |
+
"transformer.encoder.layers.0.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 19 |
+
"transformer.encoder.layers.0.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 20 |
+
"transformer.encoder.layers.0.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 21 |
+
"transformer.encoder.layers.0.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 22 |
+
"transformer.encoder.layers.0.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 23 |
+
"transformer.encoder.layers.0.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 24 |
+
"transformer.encoder.layers.0.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 25 |
+
"transformer.encoder.layers.0.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 26 |
+
"transformer.encoder.layers.0.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 27 |
+
"transformer.encoder.layers.1.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 28 |
+
"transformer.encoder.layers.1.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 29 |
+
"transformer.encoder.layers.1.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 30 |
+
"transformer.encoder.layers.1.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 31 |
+
"transformer.encoder.layers.1.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 32 |
+
"transformer.encoder.layers.1.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 33 |
+
"transformer.encoder.layers.1.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 34 |
+
"transformer.encoder.layers.1.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 35 |
+
"transformer.encoder.layers.1.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 36 |
+
"transformer.encoder.layers.1.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 37 |
+
"transformer.encoder.layers.1.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 38 |
+
"transformer.encoder.layers.1.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 39 |
+
"transformer.encoder.layers.1.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 40 |
+
"transformer.encoder.layers.1.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 41 |
+
"transformer.encoder.layers.1.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 42 |
+
"transformer.encoder.layers.1.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 43 |
+
"transformer.encoder.layers.1.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 44 |
+
"transformer.encoder.layers.1.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 45 |
+
"transformer.encoder.layers.1.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 46 |
+
"transformer.encoder.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 47 |
+
"transformer.encoder.layers.10.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 48 |
+
"transformer.encoder.layers.10.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 49 |
+
"transformer.encoder.layers.10.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 50 |
+
"transformer.encoder.layers.10.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 51 |
+
"transformer.encoder.layers.10.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 52 |
+
"transformer.encoder.layers.10.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 53 |
+
"transformer.encoder.layers.10.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 54 |
+
"transformer.encoder.layers.10.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 55 |
+
"transformer.encoder.layers.10.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 56 |
+
"transformer.encoder.layers.10.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 57 |
+
"transformer.encoder.layers.10.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 58 |
+
"transformer.encoder.layers.10.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 59 |
+
"transformer.encoder.layers.10.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 60 |
+
"transformer.encoder.layers.10.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 61 |
+
"transformer.encoder.layers.10.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 62 |
+
"transformer.encoder.layers.10.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 63 |
+
"transformer.encoder.layers.10.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 64 |
+
"transformer.encoder.layers.10.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 65 |
+
"transformer.encoder.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 66 |
+
"transformer.encoder.layers.11.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 67 |
+
"transformer.encoder.layers.11.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 68 |
+
"transformer.encoder.layers.11.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 69 |
+
"transformer.encoder.layers.11.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 70 |
+
"transformer.encoder.layers.11.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 71 |
+
"transformer.encoder.layers.11.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 72 |
+
"transformer.encoder.layers.11.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 73 |
+
"transformer.encoder.layers.11.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 74 |
+
"transformer.encoder.layers.11.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 75 |
+
"transformer.encoder.layers.11.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 76 |
+
"transformer.encoder.layers.11.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 77 |
+
"transformer.encoder.layers.11.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 78 |
+
"transformer.encoder.layers.11.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 79 |
+
"transformer.encoder.layers.11.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 80 |
+
"transformer.encoder.layers.11.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 81 |
+
"transformer.encoder.layers.11.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 82 |
+
"transformer.encoder.layers.11.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 83 |
+
"transformer.encoder.layers.11.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 84 |
+
"transformer.encoder.layers.12.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 85 |
+
"transformer.encoder.layers.12.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 86 |
+
"transformer.encoder.layers.12.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 87 |
+
"transformer.encoder.layers.12.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 88 |
+
"transformer.encoder.layers.12.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 89 |
+
"transformer.encoder.layers.12.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 90 |
+
"transformer.encoder.layers.12.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 91 |
+
"transformer.encoder.layers.12.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 92 |
+
"transformer.encoder.layers.12.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 93 |
+
"transformer.encoder.layers.12.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 94 |
+
"transformer.encoder.layers.12.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 95 |
+
"transformer.encoder.layers.12.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 96 |
+
"transformer.encoder.layers.12.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 97 |
+
"transformer.encoder.layers.12.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 98 |
+
"transformer.encoder.layers.12.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 99 |
+
"transformer.encoder.layers.12.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 100 |
+
"transformer.encoder.layers.12.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 101 |
+
"transformer.encoder.layers.12.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 102 |
+
"transformer.encoder.layers.12.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 103 |
+
"transformer.encoder.layers.13.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 104 |
+
"transformer.encoder.layers.13.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 105 |
+
"transformer.encoder.layers.13.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 106 |
+
"transformer.encoder.layers.13.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 107 |
+
"transformer.encoder.layers.13.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 108 |
+
"transformer.encoder.layers.13.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 109 |
+
"transformer.encoder.layers.13.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 110 |
+
"transformer.encoder.layers.13.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 111 |
+
"transformer.encoder.layers.13.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 112 |
+
"transformer.encoder.layers.13.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 113 |
+
"transformer.encoder.layers.13.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 114 |
+
"transformer.encoder.layers.13.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 115 |
+
"transformer.encoder.layers.13.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 116 |
+
"transformer.encoder.layers.13.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 117 |
+
"transformer.encoder.layers.13.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 118 |
+
"transformer.encoder.layers.13.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 119 |
+
"transformer.encoder.layers.13.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 120 |
+
"transformer.encoder.layers.13.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 121 |
+
"transformer.encoder.layers.13.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 122 |
+
"transformer.encoder.layers.14.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 123 |
+
"transformer.encoder.layers.14.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 124 |
+
"transformer.encoder.layers.14.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 125 |
+
"transformer.encoder.layers.14.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 126 |
+
"transformer.encoder.layers.14.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 127 |
+
"transformer.encoder.layers.14.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 128 |
+
"transformer.encoder.layers.14.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 129 |
+
"transformer.encoder.layers.14.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 130 |
+
"transformer.encoder.layers.14.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 131 |
+
"transformer.encoder.layers.14.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 132 |
+
"transformer.encoder.layers.14.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 133 |
+
"transformer.encoder.layers.14.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 134 |
+
"transformer.encoder.layers.14.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 135 |
+
"transformer.encoder.layers.14.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 136 |
+
"transformer.encoder.layers.14.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 137 |
+
"transformer.encoder.layers.14.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 138 |
+
"transformer.encoder.layers.14.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 139 |
+
"transformer.encoder.layers.14.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 140 |
+
"transformer.encoder.layers.14.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 141 |
+
"transformer.encoder.layers.15.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 142 |
+
"transformer.encoder.layers.15.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 143 |
+
"transformer.encoder.layers.15.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 144 |
+
"transformer.encoder.layers.15.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 145 |
+
"transformer.encoder.layers.15.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 146 |
+
"transformer.encoder.layers.15.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 147 |
+
"transformer.encoder.layers.15.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 148 |
+
"transformer.encoder.layers.15.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 149 |
+
"transformer.encoder.layers.15.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 150 |
+
"transformer.encoder.layers.15.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 151 |
+
"transformer.encoder.layers.15.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 152 |
+
"transformer.encoder.layers.15.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 153 |
+
"transformer.encoder.layers.15.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 154 |
+
"transformer.encoder.layers.15.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 155 |
+
"transformer.encoder.layers.15.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 156 |
+
"transformer.encoder.layers.15.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 157 |
+
"transformer.encoder.layers.15.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 158 |
+
"transformer.encoder.layers.15.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 159 |
+
"transformer.encoder.layers.15.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 160 |
+
"transformer.encoder.layers.16.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 161 |
+
"transformer.encoder.layers.16.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 162 |
+
"transformer.encoder.layers.16.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 163 |
+
"transformer.encoder.layers.16.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 164 |
+
"transformer.encoder.layers.16.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 165 |
+
"transformer.encoder.layers.16.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 166 |
+
"transformer.encoder.layers.16.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 167 |
+
"transformer.encoder.layers.16.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 168 |
+
"transformer.encoder.layers.16.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 169 |
+
"transformer.encoder.layers.16.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 170 |
+
"transformer.encoder.layers.16.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 171 |
+
"transformer.encoder.layers.16.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 172 |
+
"transformer.encoder.layers.16.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 173 |
+
"transformer.encoder.layers.16.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 174 |
+
"transformer.encoder.layers.16.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 175 |
+
"transformer.encoder.layers.16.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 176 |
+
"transformer.encoder.layers.16.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 177 |
+
"transformer.encoder.layers.16.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 178 |
+
"transformer.encoder.layers.16.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 179 |
+
"transformer.encoder.layers.17.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 180 |
+
"transformer.encoder.layers.17.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 181 |
+
"transformer.encoder.layers.17.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 182 |
+
"transformer.encoder.layers.17.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 183 |
+
"transformer.encoder.layers.17.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 184 |
+
"transformer.encoder.layers.17.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 185 |
+
"transformer.encoder.layers.17.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 186 |
+
"transformer.encoder.layers.17.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 187 |
+
"transformer.encoder.layers.17.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 188 |
+
"transformer.encoder.layers.17.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 189 |
+
"transformer.encoder.layers.17.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 190 |
+
"transformer.encoder.layers.17.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 191 |
+
"transformer.encoder.layers.17.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 192 |
+
"transformer.encoder.layers.17.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 193 |
+
"transformer.encoder.layers.17.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 194 |
+
"transformer.encoder.layers.17.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 195 |
+
"transformer.encoder.layers.17.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 196 |
+
"transformer.encoder.layers.17.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 197 |
+
"transformer.encoder.layers.17.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 198 |
+
"transformer.encoder.layers.18.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 199 |
+
"transformer.encoder.layers.18.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 200 |
+
"transformer.encoder.layers.18.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 201 |
+
"transformer.encoder.layers.18.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 202 |
+
"transformer.encoder.layers.18.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 203 |
+
"transformer.encoder.layers.18.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 204 |
+
"transformer.encoder.layers.18.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 205 |
+
"transformer.encoder.layers.18.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 206 |
+
"transformer.encoder.layers.18.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 207 |
+
"transformer.encoder.layers.18.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 208 |
+
"transformer.encoder.layers.18.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 209 |
+
"transformer.encoder.layers.18.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 210 |
+
"transformer.encoder.layers.18.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 211 |
+
"transformer.encoder.layers.18.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 212 |
+
"transformer.encoder.layers.18.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 213 |
+
"transformer.encoder.layers.18.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 214 |
+
"transformer.encoder.layers.18.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 215 |
+
"transformer.encoder.layers.18.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 216 |
+
"transformer.encoder.layers.18.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 217 |
+
"transformer.encoder.layers.19.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 218 |
+
"transformer.encoder.layers.19.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 219 |
+
"transformer.encoder.layers.19.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 220 |
+
"transformer.encoder.layers.19.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 221 |
+
"transformer.encoder.layers.19.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 222 |
+
"transformer.encoder.layers.19.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 223 |
+
"transformer.encoder.layers.19.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 224 |
+
"transformer.encoder.layers.19.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 225 |
+
"transformer.encoder.layers.19.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 226 |
+
"transformer.encoder.layers.19.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 227 |
+
"transformer.encoder.layers.19.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 228 |
+
"transformer.encoder.layers.19.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 229 |
+
"transformer.encoder.layers.19.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 230 |
+
"transformer.encoder.layers.19.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 231 |
+
"transformer.encoder.layers.19.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 232 |
+
"transformer.encoder.layers.19.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 233 |
+
"transformer.encoder.layers.19.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 234 |
+
"transformer.encoder.layers.19.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 235 |
+
"transformer.encoder.layers.19.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 236 |
+
"transformer.encoder.layers.2.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 237 |
+
"transformer.encoder.layers.2.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 238 |
+
"transformer.encoder.layers.2.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 239 |
+
"transformer.encoder.layers.2.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 240 |
+
"transformer.encoder.layers.2.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 241 |
+
"transformer.encoder.layers.2.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 242 |
+
"transformer.encoder.layers.2.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 243 |
+
"transformer.encoder.layers.2.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 244 |
+
"transformer.encoder.layers.2.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 245 |
+
"transformer.encoder.layers.2.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 246 |
+
"transformer.encoder.layers.2.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 247 |
+
"transformer.encoder.layers.2.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 248 |
+
"transformer.encoder.layers.2.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 249 |
+
"transformer.encoder.layers.2.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 250 |
+
"transformer.encoder.layers.2.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 251 |
+
"transformer.encoder.layers.2.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 252 |
+
"transformer.encoder.layers.2.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 253 |
+
"transformer.encoder.layers.2.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 254 |
+
"transformer.encoder.layers.2.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 255 |
+
"transformer.encoder.layers.20.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 256 |
+
"transformer.encoder.layers.20.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 257 |
+
"transformer.encoder.layers.20.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 258 |
+
"transformer.encoder.layers.20.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 259 |
+
"transformer.encoder.layers.20.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 260 |
+
"transformer.encoder.layers.20.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 261 |
+
"transformer.encoder.layers.20.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 262 |
+
"transformer.encoder.layers.20.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 263 |
+
"transformer.encoder.layers.20.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 264 |
+
"transformer.encoder.layers.20.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 265 |
+
"transformer.encoder.layers.20.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 266 |
+
"transformer.encoder.layers.20.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 267 |
+
"transformer.encoder.layers.20.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 268 |
+
"transformer.encoder.layers.20.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 269 |
+
"transformer.encoder.layers.20.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 270 |
+
"transformer.encoder.layers.20.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 271 |
+
"transformer.encoder.layers.20.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 272 |
+
"transformer.encoder.layers.20.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 273 |
+
"transformer.encoder.layers.20.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 274 |
+
"transformer.encoder.layers.21.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 275 |
+
"transformer.encoder.layers.21.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 276 |
+
"transformer.encoder.layers.21.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 277 |
+
"transformer.encoder.layers.21.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 278 |
+
"transformer.encoder.layers.21.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 279 |
+
"transformer.encoder.layers.21.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 280 |
+
"transformer.encoder.layers.21.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 281 |
+
"transformer.encoder.layers.21.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 282 |
+
"transformer.encoder.layers.21.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 283 |
+
"transformer.encoder.layers.21.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 284 |
+
"transformer.encoder.layers.21.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 285 |
+
"transformer.encoder.layers.21.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 286 |
+
"transformer.encoder.layers.21.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 287 |
+
"transformer.encoder.layers.21.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 288 |
+
"transformer.encoder.layers.21.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 289 |
+
"transformer.encoder.layers.21.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 290 |
+
"transformer.encoder.layers.21.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 291 |
+
"transformer.encoder.layers.21.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 292 |
+
"transformer.encoder.layers.21.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 293 |
+
"transformer.encoder.layers.22.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 294 |
+
"transformer.encoder.layers.22.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 295 |
+
"transformer.encoder.layers.22.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 296 |
+
"transformer.encoder.layers.22.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 297 |
+
"transformer.encoder.layers.22.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 298 |
+
"transformer.encoder.layers.22.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 299 |
+
"transformer.encoder.layers.22.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 300 |
+
"transformer.encoder.layers.22.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 301 |
+
"transformer.encoder.layers.22.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 302 |
+
"transformer.encoder.layers.22.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 303 |
+
"transformer.encoder.layers.22.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 304 |
+
"transformer.encoder.layers.22.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 305 |
+
"transformer.encoder.layers.22.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 306 |
+
"transformer.encoder.layers.22.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 307 |
+
"transformer.encoder.layers.22.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 308 |
+
"transformer.encoder.layers.22.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 309 |
+
"transformer.encoder.layers.22.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 310 |
+
"transformer.encoder.layers.22.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 311 |
+
"transformer.encoder.layers.22.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 312 |
+
"transformer.encoder.layers.23.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 313 |
+
"transformer.encoder.layers.23.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 314 |
+
"transformer.encoder.layers.23.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 315 |
+
"transformer.encoder.layers.23.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 316 |
+
"transformer.encoder.layers.23.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 317 |
+
"transformer.encoder.layers.23.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 318 |
+
"transformer.encoder.layers.23.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 319 |
+
"transformer.encoder.layers.23.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 320 |
+
"transformer.encoder.layers.23.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 321 |
+
"transformer.encoder.layers.23.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 322 |
+
"transformer.encoder.layers.23.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 323 |
+
"transformer.encoder.layers.23.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 324 |
+
"transformer.encoder.layers.23.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 325 |
+
"transformer.encoder.layers.23.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 326 |
+
"transformer.encoder.layers.23.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 327 |
+
"transformer.encoder.layers.23.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 328 |
+
"transformer.encoder.layers.23.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 329 |
+
"transformer.encoder.layers.23.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 330 |
+
"transformer.encoder.layers.23.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 331 |
+
"transformer.encoder.layers.24.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 332 |
+
"transformer.encoder.layers.24.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 333 |
+
"transformer.encoder.layers.24.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 334 |
+
"transformer.encoder.layers.24.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 335 |
+
"transformer.encoder.layers.24.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 336 |
+
"transformer.encoder.layers.24.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 337 |
+
"transformer.encoder.layers.24.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 338 |
+
"transformer.encoder.layers.24.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 339 |
+
"transformer.encoder.layers.24.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 340 |
+
"transformer.encoder.layers.24.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 341 |
+
"transformer.encoder.layers.24.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 342 |
+
"transformer.encoder.layers.24.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 343 |
+
"transformer.encoder.layers.24.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 344 |
+
"transformer.encoder.layers.24.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 345 |
+
"transformer.encoder.layers.24.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 346 |
+
"transformer.encoder.layers.24.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 347 |
+
"transformer.encoder.layers.24.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 348 |
+
"transformer.encoder.layers.24.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 349 |
+
"transformer.encoder.layers.24.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 350 |
+
"transformer.encoder.layers.25.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 351 |
+
"transformer.encoder.layers.25.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 352 |
+
"transformer.encoder.layers.25.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 353 |
+
"transformer.encoder.layers.25.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 354 |
+
"transformer.encoder.layers.25.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 355 |
+
"transformer.encoder.layers.25.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 356 |
+
"transformer.encoder.layers.25.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 357 |
+
"transformer.encoder.layers.25.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 358 |
+
"transformer.encoder.layers.25.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 359 |
+
"transformer.encoder.layers.25.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 360 |
+
"transformer.encoder.layers.25.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 361 |
+
"transformer.encoder.layers.25.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 362 |
+
"transformer.encoder.layers.25.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 363 |
+
"transformer.encoder.layers.25.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 364 |
+
"transformer.encoder.layers.25.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 365 |
+
"transformer.encoder.layers.25.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 366 |
+
"transformer.encoder.layers.25.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 367 |
+
"transformer.encoder.layers.25.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 368 |
+
"transformer.encoder.layers.25.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 369 |
+
"transformer.encoder.layers.26.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 370 |
+
"transformer.encoder.layers.26.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 371 |
+
"transformer.encoder.layers.26.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 372 |
+
"transformer.encoder.layers.26.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 373 |
+
"transformer.encoder.layers.26.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 374 |
+
"transformer.encoder.layers.26.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 375 |
+
"transformer.encoder.layers.26.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 376 |
+
"transformer.encoder.layers.26.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 377 |
+
"transformer.encoder.layers.26.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 378 |
+
"transformer.encoder.layers.26.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 379 |
+
"transformer.encoder.layers.26.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 380 |
+
"transformer.encoder.layers.26.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 381 |
+
"transformer.encoder.layers.26.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 382 |
+
"transformer.encoder.layers.26.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 383 |
+
"transformer.encoder.layers.26.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 384 |
+
"transformer.encoder.layers.26.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 385 |
+
"transformer.encoder.layers.26.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 386 |
+
"transformer.encoder.layers.26.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 387 |
+
"transformer.encoder.layers.26.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 388 |
+
"transformer.encoder.layers.27.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 389 |
+
"transformer.encoder.layers.27.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 390 |
+
"transformer.encoder.layers.27.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 391 |
+
"transformer.encoder.layers.27.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 392 |
+
"transformer.encoder.layers.27.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 393 |
+
"transformer.encoder.layers.27.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 394 |
+
"transformer.encoder.layers.27.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 395 |
+
"transformer.encoder.layers.27.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 396 |
+
"transformer.encoder.layers.27.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 397 |
+
"transformer.encoder.layers.27.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 398 |
+
"transformer.encoder.layers.27.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 399 |
+
"transformer.encoder.layers.27.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 400 |
+
"transformer.encoder.layers.27.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 401 |
+
"transformer.encoder.layers.27.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 402 |
+
"transformer.encoder.layers.27.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 403 |
+
"transformer.encoder.layers.27.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 404 |
+
"transformer.encoder.layers.27.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 405 |
+
"transformer.encoder.layers.27.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 406 |
+
"transformer.encoder.layers.27.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 407 |
+
"transformer.encoder.layers.28.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 408 |
+
"transformer.encoder.layers.28.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 409 |
+
"transformer.encoder.layers.28.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 410 |
+
"transformer.encoder.layers.28.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 411 |
+
"transformer.encoder.layers.28.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 412 |
+
"transformer.encoder.layers.28.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 413 |
+
"transformer.encoder.layers.28.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 414 |
+
"transformer.encoder.layers.28.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 415 |
+
"transformer.encoder.layers.28.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 416 |
+
"transformer.encoder.layers.28.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 417 |
+
"transformer.encoder.layers.28.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 418 |
+
"transformer.encoder.layers.28.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 419 |
+
"transformer.encoder.layers.28.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 420 |
+
"transformer.encoder.layers.28.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 421 |
+
"transformer.encoder.layers.28.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 422 |
+
"transformer.encoder.layers.28.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 423 |
+
"transformer.encoder.layers.28.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 424 |
+
"transformer.encoder.layers.28.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 425 |
+
"transformer.encoder.layers.28.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 426 |
+
"transformer.encoder.layers.29.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 427 |
+
"transformer.encoder.layers.29.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 428 |
+
"transformer.encoder.layers.29.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 429 |
+
"transformer.encoder.layers.29.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 430 |
+
"transformer.encoder.layers.29.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 431 |
+
"transformer.encoder.layers.29.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 432 |
+
"transformer.encoder.layers.29.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 433 |
+
"transformer.encoder.layers.29.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 434 |
+
"transformer.encoder.layers.29.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 435 |
+
"transformer.encoder.layers.29.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 436 |
+
"transformer.encoder.layers.29.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 437 |
+
"transformer.encoder.layers.29.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 438 |
+
"transformer.encoder.layers.29.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 439 |
+
"transformer.encoder.layers.29.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 440 |
+
"transformer.encoder.layers.29.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 441 |
+
"transformer.encoder.layers.29.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 442 |
+
"transformer.encoder.layers.29.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 443 |
+
"transformer.encoder.layers.29.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 444 |
+
"transformer.encoder.layers.29.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 445 |
+
"transformer.encoder.layers.3.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 446 |
+
"transformer.encoder.layers.3.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 447 |
+
"transformer.encoder.layers.3.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 448 |
+
"transformer.encoder.layers.3.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 449 |
+
"transformer.encoder.layers.3.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 450 |
+
"transformer.encoder.layers.3.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 451 |
+
"transformer.encoder.layers.3.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 452 |
+
"transformer.encoder.layers.3.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 453 |
+
"transformer.encoder.layers.3.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 454 |
+
"transformer.encoder.layers.3.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 455 |
+
"transformer.encoder.layers.3.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 456 |
+
"transformer.encoder.layers.3.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 457 |
+
"transformer.encoder.layers.3.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 458 |
+
"transformer.encoder.layers.3.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 459 |
+
"transformer.encoder.layers.3.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 460 |
+
"transformer.encoder.layers.3.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 461 |
+
"transformer.encoder.layers.3.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 462 |
+
"transformer.encoder.layers.3.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 463 |
+
"transformer.encoder.layers.3.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 464 |
+
"transformer.encoder.layers.30.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 465 |
+
"transformer.encoder.layers.30.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 466 |
+
"transformer.encoder.layers.30.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 467 |
+
"transformer.encoder.layers.30.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 468 |
+
"transformer.encoder.layers.30.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 469 |
+
"transformer.encoder.layers.30.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 470 |
+
"transformer.encoder.layers.30.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 471 |
+
"transformer.encoder.layers.30.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 472 |
+
"transformer.encoder.layers.30.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 473 |
+
"transformer.encoder.layers.30.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 474 |
+
"transformer.encoder.layers.30.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 475 |
+
"transformer.encoder.layers.30.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 476 |
+
"transformer.encoder.layers.30.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 477 |
+
"transformer.encoder.layers.30.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 478 |
+
"transformer.encoder.layers.30.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 479 |
+
"transformer.encoder.layers.30.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 480 |
+
"transformer.encoder.layers.30.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 481 |
+
"transformer.encoder.layers.30.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 482 |
+
"transformer.encoder.layers.30.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 483 |
+
"transformer.encoder.layers.31.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 484 |
+
"transformer.encoder.layers.31.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 485 |
+
"transformer.encoder.layers.31.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 486 |
+
"transformer.encoder.layers.31.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 487 |
+
"transformer.encoder.layers.31.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 488 |
+
"transformer.encoder.layers.31.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 489 |
+
"transformer.encoder.layers.31.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 490 |
+
"transformer.encoder.layers.31.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 491 |
+
"transformer.encoder.layers.31.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 492 |
+
"transformer.encoder.layers.31.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 493 |
+
"transformer.encoder.layers.31.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 494 |
+
"transformer.encoder.layers.31.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 495 |
+
"transformer.encoder.layers.31.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 496 |
+
"transformer.encoder.layers.31.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 497 |
+
"transformer.encoder.layers.31.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 498 |
+
"transformer.encoder.layers.31.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 499 |
+
"transformer.encoder.layers.31.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 500 |
+
"transformer.encoder.layers.31.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 501 |
+
"transformer.encoder.layers.31.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 502 |
+
"transformer.encoder.layers.32.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 503 |
+
"transformer.encoder.layers.32.mlp.dense_4h_to_h.weight": "model-00002-of-00002.safetensors",
|
| 504 |
+
"transformer.encoder.layers.32.mlp.dense_4h_to_h.weight.absmax": "model-00002-of-00002.safetensors",
|
| 505 |
+
"transformer.encoder.layers.32.mlp.dense_4h_to_h.weight.quant_map": "model-00002-of-00002.safetensors",
|
| 506 |
+
"transformer.encoder.layers.32.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00002-of-00002.safetensors",
|
| 507 |
+
"transformer.encoder.layers.32.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 508 |
+
"transformer.encoder.layers.32.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 509 |
+
"transformer.encoder.layers.32.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 510 |
+
"transformer.encoder.layers.32.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 511 |
+
"transformer.encoder.layers.32.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 512 |
+
"transformer.encoder.layers.32.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 513 |
+
"transformer.encoder.layers.32.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 514 |
+
"transformer.encoder.layers.32.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 515 |
+
"transformer.encoder.layers.32.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 516 |
+
"transformer.encoder.layers.32.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 517 |
+
"transformer.encoder.layers.32.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 518 |
+
"transformer.encoder.layers.32.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 519 |
+
"transformer.encoder.layers.32.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 520 |
+
"transformer.encoder.layers.32.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 521 |
+
"transformer.encoder.layers.33.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 522 |
+
"transformer.encoder.layers.33.mlp.dense_4h_to_h.weight": "model-00002-of-00002.safetensors",
|
| 523 |
+
"transformer.encoder.layers.33.mlp.dense_4h_to_h.weight.absmax": "model-00002-of-00002.safetensors",
|
| 524 |
+
"transformer.encoder.layers.33.mlp.dense_4h_to_h.weight.quant_map": "model-00002-of-00002.safetensors",
|
| 525 |
+
"transformer.encoder.layers.33.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00002-of-00002.safetensors",
|
| 526 |
+
"transformer.encoder.layers.33.mlp.dense_h_to_4h.weight": "model-00002-of-00002.safetensors",
|
| 527 |
+
"transformer.encoder.layers.33.mlp.dense_h_to_4h.weight.absmax": "model-00002-of-00002.safetensors",
|
| 528 |
+
"transformer.encoder.layers.33.mlp.dense_h_to_4h.weight.quant_map": "model-00002-of-00002.safetensors",
|
| 529 |
+
"transformer.encoder.layers.33.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00002-of-00002.safetensors",
|
| 530 |
+
"transformer.encoder.layers.33.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 531 |
+
"transformer.encoder.layers.33.self_attention.dense.weight": "model-00002-of-00002.safetensors",
|
| 532 |
+
"transformer.encoder.layers.33.self_attention.dense.weight.absmax": "model-00002-of-00002.safetensors",
|
| 533 |
+
"transformer.encoder.layers.33.self_attention.dense.weight.quant_map": "model-00002-of-00002.safetensors",
|
| 534 |
+
"transformer.encoder.layers.33.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00002-of-00002.safetensors",
|
| 535 |
+
"transformer.encoder.layers.33.self_attention.query_key_value.bias": "model-00002-of-00002.safetensors",
|
| 536 |
+
"transformer.encoder.layers.33.self_attention.query_key_value.weight": "model-00002-of-00002.safetensors",
|
| 537 |
+
"transformer.encoder.layers.33.self_attention.query_key_value.weight.absmax": "model-00002-of-00002.safetensors",
|
| 538 |
+
"transformer.encoder.layers.33.self_attention.query_key_value.weight.quant_map": "model-00002-of-00002.safetensors",
|
| 539 |
+
"transformer.encoder.layers.33.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00002-of-00002.safetensors",
|
| 540 |
+
"transformer.encoder.layers.34.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 541 |
+
"transformer.encoder.layers.34.mlp.dense_4h_to_h.weight": "model-00002-of-00002.safetensors",
|
| 542 |
+
"transformer.encoder.layers.34.mlp.dense_4h_to_h.weight.absmax": "model-00002-of-00002.safetensors",
|
| 543 |
+
"transformer.encoder.layers.34.mlp.dense_4h_to_h.weight.quant_map": "model-00002-of-00002.safetensors",
|
| 544 |
+
"transformer.encoder.layers.34.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00002-of-00002.safetensors",
|
| 545 |
+
"transformer.encoder.layers.34.mlp.dense_h_to_4h.weight": "model-00002-of-00002.safetensors",
|
| 546 |
+
"transformer.encoder.layers.34.mlp.dense_h_to_4h.weight.absmax": "model-00002-of-00002.safetensors",
|
| 547 |
+
"transformer.encoder.layers.34.mlp.dense_h_to_4h.weight.quant_map": "model-00002-of-00002.safetensors",
|
| 548 |
+
"transformer.encoder.layers.34.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00002-of-00002.safetensors",
|
| 549 |
+
"transformer.encoder.layers.34.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 550 |
+
"transformer.encoder.layers.34.self_attention.dense.weight": "model-00002-of-00002.safetensors",
|
| 551 |
+
"transformer.encoder.layers.34.self_attention.dense.weight.absmax": "model-00002-of-00002.safetensors",
|
| 552 |
+
"transformer.encoder.layers.34.self_attention.dense.weight.quant_map": "model-00002-of-00002.safetensors",
|
| 553 |
+
"transformer.encoder.layers.34.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00002-of-00002.safetensors",
|
| 554 |
+
"transformer.encoder.layers.34.self_attention.query_key_value.bias": "model-00002-of-00002.safetensors",
|
| 555 |
+
"transformer.encoder.layers.34.self_attention.query_key_value.weight": "model-00002-of-00002.safetensors",
|
| 556 |
+
"transformer.encoder.layers.34.self_attention.query_key_value.weight.absmax": "model-00002-of-00002.safetensors",
|
| 557 |
+
"transformer.encoder.layers.34.self_attention.query_key_value.weight.quant_map": "model-00002-of-00002.safetensors",
|
| 558 |
+
"transformer.encoder.layers.34.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00002-of-00002.safetensors",
|
| 559 |
+
"transformer.encoder.layers.35.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 560 |
+
"transformer.encoder.layers.35.mlp.dense_4h_to_h.weight": "model-00002-of-00002.safetensors",
|
| 561 |
+
"transformer.encoder.layers.35.mlp.dense_4h_to_h.weight.absmax": "model-00002-of-00002.safetensors",
|
| 562 |
+
"transformer.encoder.layers.35.mlp.dense_4h_to_h.weight.quant_map": "model-00002-of-00002.safetensors",
|
| 563 |
+
"transformer.encoder.layers.35.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00002-of-00002.safetensors",
|
| 564 |
+
"transformer.encoder.layers.35.mlp.dense_h_to_4h.weight": "model-00002-of-00002.safetensors",
|
| 565 |
+
"transformer.encoder.layers.35.mlp.dense_h_to_4h.weight.absmax": "model-00002-of-00002.safetensors",
|
| 566 |
+
"transformer.encoder.layers.35.mlp.dense_h_to_4h.weight.quant_map": "model-00002-of-00002.safetensors",
|
| 567 |
+
"transformer.encoder.layers.35.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00002-of-00002.safetensors",
|
| 568 |
+
"transformer.encoder.layers.35.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 569 |
+
"transformer.encoder.layers.35.self_attention.dense.weight": "model-00002-of-00002.safetensors",
|
| 570 |
+
"transformer.encoder.layers.35.self_attention.dense.weight.absmax": "model-00002-of-00002.safetensors",
|
| 571 |
+
"transformer.encoder.layers.35.self_attention.dense.weight.quant_map": "model-00002-of-00002.safetensors",
|
| 572 |
+
"transformer.encoder.layers.35.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00002-of-00002.safetensors",
|
| 573 |
+
"transformer.encoder.layers.35.self_attention.query_key_value.bias": "model-00002-of-00002.safetensors",
|
| 574 |
+
"transformer.encoder.layers.35.self_attention.query_key_value.weight": "model-00002-of-00002.safetensors",
|
| 575 |
+
"transformer.encoder.layers.35.self_attention.query_key_value.weight.absmax": "model-00002-of-00002.safetensors",
|
| 576 |
+
"transformer.encoder.layers.35.self_attention.query_key_value.weight.quant_map": "model-00002-of-00002.safetensors",
|
| 577 |
+
"transformer.encoder.layers.35.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00002-of-00002.safetensors",
|
| 578 |
+
"transformer.encoder.layers.36.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 579 |
+
"transformer.encoder.layers.36.mlp.dense_4h_to_h.weight": "model-00002-of-00002.safetensors",
|
| 580 |
+
"transformer.encoder.layers.36.mlp.dense_4h_to_h.weight.absmax": "model-00002-of-00002.safetensors",
|
| 581 |
+
"transformer.encoder.layers.36.mlp.dense_4h_to_h.weight.quant_map": "model-00002-of-00002.safetensors",
|
| 582 |
+
"transformer.encoder.layers.36.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00002-of-00002.safetensors",
|
| 583 |
+
"transformer.encoder.layers.36.mlp.dense_h_to_4h.weight": "model-00002-of-00002.safetensors",
|
| 584 |
+
"transformer.encoder.layers.36.mlp.dense_h_to_4h.weight.absmax": "model-00002-of-00002.safetensors",
|
| 585 |
+
"transformer.encoder.layers.36.mlp.dense_h_to_4h.weight.quant_map": "model-00002-of-00002.safetensors",
|
| 586 |
+
"transformer.encoder.layers.36.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00002-of-00002.safetensors",
|
| 587 |
+
"transformer.encoder.layers.36.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 588 |
+
"transformer.encoder.layers.36.self_attention.dense.weight": "model-00002-of-00002.safetensors",
|
| 589 |
+
"transformer.encoder.layers.36.self_attention.dense.weight.absmax": "model-00002-of-00002.safetensors",
|
| 590 |
+
"transformer.encoder.layers.36.self_attention.dense.weight.quant_map": "model-00002-of-00002.safetensors",
|
| 591 |
+
"transformer.encoder.layers.36.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00002-of-00002.safetensors",
|
| 592 |
+
"transformer.encoder.layers.36.self_attention.query_key_value.bias": "model-00002-of-00002.safetensors",
|
| 593 |
+
"transformer.encoder.layers.36.self_attention.query_key_value.weight": "model-00002-of-00002.safetensors",
|
| 594 |
+
"transformer.encoder.layers.36.self_attention.query_key_value.weight.absmax": "model-00002-of-00002.safetensors",
|
| 595 |
+
"transformer.encoder.layers.36.self_attention.query_key_value.weight.quant_map": "model-00002-of-00002.safetensors",
|
| 596 |
+
"transformer.encoder.layers.36.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00002-of-00002.safetensors",
|
| 597 |
+
"transformer.encoder.layers.37.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 598 |
+
"transformer.encoder.layers.37.mlp.dense_4h_to_h.weight": "model-00002-of-00002.safetensors",
|
| 599 |
+
"transformer.encoder.layers.37.mlp.dense_4h_to_h.weight.absmax": "model-00002-of-00002.safetensors",
|
| 600 |
+
"transformer.encoder.layers.37.mlp.dense_4h_to_h.weight.quant_map": "model-00002-of-00002.safetensors",
|
| 601 |
+
"transformer.encoder.layers.37.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00002-of-00002.safetensors",
|
| 602 |
+
"transformer.encoder.layers.37.mlp.dense_h_to_4h.weight": "model-00002-of-00002.safetensors",
|
| 603 |
+
"transformer.encoder.layers.37.mlp.dense_h_to_4h.weight.absmax": "model-00002-of-00002.safetensors",
|
| 604 |
+
"transformer.encoder.layers.37.mlp.dense_h_to_4h.weight.quant_map": "model-00002-of-00002.safetensors",
|
| 605 |
+
"transformer.encoder.layers.37.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00002-of-00002.safetensors",
|
| 606 |
+
"transformer.encoder.layers.37.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 607 |
+
"transformer.encoder.layers.37.self_attention.dense.weight": "model-00002-of-00002.safetensors",
|
| 608 |
+
"transformer.encoder.layers.37.self_attention.dense.weight.absmax": "model-00002-of-00002.safetensors",
|
| 609 |
+
"transformer.encoder.layers.37.self_attention.dense.weight.quant_map": "model-00002-of-00002.safetensors",
|
| 610 |
+
"transformer.encoder.layers.37.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00002-of-00002.safetensors",
|
| 611 |
+
"transformer.encoder.layers.37.self_attention.query_key_value.bias": "model-00002-of-00002.safetensors",
|
| 612 |
+
"transformer.encoder.layers.37.self_attention.query_key_value.weight": "model-00002-of-00002.safetensors",
|
| 613 |
+
"transformer.encoder.layers.37.self_attention.query_key_value.weight.absmax": "model-00002-of-00002.safetensors",
|
| 614 |
+
"transformer.encoder.layers.37.self_attention.query_key_value.weight.quant_map": "model-00002-of-00002.safetensors",
|
| 615 |
+
"transformer.encoder.layers.37.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00002-of-00002.safetensors",
|
| 616 |
+
"transformer.encoder.layers.38.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 617 |
+
"transformer.encoder.layers.38.mlp.dense_4h_to_h.weight": "model-00002-of-00002.safetensors",
|
| 618 |
+
"transformer.encoder.layers.38.mlp.dense_4h_to_h.weight.absmax": "model-00002-of-00002.safetensors",
|
| 619 |
+
"transformer.encoder.layers.38.mlp.dense_4h_to_h.weight.quant_map": "model-00002-of-00002.safetensors",
|
| 620 |
+
"transformer.encoder.layers.38.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00002-of-00002.safetensors",
|
| 621 |
+
"transformer.encoder.layers.38.mlp.dense_h_to_4h.weight": "model-00002-of-00002.safetensors",
|
| 622 |
+
"transformer.encoder.layers.38.mlp.dense_h_to_4h.weight.absmax": "model-00002-of-00002.safetensors",
|
| 623 |
+
"transformer.encoder.layers.38.mlp.dense_h_to_4h.weight.quant_map": "model-00002-of-00002.safetensors",
|
| 624 |
+
"transformer.encoder.layers.38.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00002-of-00002.safetensors",
|
| 625 |
+
"transformer.encoder.layers.38.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 626 |
+
"transformer.encoder.layers.38.self_attention.dense.weight": "model-00002-of-00002.safetensors",
|
| 627 |
+
"transformer.encoder.layers.38.self_attention.dense.weight.absmax": "model-00002-of-00002.safetensors",
|
| 628 |
+
"transformer.encoder.layers.38.self_attention.dense.weight.quant_map": "model-00002-of-00002.safetensors",
|
| 629 |
+
"transformer.encoder.layers.38.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00002-of-00002.safetensors",
|
| 630 |
+
"transformer.encoder.layers.38.self_attention.query_key_value.bias": "model-00002-of-00002.safetensors",
|
| 631 |
+
"transformer.encoder.layers.38.self_attention.query_key_value.weight": "model-00002-of-00002.safetensors",
|
| 632 |
+
"transformer.encoder.layers.38.self_attention.query_key_value.weight.absmax": "model-00002-of-00002.safetensors",
|
| 633 |
+
"transformer.encoder.layers.38.self_attention.query_key_value.weight.quant_map": "model-00002-of-00002.safetensors",
|
| 634 |
+
"transformer.encoder.layers.38.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00002-of-00002.safetensors",
|
| 635 |
+
"transformer.encoder.layers.39.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 636 |
+
"transformer.encoder.layers.39.mlp.dense_4h_to_h.weight": "model-00002-of-00002.safetensors",
|
| 637 |
+
"transformer.encoder.layers.39.mlp.dense_4h_to_h.weight.absmax": "model-00002-of-00002.safetensors",
|
| 638 |
+
"transformer.encoder.layers.39.mlp.dense_4h_to_h.weight.quant_map": "model-00002-of-00002.safetensors",
|
| 639 |
+
"transformer.encoder.layers.39.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00002-of-00002.safetensors",
|
| 640 |
+
"transformer.encoder.layers.39.mlp.dense_h_to_4h.weight": "model-00002-of-00002.safetensors",
|
| 641 |
+
"transformer.encoder.layers.39.mlp.dense_h_to_4h.weight.absmax": "model-00002-of-00002.safetensors",
|
| 642 |
+
"transformer.encoder.layers.39.mlp.dense_h_to_4h.weight.quant_map": "model-00002-of-00002.safetensors",
|
| 643 |
+
"transformer.encoder.layers.39.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00002-of-00002.safetensors",
|
| 644 |
+
"transformer.encoder.layers.39.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 645 |
+
"transformer.encoder.layers.39.self_attention.dense.weight": "model-00002-of-00002.safetensors",
|
| 646 |
+
"transformer.encoder.layers.39.self_attention.dense.weight.absmax": "model-00002-of-00002.safetensors",
|
| 647 |
+
"transformer.encoder.layers.39.self_attention.dense.weight.quant_map": "model-00002-of-00002.safetensors",
|
| 648 |
+
"transformer.encoder.layers.39.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00002-of-00002.safetensors",
|
| 649 |
+
"transformer.encoder.layers.39.self_attention.query_key_value.bias": "model-00002-of-00002.safetensors",
|
| 650 |
+
"transformer.encoder.layers.39.self_attention.query_key_value.weight": "model-00002-of-00002.safetensors",
|
| 651 |
+
"transformer.encoder.layers.39.self_attention.query_key_value.weight.absmax": "model-00002-of-00002.safetensors",
|
| 652 |
+
"transformer.encoder.layers.39.self_attention.query_key_value.weight.quant_map": "model-00002-of-00002.safetensors",
|
| 653 |
+
"transformer.encoder.layers.39.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00002-of-00002.safetensors",
|
| 654 |
+
"transformer.encoder.layers.4.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 655 |
+
"transformer.encoder.layers.4.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 656 |
+
"transformer.encoder.layers.4.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 657 |
+
"transformer.encoder.layers.4.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 658 |
+
"transformer.encoder.layers.4.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 659 |
+
"transformer.encoder.layers.4.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 660 |
+
"transformer.encoder.layers.4.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 661 |
+
"transformer.encoder.layers.4.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 662 |
+
"transformer.encoder.layers.4.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 663 |
+
"transformer.encoder.layers.4.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 664 |
+
"transformer.encoder.layers.4.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 665 |
+
"transformer.encoder.layers.4.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 666 |
+
"transformer.encoder.layers.4.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 667 |
+
"transformer.encoder.layers.4.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 668 |
+
"transformer.encoder.layers.4.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 669 |
+
"transformer.encoder.layers.4.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 670 |
+
"transformer.encoder.layers.4.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 671 |
+
"transformer.encoder.layers.4.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 672 |
+
"transformer.encoder.layers.4.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 673 |
+
"transformer.encoder.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 674 |
+
"transformer.encoder.layers.5.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 675 |
+
"transformer.encoder.layers.5.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 676 |
+
"transformer.encoder.layers.5.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 677 |
+
"transformer.encoder.layers.5.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 678 |
+
"transformer.encoder.layers.5.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 679 |
+
"transformer.encoder.layers.5.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 680 |
+
"transformer.encoder.layers.5.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 681 |
+
"transformer.encoder.layers.5.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 682 |
+
"transformer.encoder.layers.5.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 683 |
+
"transformer.encoder.layers.5.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 684 |
+
"transformer.encoder.layers.5.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 685 |
+
"transformer.encoder.layers.5.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 686 |
+
"transformer.encoder.layers.5.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 687 |
+
"transformer.encoder.layers.5.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 688 |
+
"transformer.encoder.layers.5.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 689 |
+
"transformer.encoder.layers.5.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 690 |
+
"transformer.encoder.layers.5.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 691 |
+
"transformer.encoder.layers.5.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 692 |
+
"transformer.encoder.layers.6.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 693 |
+
"transformer.encoder.layers.6.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 694 |
+
"transformer.encoder.layers.6.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 695 |
+
"transformer.encoder.layers.6.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 696 |
+
"transformer.encoder.layers.6.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 697 |
+
"transformer.encoder.layers.6.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 698 |
+
"transformer.encoder.layers.6.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 699 |
+
"transformer.encoder.layers.6.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 700 |
+
"transformer.encoder.layers.6.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 701 |
+
"transformer.encoder.layers.6.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 702 |
+
"transformer.encoder.layers.6.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 703 |
+
"transformer.encoder.layers.6.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 704 |
+
"transformer.encoder.layers.6.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 705 |
+
"transformer.encoder.layers.6.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 706 |
+
"transformer.encoder.layers.6.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 707 |
+
"transformer.encoder.layers.6.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 708 |
+
"transformer.encoder.layers.6.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 709 |
+
"transformer.encoder.layers.6.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 710 |
+
"transformer.encoder.layers.6.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 711 |
+
"transformer.encoder.layers.7.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 712 |
+
"transformer.encoder.layers.7.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 713 |
+
"transformer.encoder.layers.7.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 714 |
+
"transformer.encoder.layers.7.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 715 |
+
"transformer.encoder.layers.7.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 716 |
+
"transformer.encoder.layers.7.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 717 |
+
"transformer.encoder.layers.7.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 718 |
+
"transformer.encoder.layers.7.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 719 |
+
"transformer.encoder.layers.7.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 720 |
+
"transformer.encoder.layers.7.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 721 |
+
"transformer.encoder.layers.7.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 722 |
+
"transformer.encoder.layers.7.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 723 |
+
"transformer.encoder.layers.7.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 724 |
+
"transformer.encoder.layers.7.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 725 |
+
"transformer.encoder.layers.7.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 726 |
+
"transformer.encoder.layers.7.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 727 |
+
"transformer.encoder.layers.7.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 728 |
+
"transformer.encoder.layers.7.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 729 |
+
"transformer.encoder.layers.7.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 730 |
+
"transformer.encoder.layers.8.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 731 |
+
"transformer.encoder.layers.8.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 732 |
+
"transformer.encoder.layers.8.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 733 |
+
"transformer.encoder.layers.8.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 734 |
+
"transformer.encoder.layers.8.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 735 |
+
"transformer.encoder.layers.8.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 736 |
+
"transformer.encoder.layers.8.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 737 |
+
"transformer.encoder.layers.8.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 738 |
+
"transformer.encoder.layers.8.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 739 |
+
"transformer.encoder.layers.8.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 740 |
+
"transformer.encoder.layers.8.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 741 |
+
"transformer.encoder.layers.8.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 742 |
+
"transformer.encoder.layers.8.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 743 |
+
"transformer.encoder.layers.8.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 744 |
+
"transformer.encoder.layers.8.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 745 |
+
"transformer.encoder.layers.8.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 746 |
+
"transformer.encoder.layers.8.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 747 |
+
"transformer.encoder.layers.8.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 748 |
+
"transformer.encoder.layers.8.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 749 |
+
"transformer.encoder.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 750 |
+
"transformer.encoder.layers.9.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
|
| 751 |
+
"transformer.encoder.layers.9.mlp.dense_4h_to_h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 752 |
+
"transformer.encoder.layers.9.mlp.dense_4h_to_h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 753 |
+
"transformer.encoder.layers.9.mlp.dense_4h_to_h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 754 |
+
"transformer.encoder.layers.9.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
|
| 755 |
+
"transformer.encoder.layers.9.mlp.dense_h_to_4h.weight.absmax": "model-00001-of-00002.safetensors",
|
| 756 |
+
"transformer.encoder.layers.9.mlp.dense_h_to_4h.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 757 |
+
"transformer.encoder.layers.9.mlp.dense_h_to_4h.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 758 |
+
"transformer.encoder.layers.9.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 759 |
+
"transformer.encoder.layers.9.self_attention.dense.weight": "model-00001-of-00002.safetensors",
|
| 760 |
+
"transformer.encoder.layers.9.self_attention.dense.weight.absmax": "model-00001-of-00002.safetensors",
|
| 761 |
+
"transformer.encoder.layers.9.self_attention.dense.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 762 |
+
"transformer.encoder.layers.9.self_attention.dense.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 763 |
+
"transformer.encoder.layers.9.self_attention.query_key_value.bias": "model-00001-of-00002.safetensors",
|
| 764 |
+
"transformer.encoder.layers.9.self_attention.query_key_value.weight": "model-00001-of-00002.safetensors",
|
| 765 |
+
"transformer.encoder.layers.9.self_attention.query_key_value.weight.absmax": "model-00001-of-00002.safetensors",
|
| 766 |
+
"transformer.encoder.layers.9.self_attention.query_key_value.weight.quant_map": "model-00001-of-00002.safetensors",
|
| 767 |
+
"transformer.encoder.layers.9.self_attention.query_key_value.weight.quant_state.bitsandbytes__fp4": "model-00001-of-00002.safetensors",
|
| 768 |
+
"transformer.output_layer.weight": "model-00002-of-00002.safetensors",
|
| 769 |
+
"transformer.output_layer.weight.absmax": "model-00002-of-00002.safetensors",
|
| 770 |
+
"transformer.output_layer.weight.quant_map": "model-00002-of-00002.safetensors",
|
| 771 |
+
"transformer.output_layer.weight.quant_state.bitsandbytes__fp4": "model-00002-of-00002.safetensors",
|
| 772 |
+
"transformer.rotary_pos_emb.inv_freq": "model-00001-of-00002.safetensors"
|
| 773 |
+
}
|
| 774 |
+
}
|
modeling_chatglm.py
ADDED
|
@@ -0,0 +1,898 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
""" PyTorch ChatGLM model. """
|
| 2 |
+
|
| 3 |
+
import math
|
| 4 |
+
import copy
|
| 5 |
+
import warnings
|
| 6 |
+
import re
|
| 7 |
+
import sys
|
| 8 |
+
|
| 9 |
+
import torch
|
| 10 |
+
import torch.utils.checkpoint
|
| 11 |
+
import torch.nn.functional as F
|
| 12 |
+
from torch import nn
|
| 13 |
+
from torch.nn import CrossEntropyLoss, LayerNorm
|
| 14 |
+
from torch.nn.utils import skip_init
|
| 15 |
+
from typing import Optional, Tuple, Union, List, Callable, Dict, Any
|
| 16 |
+
|
| 17 |
+
from transformers.modeling_outputs import (
|
| 18 |
+
BaseModelOutputWithPast,
|
| 19 |
+
CausalLMOutputWithPast,
|
| 20 |
+
)
|
| 21 |
+
from transformers.modeling_utils import PreTrainedModel
|
| 22 |
+
from transformers.utils import logging
|
| 23 |
+
from transformers.generation.logits_process import LogitsProcessor
|
| 24 |
+
from transformers.generation.utils import LogitsProcessorList, StoppingCriteriaList, GenerationConfig, ModelOutput
|
| 25 |
+
|
| 26 |
+
from .configuration_chatglm import ChatGLMConfig
|
| 27 |
+
from einops import rearrange
|
| 28 |
+
try:
|
| 29 |
+
from flash_attn.flash_attn_interface import flash_attn_unpadded_func
|
| 30 |
+
except ImportError:
|
| 31 |
+
try:
|
| 32 |
+
# FlashAttention-2
|
| 33 |
+
from flash_attn.flash_attn_interface import flash_attn_varlen_func as flash_attn_unpadded_func
|
| 34 |
+
except ImportError:
|
| 35 |
+
flash_attn_unpadded_func = None
|
| 36 |
+
|
| 37 |
+
# flags required to enable jit fusion kernels
|
| 38 |
+
|
| 39 |
+
if sys.platform != 'darwin':
|
| 40 |
+
torch._C._jit_set_profiling_mode(False)
|
| 41 |
+
torch._C._jit_set_profiling_executor(False)
|
| 42 |
+
torch._C._jit_override_can_fuse_on_cpu(True)
|
| 43 |
+
torch._C._jit_override_can_fuse_on_gpu(True)
|
| 44 |
+
|
| 45 |
+
logger = logging.get_logger(__name__)
|
| 46 |
+
|
| 47 |
+
_CHECKPOINT_FOR_DOC = "THUDM/ChatGLM2-6B"
|
| 48 |
+
_CONFIG_FOR_DOC = "ChatGLM6BConfig"
|
| 49 |
+
|
| 50 |
+
CHATGLM_6B_PRETRAINED_MODEL_ARCHIVE_LIST = [
|
| 51 |
+
"THUDM/chatglm2-6b",
|
| 52 |
+
# See all ChatGLM models at https://huggingface.co/models?filter=chatglm
|
| 53 |
+
]
|
| 54 |
+
|
| 55 |
+
def default_init(cls, *args, **kwargs):
|
| 56 |
+
return cls(*args, **kwargs)
|
| 57 |
+
|
| 58 |
+
|
| 59 |
+
class InvalidScoreLogitsProcessor(LogitsProcessor):
|
| 60 |
+
def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor) -> torch.FloatTensor:
|
| 61 |
+
if torch.isnan(scores).any() or torch.isinf(scores).any():
|
| 62 |
+
scores.zero_()
|
| 63 |
+
scores[..., 5] = 5e4
|
| 64 |
+
return scores
|
| 65 |
+
|
| 66 |
+
def split_tensor_along_last_dim(
|
| 67 |
+
tensor: torch.Tensor,
|
| 68 |
+
num_partitions: int,
|
| 69 |
+
contiguous_split_chunks: bool = False,
|
| 70 |
+
) -> List[torch.Tensor]:
|
| 71 |
+
"""Split a tensor along its last dimension.
|
| 72 |
+
|
| 73 |
+
Arguments:
|
| 74 |
+
tensor: input tensor.
|
| 75 |
+
num_partitions: number of partitions to split the tensor
|
| 76 |
+
contiguous_split_chunks: If True, make each chunk contiguous
|
| 77 |
+
in memory.
|
| 78 |
+
|
| 79 |
+
Returns:
|
| 80 |
+
A list of Tensors
|
| 81 |
+
"""
|
| 82 |
+
# Get the size and dimension.
|
| 83 |
+
last_dim = tensor.dim() - 1
|
| 84 |
+
last_dim_size = tensor.size()[last_dim] // num_partitions
|
| 85 |
+
# Split.
|
| 86 |
+
tensor_list = torch.split(tensor, last_dim_size, dim=last_dim)
|
| 87 |
+
# Note: torch.split does not create contiguous tensors by default.
|
| 88 |
+
if contiguous_split_chunks:
|
| 89 |
+
return tuple(chunk.contiguous() for chunk in tensor_list)
|
| 90 |
+
|
| 91 |
+
return tensor_list
|
| 92 |
+
|
| 93 |
+
|
| 94 |
+
class RotaryEmbedding(nn.Module):
|
| 95 |
+
def __init__(self, dim, rope_ratio=1, original_impl=False, device=None, dtype=None):
|
| 96 |
+
super().__init__()
|
| 97 |
+
inv_freq = 1.0 / (10000 ** (torch.arange(0, dim, 2, device=device).to(dtype=dtype) / dim))
|
| 98 |
+
self.register_buffer("inv_freq", inv_freq)
|
| 99 |
+
self.dim = dim
|
| 100 |
+
self.original_impl = original_impl
|
| 101 |
+
self.rope_ratio = rope_ratio
|
| 102 |
+
|
| 103 |
+
def forward_impl(
|
| 104 |
+
self, seq_len: int, n_elem: int, dtype: torch.dtype, device: torch.device, base: int = 10000
|
| 105 |
+
):
|
| 106 |
+
"""Enhanced Transformer with Rotary Position Embedding.
|
| 107 |
+
|
| 108 |
+
Derived from: https://github.com/labmlai/annotated_deep_learning_paper_implementations/blob/master/labml_nn/
|
| 109 |
+
transformers/rope/__init__.py. MIT License:
|
| 110 |
+
https://github.com/labmlai/annotated_deep_learning_paper_implementations/blob/master/license.
|
| 111 |
+
"""
|
| 112 |
+
# $\Theta = {\theta_i = 10000^{\frac{2(i-1)}{d}}, i \in [1, 2, ..., \frac{d}{2}]}$
|
| 113 |
+
|
| 114 |
+
base = base * self.rope_ratio
|
| 115 |
+
theta = 1.0 / (base ** (torch.arange(0, n_elem, 2, dtype=torch.float, device=device) / n_elem))
|
| 116 |
+
|
| 117 |
+
# Create position indexes `[0, 1, ..., seq_len - 1]`
|
| 118 |
+
seq_idx = torch.arange(seq_len, dtype=torch.float, device=device)
|
| 119 |
+
|
| 120 |
+
# Calculate the product of position index and $\theta_i$
|
| 121 |
+
idx_theta = torch.outer(seq_idx, theta).float()
|
| 122 |
+
|
| 123 |
+
cache = torch.stack([torch.cos(idx_theta), torch.sin(idx_theta)], dim=-1)
|
| 124 |
+
|
| 125 |
+
# this is to mimic the behaviour of complex32, else we will get different results
|
| 126 |
+
if dtype in (torch.float16, torch.bfloat16, torch.int8):
|
| 127 |
+
cache = cache.bfloat16() if dtype == torch.bfloat16 else cache.half()
|
| 128 |
+
return cache
|
| 129 |
+
|
| 130 |
+
def forward(self, max_seq_len, offset=0):
|
| 131 |
+
return self.forward_impl(
|
| 132 |
+
max_seq_len, self.dim, dtype=self.inv_freq.dtype, device=self.inv_freq.device
|
| 133 |
+
)
|
| 134 |
+
|
| 135 |
+
|
| 136 |
+
@torch.jit.script
|
| 137 |
+
def apply_rotary_pos_emb(x: torch.Tensor, rope_cache: torch.Tensor) -> torch.Tensor:
|
| 138 |
+
# x: [sq, b, np, hn]
|
| 139 |
+
sq, b, np, hn = x.size(0), x.size(1), x.size(2), x.size(3)
|
| 140 |
+
rot_dim = rope_cache.shape[-2] * 2
|
| 141 |
+
x, x_pass = x[..., :rot_dim], x[..., rot_dim:]
|
| 142 |
+
# truncate to support variable sizes
|
| 143 |
+
rope_cache = rope_cache[:sq]
|
| 144 |
+
xshaped = x.reshape(sq, -1, np, rot_dim // 2, 2)
|
| 145 |
+
rope_cache = rope_cache.view(sq, -1, 1, xshaped.size(3), 2)
|
| 146 |
+
x_out2 = torch.stack(
|
| 147 |
+
[
|
| 148 |
+
xshaped[..., 0] * rope_cache[..., 0] - xshaped[..., 1] * rope_cache[..., 1],
|
| 149 |
+
xshaped[..., 1] * rope_cache[..., 0] + xshaped[..., 0] * rope_cache[..., 1],
|
| 150 |
+
],
|
| 151 |
+
-1,
|
| 152 |
+
)
|
| 153 |
+
x_out2 = x_out2.flatten(3)
|
| 154 |
+
return torch.cat((x_out2, x_pass), dim=-1)
|
| 155 |
+
|
| 156 |
+
|
| 157 |
+
class RMSNorm(torch.nn.Module):
|
| 158 |
+
def __init__(self, normalized_shape, eps=1e-5, device=None, dtype=None, **kwargs):
|
| 159 |
+
super().__init__()
|
| 160 |
+
self.weight = torch.nn.Parameter(torch.empty(normalized_shape, device=device, dtype=dtype))
|
| 161 |
+
self.eps = eps
|
| 162 |
+
|
| 163 |
+
def forward(self, hidden_states: torch.Tensor):
|
| 164 |
+
input_dtype = hidden_states.dtype
|
| 165 |
+
variance = hidden_states.to(torch.float32).pow(2).mean(-1, keepdim=True)
|
| 166 |
+
hidden_states = hidden_states * torch.rsqrt(variance + self.eps)
|
| 167 |
+
|
| 168 |
+
return (self.weight * hidden_states).to(input_dtype)
|
| 169 |
+
|
| 170 |
+
|
| 171 |
+
class CoreAttention(torch.nn.Module):
|
| 172 |
+
def __init__(self, config: ChatGLMConfig, layer_number):
|
| 173 |
+
super(CoreAttention, self).__init__()
|
| 174 |
+
|
| 175 |
+
self.apply_query_key_layer_scaling = config.apply_query_key_layer_scaling
|
| 176 |
+
self.attention_softmax_in_fp32 = config.attention_softmax_in_fp32
|
| 177 |
+
if self.apply_query_key_layer_scaling:
|
| 178 |
+
self.attention_softmax_in_fp32 = True
|
| 179 |
+
self.layer_number = max(1, layer_number)
|
| 180 |
+
|
| 181 |
+
projection_size = config.kv_channels * config.num_attention_heads
|
| 182 |
+
|
| 183 |
+
# Per attention head and per partition values.
|
| 184 |
+
self.hidden_size_per_partition = projection_size
|
| 185 |
+
self.hidden_size_per_attention_head = projection_size // config.num_attention_heads
|
| 186 |
+
self.num_attention_heads_per_partition = config.num_attention_heads
|
| 187 |
+
|
| 188 |
+
self.norm_factor = math.sqrt(self.hidden_size_per_attention_head)
|
| 189 |
+
self.attention_dropout = config.attention_dropout
|
| 190 |
+
|
| 191 |
+
def forward(self, query_layer, key_layer, value_layer, attention_mask):
|
| 192 |
+
seqlen_q, batch_size = query_layer.shape[0], query_layer.shape[1]
|
| 193 |
+
seqlen_k = key_layer.shape[0]
|
| 194 |
+
query_layer, key_layer, value_layer = [rearrange(x, 's b ... -> (b s) ...') for x in [query_layer, key_layer, value_layer]]
|
| 195 |
+
# DO flash_attn_varlen_func
|
| 196 |
+
if attention_mask is None or attention_mask.ndim != 1:
|
| 197 |
+
cu_seqlens_q = torch.arange(0, (batch_size + 1) * seqlen_q, step=seqlen_q, dtype=torch.int32,
|
| 198 |
+
device=query_layer.device)
|
| 199 |
+
else:
|
| 200 |
+
assert seqlen_q == seqlen_k
|
| 201 |
+
cu_seqlens_q = attention_mask
|
| 202 |
+
if self.training:
|
| 203 |
+
assert seqlen_k == seqlen_q
|
| 204 |
+
is_causal = True
|
| 205 |
+
cu_seqlens_k = cu_seqlens_q
|
| 206 |
+
else:
|
| 207 |
+
is_causal = seqlen_q == seqlen_k
|
| 208 |
+
cu_seqlens_k = torch.arange(0, (batch_size + 1) * seqlen_k, step=seqlen_k, dtype=torch.int32,
|
| 209 |
+
device=query_layer.device) if not is_causal else cu_seqlens_q
|
| 210 |
+
self.attention_dropout = 0
|
| 211 |
+
context_layer = flash_attn_unpadded_func(
|
| 212 |
+
query_layer, key_layer, value_layer, cu_seqlens_q, cu_seqlens_k, seqlen_q, seqlen_k,
|
| 213 |
+
self.attention_dropout,
|
| 214 |
+
softmax_scale=1.0 / self.norm_factor, causal=is_causal
|
| 215 |
+
)
|
| 216 |
+
context_layer = rearrange(context_layer, '(b s) ... -> s b ...', b=batch_size)
|
| 217 |
+
new_context_layer_shape = context_layer.size()[:-2] + (self.hidden_size_per_partition,)
|
| 218 |
+
context_layer = context_layer.reshape(*new_context_layer_shape)
|
| 219 |
+
return context_layer
|
| 220 |
+
|
| 221 |
+
|
| 222 |
+
class SelfAttention(torch.nn.Module):
|
| 223 |
+
"""Parallel self-attention layer abstract class.
|
| 224 |
+
|
| 225 |
+
Self-attention layer takes input with size [s, b, h]
|
| 226 |
+
and returns output of the same size.
|
| 227 |
+
"""
|
| 228 |
+
|
| 229 |
+
def __init__(self, config: ChatGLMConfig, layer_number, device=None):
|
| 230 |
+
super(SelfAttention, self).__init__()
|
| 231 |
+
self.layer_number = max(1, layer_number)
|
| 232 |
+
|
| 233 |
+
self.projection_size = config.kv_channels * config.num_attention_heads
|
| 234 |
+
|
| 235 |
+
# Per attention head and per partition values.
|
| 236 |
+
self.hidden_size_per_attention_head = self.projection_size // config.num_attention_heads
|
| 237 |
+
self.num_attention_heads_per_partition = config.num_attention_heads
|
| 238 |
+
|
| 239 |
+
self.multi_query_attention = config.multi_query_attention
|
| 240 |
+
self.qkv_hidden_size = 3 * self.projection_size
|
| 241 |
+
if self.multi_query_attention:
|
| 242 |
+
self.num_multi_query_groups_per_partition = config.multi_query_group_num
|
| 243 |
+
self.qkv_hidden_size = (
|
| 244 |
+
self.projection_size + 2 * self.hidden_size_per_attention_head * config.multi_query_group_num
|
| 245 |
+
)
|
| 246 |
+
self.query_key_value = nn.Linear(config.hidden_size, self.qkv_hidden_size,
|
| 247 |
+
bias=config.add_bias_linear or config.add_qkv_bias,
|
| 248 |
+
device=device, **_config_to_kwargs(config)
|
| 249 |
+
)
|
| 250 |
+
|
| 251 |
+
self.core_attention = CoreAttention(config, self.layer_number)
|
| 252 |
+
|
| 253 |
+
# Output.
|
| 254 |
+
self.dense = nn.Linear(self.projection_size, config.hidden_size, bias=config.add_bias_linear,
|
| 255 |
+
device=device, **_config_to_kwargs(config)
|
| 256 |
+
)
|
| 257 |
+
|
| 258 |
+
def _allocate_memory(self, inference_max_sequence_len, batch_size, device=None, dtype=None):
|
| 259 |
+
if self.multi_query_attention:
|
| 260 |
+
num_attention_heads = self.num_multi_query_groups_per_partition
|
| 261 |
+
else:
|
| 262 |
+
num_attention_heads = self.num_attention_heads_per_partition
|
| 263 |
+
return torch.empty(
|
| 264 |
+
inference_max_sequence_len,
|
| 265 |
+
batch_size,
|
| 266 |
+
num_attention_heads,
|
| 267 |
+
self.hidden_size_per_attention_head,
|
| 268 |
+
dtype=dtype,
|
| 269 |
+
device=device,
|
| 270 |
+
)
|
| 271 |
+
|
| 272 |
+
def forward(
|
| 273 |
+
self, hidden_states, attention_mask, rotary_pos_emb, kv_cache=None, use_cache=True
|
| 274 |
+
):
|
| 275 |
+
# hidden_states: [sq, b, h]
|
| 276 |
+
|
| 277 |
+
# =================================================
|
| 278 |
+
# Pre-allocate memory for key-values for inference.
|
| 279 |
+
# =================================================
|
| 280 |
+
# =====================
|
| 281 |
+
# Query, Key, and Value
|
| 282 |
+
# =====================
|
| 283 |
+
|
| 284 |
+
# Attention heads [sq, b, h] --> [sq, b, (np * 3 * hn)]
|
| 285 |
+
mixed_x_layer = self.query_key_value(hidden_states)
|
| 286 |
+
|
| 287 |
+
if self.multi_query_attention:
|
| 288 |
+
(query_layer, key_layer, value_layer) = mixed_x_layer.split(
|
| 289 |
+
[
|
| 290 |
+
self.num_attention_heads_per_partition * self.hidden_size_per_attention_head,
|
| 291 |
+
self.num_multi_query_groups_per_partition * self.hidden_size_per_attention_head,
|
| 292 |
+
self.num_multi_query_groups_per_partition * self.hidden_size_per_attention_head,
|
| 293 |
+
],
|
| 294 |
+
dim=-1,
|
| 295 |
+
)
|
| 296 |
+
query_layer = query_layer.view(
|
| 297 |
+
query_layer.size()[:-1] + (self.num_attention_heads_per_partition, self.hidden_size_per_attention_head)
|
| 298 |
+
)
|
| 299 |
+
key_layer = key_layer.view(
|
| 300 |
+
key_layer.size()[:-1] + (self.num_multi_query_groups_per_partition, self.hidden_size_per_attention_head)
|
| 301 |
+
)
|
| 302 |
+
value_layer = value_layer.view(
|
| 303 |
+
value_layer.size()[:-1]
|
| 304 |
+
+ (self.num_multi_query_groups_per_partition, self.hidden_size_per_attention_head)
|
| 305 |
+
)
|
| 306 |
+
else:
|
| 307 |
+
new_tensor_shape = mixed_x_layer.size()[:-1] + \
|
| 308 |
+
(self.num_attention_heads_per_partition,
|
| 309 |
+
3 * self.hidden_size_per_attention_head)
|
| 310 |
+
mixed_x_layer = mixed_x_layer.view(*new_tensor_shape)
|
| 311 |
+
|
| 312 |
+
# [sq, b, np, 3 * hn] --> 3 [sq, b, np, hn]
|
| 313 |
+
(query_layer, key_layer, value_layer) = split_tensor_along_last_dim(mixed_x_layer, 3)
|
| 314 |
+
|
| 315 |
+
# apply relative positional encoding (rotary embedding)
|
| 316 |
+
if rotary_pos_emb is not None:
|
| 317 |
+
query_layer = apply_rotary_pos_emb(query_layer, rotary_pos_emb)
|
| 318 |
+
key_layer = apply_rotary_pos_emb(key_layer, rotary_pos_emb)
|
| 319 |
+
|
| 320 |
+
# adjust key and value for inference
|
| 321 |
+
if use_cache:
|
| 322 |
+
if kv_cache is not None:
|
| 323 |
+
cache_k, cache_v = kv_cache
|
| 324 |
+
key_layer = torch.cat((cache_k, key_layer), dim=0)
|
| 325 |
+
value_layer = torch.cat((cache_v, value_layer), dim=0)
|
| 326 |
+
kv_cache = (key_layer, value_layer)
|
| 327 |
+
else:
|
| 328 |
+
kv_cache = None
|
| 329 |
+
|
| 330 |
+
|
| 331 |
+
if self.multi_query_attention:
|
| 332 |
+
key_layer = key_layer.unsqueeze(-2)
|
| 333 |
+
key_layer = key_layer.expand(
|
| 334 |
+
-1, -1, -1, self.num_attention_heads_per_partition // self.num_multi_query_groups_per_partition, -1
|
| 335 |
+
)
|
| 336 |
+
key_layer = key_layer.contiguous().view(
|
| 337 |
+
key_layer.size()[:2] + (self.num_attention_heads_per_partition, self.hidden_size_per_attention_head)
|
| 338 |
+
)
|
| 339 |
+
value_layer = value_layer.unsqueeze(-2)
|
| 340 |
+
value_layer = value_layer.expand(
|
| 341 |
+
-1, -1, -1, self.num_attention_heads_per_partition // self.num_multi_query_groups_per_partition, -1
|
| 342 |
+
)
|
| 343 |
+
value_layer = value_layer.contiguous().view(
|
| 344 |
+
value_layer.size()[:2] + (self.num_attention_heads_per_partition, self.hidden_size_per_attention_head)
|
| 345 |
+
)
|
| 346 |
+
|
| 347 |
+
# ==================================
|
| 348 |
+
# core attention computation
|
| 349 |
+
# ==================================
|
| 350 |
+
|
| 351 |
+
context_layer = self.core_attention(query_layer, key_layer, value_layer, attention_mask)
|
| 352 |
+
|
| 353 |
+
# =================
|
| 354 |
+
# Output. [sq, b, h]
|
| 355 |
+
# =================
|
| 356 |
+
|
| 357 |
+
output = self.dense(context_layer)
|
| 358 |
+
|
| 359 |
+
return output, kv_cache
|
| 360 |
+
|
| 361 |
+
|
| 362 |
+
def _config_to_kwargs(args):
|
| 363 |
+
common_kwargs = {
|
| 364 |
+
"dtype": args.torch_dtype,
|
| 365 |
+
}
|
| 366 |
+
return common_kwargs
|
| 367 |
+
|
| 368 |
+
|
| 369 |
+
class MLP(torch.nn.Module):
|
| 370 |
+
"""MLP.
|
| 371 |
+
|
| 372 |
+
MLP will take the input with h hidden state, project it to 4*h
|
| 373 |
+
hidden dimension, perform nonlinear transformation, and project the
|
| 374 |
+
state back into h hidden dimension.
|
| 375 |
+
"""
|
| 376 |
+
|
| 377 |
+
def __init__(self, config: ChatGLMConfig, device=None):
|
| 378 |
+
super(MLP, self).__init__()
|
| 379 |
+
|
| 380 |
+
self.add_bias = config.add_bias_linear
|
| 381 |
+
|
| 382 |
+
# Project to 4h. If using swiglu double the output width, see https://arxiv.org/pdf/2002.05202.pdf
|
| 383 |
+
self.dense_h_to_4h = nn.Linear(
|
| 384 |
+
config.hidden_size,
|
| 385 |
+
config.ffn_hidden_size * 2,
|
| 386 |
+
bias=self.add_bias,
|
| 387 |
+
device=device,
|
| 388 |
+
**_config_to_kwargs(config)
|
| 389 |
+
)
|
| 390 |
+
|
| 391 |
+
def swiglu(x):
|
| 392 |
+
x = torch.chunk(x, 2, dim=-1)
|
| 393 |
+
return F.silu(x[0]) * x[1]
|
| 394 |
+
|
| 395 |
+
self.activation_func = swiglu
|
| 396 |
+
|
| 397 |
+
# Project back to h.
|
| 398 |
+
self.dense_4h_to_h = nn.Linear(
|
| 399 |
+
config.ffn_hidden_size,
|
| 400 |
+
config.hidden_size,
|
| 401 |
+
bias=self.add_bias,
|
| 402 |
+
device=device,
|
| 403 |
+
**_config_to_kwargs(config)
|
| 404 |
+
)
|
| 405 |
+
|
| 406 |
+
def forward(self, hidden_states):
|
| 407 |
+
# [s, b, 4hp]
|
| 408 |
+
intermediate_parallel = self.dense_h_to_4h(hidden_states)
|
| 409 |
+
intermediate_parallel = self.activation_func(intermediate_parallel)
|
| 410 |
+
# [s, b, h]
|
| 411 |
+
output = self.dense_4h_to_h(intermediate_parallel)
|
| 412 |
+
return output
|
| 413 |
+
|
| 414 |
+
|
| 415 |
+
class GLMBlock(torch.nn.Module):
|
| 416 |
+
"""A single transformer layer.
|
| 417 |
+
|
| 418 |
+
Transformer layer takes input with size [s, b, h] and returns an
|
| 419 |
+
output of the same size.
|
| 420 |
+
"""
|
| 421 |
+
|
| 422 |
+
def __init__(self, config: ChatGLMConfig, layer_number, device=None):
|
| 423 |
+
super(GLMBlock, self).__init__()
|
| 424 |
+
self.layer_number = layer_number
|
| 425 |
+
|
| 426 |
+
self.apply_residual_connection_post_layernorm = config.apply_residual_connection_post_layernorm
|
| 427 |
+
|
| 428 |
+
self.fp32_residual_connection = config.fp32_residual_connection
|
| 429 |
+
|
| 430 |
+
LayerNormFunc = RMSNorm if config.rmsnorm else LayerNorm
|
| 431 |
+
# Layernorm on the input data.
|
| 432 |
+
self.input_layernorm = LayerNormFunc(config.hidden_size, eps=config.layernorm_epsilon, device=device,
|
| 433 |
+
dtype=config.torch_dtype)
|
| 434 |
+
|
| 435 |
+
# Self attention.
|
| 436 |
+
self.self_attention = SelfAttention(config, layer_number, device=device)
|
| 437 |
+
self.hidden_dropout = config.hidden_dropout
|
| 438 |
+
|
| 439 |
+
# Layernorm on the attention output
|
| 440 |
+
self.post_attention_layernorm = LayerNormFunc(config.hidden_size, eps=config.layernorm_epsilon, device=device,
|
| 441 |
+
dtype=config.torch_dtype)
|
| 442 |
+
|
| 443 |
+
# MLP
|
| 444 |
+
self.mlp = MLP(config, device=device)
|
| 445 |
+
|
| 446 |
+
def forward(
|
| 447 |
+
self, hidden_states, attention_mask, rotary_pos_emb, kv_cache=None, use_cache=True,
|
| 448 |
+
):
|
| 449 |
+
# hidden_states: [s, b, h]
|
| 450 |
+
|
| 451 |
+
# Layer norm at the beginning of the transformer layer.
|
| 452 |
+
layernorm_output = self.input_layernorm(hidden_states)
|
| 453 |
+
# Self attention.
|
| 454 |
+
attention_output, kv_cache = self.self_attention(
|
| 455 |
+
layernorm_output,
|
| 456 |
+
attention_mask,
|
| 457 |
+
rotary_pos_emb,
|
| 458 |
+
kv_cache=kv_cache,
|
| 459 |
+
use_cache=use_cache
|
| 460 |
+
)
|
| 461 |
+
|
| 462 |
+
# Residual connection.
|
| 463 |
+
if self.apply_residual_connection_post_layernorm:
|
| 464 |
+
residual = layernorm_output
|
| 465 |
+
else:
|
| 466 |
+
residual = hidden_states
|
| 467 |
+
|
| 468 |
+
layernorm_input = torch.nn.functional.dropout(attention_output, p=self.hidden_dropout, training=self.training)
|
| 469 |
+
layernorm_input = residual + layernorm_input
|
| 470 |
+
|
| 471 |
+
# Layer norm post the self attention.
|
| 472 |
+
layernorm_output = self.post_attention_layernorm(layernorm_input)
|
| 473 |
+
|
| 474 |
+
# MLP.
|
| 475 |
+
mlp_output = self.mlp(layernorm_output)
|
| 476 |
+
|
| 477 |
+
# Second residual connection.
|
| 478 |
+
if self.apply_residual_connection_post_layernorm:
|
| 479 |
+
residual = layernorm_output
|
| 480 |
+
else:
|
| 481 |
+
residual = layernorm_input
|
| 482 |
+
|
| 483 |
+
output = torch.nn.functional.dropout(mlp_output, p=self.hidden_dropout, training=self.training)
|
| 484 |
+
output = residual + output
|
| 485 |
+
|
| 486 |
+
return output, kv_cache
|
| 487 |
+
|
| 488 |
+
|
| 489 |
+
class GLMTransformer(torch.nn.Module):
|
| 490 |
+
"""Transformer class."""
|
| 491 |
+
|
| 492 |
+
def __init__(self, config: ChatGLMConfig, device=None):
|
| 493 |
+
super(GLMTransformer, self).__init__()
|
| 494 |
+
|
| 495 |
+
self.fp32_residual_connection = config.fp32_residual_connection
|
| 496 |
+
self.post_layer_norm = config.post_layer_norm
|
| 497 |
+
|
| 498 |
+
# Number of layers.
|
| 499 |
+
self.num_layers = config.num_layers
|
| 500 |
+
|
| 501 |
+
# Transformer layers.
|
| 502 |
+
def build_layer(layer_number):
|
| 503 |
+
return GLMBlock(config, layer_number, device=device)
|
| 504 |
+
|
| 505 |
+
self.layers = torch.nn.ModuleList([build_layer(i + 1) for i in range(self.num_layers)])
|
| 506 |
+
|
| 507 |
+
if self.post_layer_norm:
|
| 508 |
+
LayerNormFunc = RMSNorm if config.rmsnorm else LayerNorm
|
| 509 |
+
# Final layer norm before output.
|
| 510 |
+
self.final_layernorm = LayerNormFunc(config.hidden_size, eps=config.layernorm_epsilon, device=device,
|
| 511 |
+
dtype=config.torch_dtype)
|
| 512 |
+
|
| 513 |
+
self.gradient_checkpointing = False
|
| 514 |
+
|
| 515 |
+
def _get_layer(self, layer_number):
|
| 516 |
+
return self.layers[layer_number]
|
| 517 |
+
|
| 518 |
+
def forward(
|
| 519 |
+
self, hidden_states, attention_mask, rotary_pos_emb, kv_caches=None,
|
| 520 |
+
use_cache: Optional[bool] = True,
|
| 521 |
+
output_hidden_states: Optional[bool] = False,
|
| 522 |
+
):
|
| 523 |
+
if not kv_caches:
|
| 524 |
+
kv_caches = [None for _ in range(self.num_layers)]
|
| 525 |
+
presents = () if use_cache else None
|
| 526 |
+
if self.gradient_checkpointing and self.training:
|
| 527 |
+
if use_cache:
|
| 528 |
+
# logger.warning_once(
|
| 529 |
+
# "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`..."
|
| 530 |
+
# )
|
| 531 |
+
use_cache = False
|
| 532 |
+
|
| 533 |
+
all_self_attentions = None
|
| 534 |
+
all_hidden_states = () if output_hidden_states else None
|
| 535 |
+
for index in range(self.num_layers):
|
| 536 |
+
if output_hidden_states:
|
| 537 |
+
all_hidden_states = all_hidden_states + (hidden_states,)
|
| 538 |
+
|
| 539 |
+
layer = self._get_layer(index)
|
| 540 |
+
if self.gradient_checkpointing and self.training:
|
| 541 |
+
layer_ret = torch.utils.checkpoint.checkpoint(
|
| 542 |
+
layer,
|
| 543 |
+
hidden_states,
|
| 544 |
+
attention_mask,
|
| 545 |
+
rotary_pos_emb,
|
| 546 |
+
kv_caches[index],
|
| 547 |
+
use_cache,
|
| 548 |
+
use_reentrant=False
|
| 549 |
+
)
|
| 550 |
+
else:
|
| 551 |
+
layer_ret = layer(
|
| 552 |
+
hidden_states,
|
| 553 |
+
attention_mask,
|
| 554 |
+
rotary_pos_emb,
|
| 555 |
+
kv_cache=kv_caches[index],
|
| 556 |
+
use_cache=use_cache
|
| 557 |
+
)
|
| 558 |
+
hidden_states, kv_cache = layer_ret
|
| 559 |
+
if use_cache:
|
| 560 |
+
presents = presents + (kv_cache,)
|
| 561 |
+
|
| 562 |
+
if output_hidden_states:
|
| 563 |
+
all_hidden_states = all_hidden_states + (hidden_states,)
|
| 564 |
+
|
| 565 |
+
# Final layer norm.
|
| 566 |
+
if self.post_layer_norm:
|
| 567 |
+
hidden_states = self.final_layernorm(hidden_states)
|
| 568 |
+
|
| 569 |
+
return hidden_states, presents, all_hidden_states, all_self_attentions
|
| 570 |
+
|
| 571 |
+
|
| 572 |
+
class ChatGLMPreTrainedModel(PreTrainedModel):
|
| 573 |
+
"""
|
| 574 |
+
An abstract class to handle weights initialization and
|
| 575 |
+
a simple interface for downloading and loading pretrained models.
|
| 576 |
+
"""
|
| 577 |
+
|
| 578 |
+
is_parallelizable = False
|
| 579 |
+
supports_gradient_checkpointing = True
|
| 580 |
+
config_class = ChatGLMConfig
|
| 581 |
+
base_model_prefix = "transformer"
|
| 582 |
+
_no_split_modules = ["GLMBlock"]
|
| 583 |
+
|
| 584 |
+
def _init_weights(self, module: nn.Module):
|
| 585 |
+
"""Initialize the weights."""
|
| 586 |
+
return
|
| 587 |
+
|
| 588 |
+
def get_masks(self, input_ids, past_key_values, padding_mask=None):
|
| 589 |
+
batch_size, seq_length = input_ids.shape
|
| 590 |
+
full_attention_mask = torch.ones(batch_size, seq_length, seq_length, device=input_ids.device)
|
| 591 |
+
full_attention_mask.tril_()
|
| 592 |
+
past_length = 0
|
| 593 |
+
if past_key_values:
|
| 594 |
+
past_length = past_key_values[0][0].shape[0]
|
| 595 |
+
if past_length:
|
| 596 |
+
full_attention_mask = torch.cat((torch.ones(batch_size, seq_length, past_length,
|
| 597 |
+
device=input_ids.device), full_attention_mask), dim=-1)
|
| 598 |
+
if padding_mask is not None:
|
| 599 |
+
full_attention_mask = full_attention_mask * padding_mask.unsqueeze(1)
|
| 600 |
+
if not past_length and padding_mask is not None:
|
| 601 |
+
full_attention_mask -= padding_mask.unsqueeze(-1) - 1
|
| 602 |
+
full_attention_mask = (full_attention_mask < 0.5).bool()
|
| 603 |
+
full_attention_mask.unsqueeze_(1)
|
| 604 |
+
return full_attention_mask
|
| 605 |
+
|
| 606 |
+
def get_position_ids(self, input_ids, device):
|
| 607 |
+
batch_size, seq_length = input_ids.shape
|
| 608 |
+
position_ids = torch.arange(seq_length, dtype=torch.long, device=device).unsqueeze(0).repeat(batch_size, 1)
|
| 609 |
+
return position_ids
|
| 610 |
+
|
| 611 |
+
def _set_gradient_checkpointing(self, module, value=False):
|
| 612 |
+
if isinstance(module, GLMTransformer):
|
| 613 |
+
module.gradient_checkpointing = value
|
| 614 |
+
|
| 615 |
+
|
| 616 |
+
class Embedding(torch.nn.Module):
|
| 617 |
+
"""Language model embeddings."""
|
| 618 |
+
|
| 619 |
+
def __init__(self, config: ChatGLMConfig, device=None):
|
| 620 |
+
super(Embedding, self).__init__()
|
| 621 |
+
|
| 622 |
+
self.hidden_size = config.hidden_size
|
| 623 |
+
# Word embeddings (parallel).
|
| 624 |
+
self.word_embeddings = nn.Embedding(
|
| 625 |
+
config.padded_vocab_size,
|
| 626 |
+
self.hidden_size,
|
| 627 |
+
dtype=config.torch_dtype,
|
| 628 |
+
device=device
|
| 629 |
+
)
|
| 630 |
+
self.fp32_residual_connection = config.fp32_residual_connection
|
| 631 |
+
|
| 632 |
+
def forward(self, input_ids):
|
| 633 |
+
# Embeddings.
|
| 634 |
+
words_embeddings = self.word_embeddings(input_ids)
|
| 635 |
+
embeddings = words_embeddings
|
| 636 |
+
# Data format change to avoid explicit tranposes : [b s h] --> [s b h].
|
| 637 |
+
embeddings = embeddings.transpose(0, 1).contiguous()
|
| 638 |
+
# If the input flag for fp32 residual connection is set, convert for float.
|
| 639 |
+
if self.fp32_residual_connection:
|
| 640 |
+
embeddings = embeddings.float()
|
| 641 |
+
return embeddings
|
| 642 |
+
|
| 643 |
+
|
| 644 |
+
class ChatGLMModel(ChatGLMPreTrainedModel):
|
| 645 |
+
def __init__(self, config: ChatGLMConfig, device=None, empty_init=True):
|
| 646 |
+
super().__init__(config)
|
| 647 |
+
if empty_init:
|
| 648 |
+
init_method = skip_init
|
| 649 |
+
else:
|
| 650 |
+
init_method = default_init
|
| 651 |
+
init_kwargs = {}
|
| 652 |
+
if device is not None:
|
| 653 |
+
init_kwargs["device"] = device
|
| 654 |
+
self.embedding = init_method(Embedding, config, **init_kwargs)
|
| 655 |
+
|
| 656 |
+
# Rotary positional embeddings
|
| 657 |
+
self.seq_length = config.seq_length
|
| 658 |
+
rotary_dim = (
|
| 659 |
+
config.hidden_size // config.num_attention_heads if config.kv_channels is None else config.kv_channels
|
| 660 |
+
)
|
| 661 |
+
|
| 662 |
+
self.rotary_pos_emb = RotaryEmbedding(rotary_dim // 2, rope_ratio=config.rope_ratio, original_impl=config.original_rope,
|
| 663 |
+
device=device, dtype=config.torch_dtype)
|
| 664 |
+
self.encoder = init_method(GLMTransformer, config, **init_kwargs)
|
| 665 |
+
self.output_layer = init_method(nn.Linear, config.hidden_size, config.padded_vocab_size, bias=False,
|
| 666 |
+
dtype=config.torch_dtype, **init_kwargs)
|
| 667 |
+
|
| 668 |
+
def get_input_embeddings(self):
|
| 669 |
+
return self.embedding.word_embeddings
|
| 670 |
+
|
| 671 |
+
def forward(
|
| 672 |
+
self,
|
| 673 |
+
input_ids,
|
| 674 |
+
position_ids: Optional[torch.Tensor] = None,
|
| 675 |
+
attention_mask: Optional[torch.BoolTensor] = None,
|
| 676 |
+
full_attention_mask: Optional[torch.BoolTensor] = None,
|
| 677 |
+
past_key_values: Optional[Tuple[Tuple[torch.Tensor, torch.Tensor], ...]] = None,
|
| 678 |
+
inputs_embeds: Optional[torch.Tensor] = None,
|
| 679 |
+
use_cache: Optional[bool] = None,
|
| 680 |
+
output_hidden_states: Optional[bool] = None,
|
| 681 |
+
return_dict: Optional[bool] = None,
|
| 682 |
+
):
|
| 683 |
+
output_hidden_states = (
|
| 684 |
+
output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
|
| 685 |
+
)
|
| 686 |
+
use_cache = use_cache if use_cache is not None else self.config.use_cache
|
| 687 |
+
return_dict = return_dict if return_dict is not None else self.config.use_return_dict
|
| 688 |
+
|
| 689 |
+
batch_size, seq_length = input_ids.shape
|
| 690 |
+
|
| 691 |
+
if inputs_embeds is None:
|
| 692 |
+
inputs_embeds = self.embedding(input_ids)
|
| 693 |
+
|
| 694 |
+
# if full_attention_mask is None:
|
| 695 |
+
# if (attention_mask is not None and not attention_mask.all()) or (past_key_values and seq_length != 1):
|
| 696 |
+
# full_attention_mask = self.get_masks(input_ids, past_key_values, padding_mask=attention_mask)
|
| 697 |
+
|
| 698 |
+
# Rotary positional embeddings
|
| 699 |
+
rotary_pos_emb = self.rotary_pos_emb(self.seq_length)
|
| 700 |
+
if position_ids is not None:
|
| 701 |
+
rotary_pos_emb = rotary_pos_emb[position_ids]
|
| 702 |
+
else:
|
| 703 |
+
rotary_pos_emb = rotary_pos_emb[None, :seq_length]
|
| 704 |
+
rotary_pos_emb = rotary_pos_emb.transpose(0, 1).contiguous()
|
| 705 |
+
|
| 706 |
+
# Run encoder.
|
| 707 |
+
hidden_states, presents, all_hidden_states, all_self_attentions = self.encoder(
|
| 708 |
+
inputs_embeds, attention_mask, rotary_pos_emb=rotary_pos_emb,
|
| 709 |
+
kv_caches=past_key_values, use_cache=use_cache, output_hidden_states=output_hidden_states
|
| 710 |
+
)
|
| 711 |
+
|
| 712 |
+
if not return_dict:
|
| 713 |
+
return tuple(v for v in [hidden_states, presents, all_hidden_states, all_self_attentions] if v is not None)
|
| 714 |
+
|
| 715 |
+
return BaseModelOutputWithPast(
|
| 716 |
+
last_hidden_state=hidden_states,
|
| 717 |
+
past_key_values=presents,
|
| 718 |
+
hidden_states=all_hidden_states,
|
| 719 |
+
attentions=all_self_attentions,
|
| 720 |
+
)
|
| 721 |
+
|
| 722 |
+
|
| 723 |
+
class ChatGLMForConditionalGeneration(ChatGLMPreTrainedModel):
|
| 724 |
+
def __init__(self, config: ChatGLMConfig, empty_init=True, device=None):
|
| 725 |
+
super().__init__(config)
|
| 726 |
+
|
| 727 |
+
self.max_sequence_length = config.max_length
|
| 728 |
+
self.transformer = ChatGLMModel(config, empty_init=empty_init, device=device)
|
| 729 |
+
self.config = config
|
| 730 |
+
self.pack_loss = False
|
| 731 |
+
|
| 732 |
+
def _update_model_kwargs_for_generation(
|
| 733 |
+
self,
|
| 734 |
+
outputs: ModelOutput,
|
| 735 |
+
model_kwargs: Dict[str, Any],
|
| 736 |
+
is_encoder_decoder: bool = False,
|
| 737 |
+
) -> Dict[str, Any]:
|
| 738 |
+
# update past_key_values
|
| 739 |
+
cache_name, cache = self._extract_past_from_model_output(outputs)
|
| 740 |
+
model_kwargs[cache_name] = cache
|
| 741 |
+
|
| 742 |
+
# update attention mask
|
| 743 |
+
if "attention_mask" in model_kwargs:
|
| 744 |
+
attention_mask = model_kwargs["attention_mask"]
|
| 745 |
+
model_kwargs["attention_mask"] = torch.cat(
|
| 746 |
+
[attention_mask, attention_mask.new_ones((attention_mask.shape[0], 1))], dim=-1
|
| 747 |
+
)
|
| 748 |
+
|
| 749 |
+
# update position ids
|
| 750 |
+
if "position_ids" in model_kwargs:
|
| 751 |
+
position_ids = model_kwargs["position_ids"]
|
| 752 |
+
new_position_id = position_ids[..., -1:].clone()
|
| 753 |
+
new_position_id += 1
|
| 754 |
+
model_kwargs["position_ids"] = torch.cat(
|
| 755 |
+
[position_ids, new_position_id], dim=-1
|
| 756 |
+
)
|
| 757 |
+
|
| 758 |
+
model_kwargs["is_first_forward"] = False
|
| 759 |
+
return model_kwargs
|
| 760 |
+
|
| 761 |
+
def prepare_inputs_for_generation(
|
| 762 |
+
self,
|
| 763 |
+
input_ids: torch.LongTensor,
|
| 764 |
+
past_key_values: Optional[torch.Tensor] = None,
|
| 765 |
+
attention_mask: Optional[torch.Tensor] = None,
|
| 766 |
+
position_ids: Optional[torch.Tensor] = None,
|
| 767 |
+
is_first_forward: bool = True,
|
| 768 |
+
**kwargs
|
| 769 |
+
) -> dict:
|
| 770 |
+
# only last token for input_ids if past is not None
|
| 771 |
+
if position_ids is None:
|
| 772 |
+
position_ids = self.get_position_ids(input_ids, device=input_ids.device)
|
| 773 |
+
if not is_first_forward:
|
| 774 |
+
position_ids = position_ids[..., -1:]
|
| 775 |
+
input_ids = input_ids[:, -1:]
|
| 776 |
+
return {
|
| 777 |
+
"input_ids": input_ids,
|
| 778 |
+
"past_key_values": past_key_values,
|
| 779 |
+
"position_ids": position_ids,
|
| 780 |
+
"attention_mask": attention_mask,
|
| 781 |
+
"return_last_logit": True
|
| 782 |
+
}
|
| 783 |
+
|
| 784 |
+
def forward(
|
| 785 |
+
self,
|
| 786 |
+
input_ids: Optional[torch.Tensor] = None,
|
| 787 |
+
position_ids: Optional[torch.Tensor] = None,
|
| 788 |
+
attention_mask: Optional[torch.Tensor] = None,
|
| 789 |
+
past_key_values: Optional[Tuple[torch.FloatTensor]] = None,
|
| 790 |
+
inputs_embeds: Optional[torch.Tensor] = None,
|
| 791 |
+
labels: Optional[Tuple[torch.Tensor]] = None,
|
| 792 |
+
use_cache: Optional[bool] = None,
|
| 793 |
+
output_attentions: Optional[bool] = None,
|
| 794 |
+
output_hidden_states: Optional[bool] = None,
|
| 795 |
+
return_dict: Optional[bool] = None,
|
| 796 |
+
return_last_logit: Optional[bool] = False,
|
| 797 |
+
):
|
| 798 |
+
use_cache = use_cache if use_cache is not None else self.config.use_cache
|
| 799 |
+
return_dict = return_dict if return_dict is not None else self.config.use_return_dict
|
| 800 |
+
|
| 801 |
+
transformer_outputs = self.transformer(
|
| 802 |
+
input_ids=input_ids,
|
| 803 |
+
position_ids=position_ids,
|
| 804 |
+
attention_mask=attention_mask,
|
| 805 |
+
past_key_values=past_key_values,
|
| 806 |
+
inputs_embeds=inputs_embeds,
|
| 807 |
+
use_cache=use_cache,
|
| 808 |
+
output_hidden_states=output_hidden_states,
|
| 809 |
+
return_dict=return_dict,
|
| 810 |
+
)
|
| 811 |
+
|
| 812 |
+
hidden_states = transformer_outputs[0]
|
| 813 |
+
if return_last_logit:
|
| 814 |
+
hidden_states = hidden_states[-1:]
|
| 815 |
+
lm_logits = self.transformer.output_layer(hidden_states)
|
| 816 |
+
lm_logits = lm_logits.transpose(0, 1).contiguous()
|
| 817 |
+
|
| 818 |
+
loss = None
|
| 819 |
+
if labels is not None:
|
| 820 |
+
lm_logits = lm_logits.to(torch.float32)
|
| 821 |
+
# Shift so that tokens < n predict n
|
| 822 |
+
shift_logits = lm_logits[..., :-1, :].contiguous()
|
| 823 |
+
if isinstance(labels, tuple) or isinstance(labels, list):
|
| 824 |
+
labels, weights = labels
|
| 825 |
+
shift_labels = labels[..., 1:].contiguous()
|
| 826 |
+
if self.pack_loss:
|
| 827 |
+
loss_fct = CrossEntropyLoss(ignore_index=-100)#, reduction='none')
|
| 828 |
+
loss = loss_fct(shift_logits.view(-1, shift_logits.size(-1)), shift_labels.view(-1))
|
| 829 |
+
loss *= weights
|
| 830 |
+
# if self.pack_loss:
|
| 831 |
+
# shift_weights = weights[..., 1:].contiguous()
|
| 832 |
+
# loss_fct = CrossEntropyLoss(ignore_index=-100, reduction='none')
|
| 833 |
+
# loss = loss_fct(shift_logits.view(-1, shift_logits.size(-1)), shift_labels.view(-1))
|
| 834 |
+
# loss = (loss * shift_weights).sum()
|
| 835 |
+
else:
|
| 836 |
+
loss_fct = CrossEntropyLoss(ignore_index=-100)
|
| 837 |
+
loss = loss_fct(shift_logits.view(-1, shift_logits.size(-1)), shift_labels.view(-1))
|
| 838 |
+
|
| 839 |
+
lm_logits = lm_logits.to(hidden_states.dtype)
|
| 840 |
+
loss = loss.to(hidden_states.dtype)
|
| 841 |
+
|
| 842 |
+
if not return_dict:
|
| 843 |
+
output = (lm_logits,) + transformer_outputs[1:]
|
| 844 |
+
return ((loss,) + output) if loss is not None else output
|
| 845 |
+
|
| 846 |
+
return CausalLMOutputWithPast(
|
| 847 |
+
loss=loss,
|
| 848 |
+
logits=lm_logits,
|
| 849 |
+
past_key_values=transformer_outputs.past_key_values,
|
| 850 |
+
hidden_states=transformer_outputs.hidden_states,
|
| 851 |
+
attentions=transformer_outputs.attentions,
|
| 852 |
+
)
|
| 853 |
+
|
| 854 |
+
@staticmethod
|
| 855 |
+
def _reorder_cache(
|
| 856 |
+
past: Tuple[Tuple[torch.Tensor, torch.Tensor], ...], beam_idx: torch.LongTensor
|
| 857 |
+
) -> Tuple[Tuple[torch.Tensor, torch.Tensor], ...]:
|
| 858 |
+
"""
|
| 859 |
+
This function is used to re-order the `past_key_values` cache if [`~PreTrainedModel.beam_search`] or
|
| 860 |
+
[`~PreTrainedModel.beam_sample`] is called. This is required to match `past_key_values` with the correct
|
| 861 |
+
beam_idx at every generation step.
|
| 862 |
+
|
| 863 |
+
Output shares the same memory storage as `past`.
|
| 864 |
+
"""
|
| 865 |
+
return tuple(
|
| 866 |
+
(
|
| 867 |
+
layer_past[0].index_select(1, beam_idx.to(layer_past[0].device)),
|
| 868 |
+
layer_past[1].index_select(1, beam_idx.to(layer_past[1].device)),
|
| 869 |
+
)
|
| 870 |
+
for layer_past in past
|
| 871 |
+
)
|
| 872 |
+
|
| 873 |
+
def process_response(self, response):
|
| 874 |
+
response = response.strip()
|
| 875 |
+
response = response.replace("[[训练时间]]", "2023年")
|
| 876 |
+
return response
|
| 877 |
+
|
| 878 |
+
@torch.inference_mode()
|
| 879 |
+
def chat(self, tokenizer, query: str, history: List[Dict] = None, role: str = "user",
|
| 880 |
+
max_length: int = 8192, num_beams=1, do_sample=True, top_p=0.8, temperature=0.8, logits_processor=None,
|
| 881 |
+
**kwargs):
|
| 882 |
+
if history is None:
|
| 883 |
+
history = []
|
| 884 |
+
if logits_processor is None:
|
| 885 |
+
logits_processor = LogitsProcessorList()
|
| 886 |
+
logits_processor.append(InvalidScoreLogitsProcessor())
|
| 887 |
+
gen_kwargs = {"max_length": max_length, "num_beams": num_beams, "do_sample": do_sample, "top_p": top_p,
|
| 888 |
+
"temperature": temperature, "logits_processor": logits_processor, **kwargs}
|
| 889 |
+
inputs = tokenizer.build_chat_input(query, history=history, role=role)
|
| 890 |
+
inputs = inputs.to(self.device)
|
| 891 |
+
eos_token_id = [tokenizer.eos_token_id, tokenizer.get_command("<|user|>"),
|
| 892 |
+
tokenizer.get_command("<|observation|>")]
|
| 893 |
+
outputs = self.generate(**inputs, **gen_kwargs, eos_token_id=eos_token_id)
|
| 894 |
+
outputs = outputs.tolist()[0][len(inputs["input_ids"][0]):-1]
|
| 895 |
+
response = tokenizer.decode(outputs)
|
| 896 |
+
history.append({"role": role, "content": query})
|
| 897 |
+
response = self.process_response(response)
|
| 898 |
+
return response, history
|
smash_config.json
ADDED
|
@@ -0,0 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"api_key": null,
|
| 3 |
+
"verify_url": "http://johnrachwan.pythonanywhere.com",
|
| 4 |
+
"smash_config": {
|
| 5 |
+
"pruners": "None",
|
| 6 |
+
"pruning_ratio": 0.0,
|
| 7 |
+
"factorizers": "None",
|
| 8 |
+
"quantizers": "['llm-int8']",
|
| 9 |
+
"weight_quantization_bits": 4,
|
| 10 |
+
"output_deviation": 0.005,
|
| 11 |
+
"compilers": "None",
|
| 12 |
+
"static_batch": true,
|
| 13 |
+
"static_shape": true,
|
| 14 |
+
"controlnet": "None",
|
| 15 |
+
"unet_dim": 4,
|
| 16 |
+
"device": "cuda",
|
| 17 |
+
"cache_dir": "/ceph/hdd/staff/charpent/.cache/modelsdhkyhu_n",
|
| 18 |
+
"batch_size": 1,
|
| 19 |
+
"model_name": "THUDM/LongWriter-glm4-9b",
|
| 20 |
+
"task": "text_text_generation",
|
| 21 |
+
"max_batch_size": 1,
|
| 22 |
+
"qtype_weight": "torch.qint8",
|
| 23 |
+
"qtype_activation": "torch.quint8",
|
| 24 |
+
"qobserver": "<class 'torch.ao.quantization.observer.MinMaxObserver'>",
|
| 25 |
+
"qscheme": "torch.per_tensor_symmetric",
|
| 26 |
+
"qconfig": "x86",
|
| 27 |
+
"group_size": 128,
|
| 28 |
+
"damp_percent": 0.1,
|
| 29 |
+
"save_load_fn": "bitsandbytes"
|
| 30 |
+
}
|
| 31 |
+
}
|
special_tokens_map.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"eos_token": {
|
| 3 |
+
"content": "<|endoftext|>",
|
| 4 |
+
"lstrip": false,
|
| 5 |
+
"normalized": false,
|
| 6 |
+
"rstrip": false,
|
| 7 |
+
"single_word": false
|
| 8 |
+
}
|
| 9 |
+
}
|
tokenizer.model
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5a493598071550244b2ee7f26118f3edec2150b9dfa967929a99052ac83fe716
|
| 3 |
+
size 2623634
|
tokenizer_config.json
ADDED
|
@@ -0,0 +1,27 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"added_tokens_decoder": {
|
| 3 |
+
"151329": {
|
| 4 |
+
"content": "<|endoftext|>",
|
| 5 |
+
"lstrip": false,
|
| 6 |
+
"normalized": false,
|
| 7 |
+
"rstrip": false,
|
| 8 |
+
"single_word": false,
|
| 9 |
+
"special": true
|
| 10 |
+
}
|
| 11 |
+
},
|
| 12 |
+
"auto_map": {
|
| 13 |
+
"AutoTokenizer": [
|
| 14 |
+
"THUDM/LongWriter-glm4-9b--tokenization_chatglm.ChatGLM4Tokenizer",
|
| 15 |
+
null
|
| 16 |
+
]
|
| 17 |
+
},
|
| 18 |
+
"chat_template": "{% for message in messages %}{% if loop.first %}[gMASK]sop<|{{ message['role'] }}|>\n {{ message['content'] }}{% else %}<|{{ message['role'] }}|>\n {{ message['content'] }}{% endif %}{% endfor %}{% if add_generation_prompt %}<|assistant|>{% endif %}",
|
| 19 |
+
"clean_up_tokenization_spaces": false,
|
| 20 |
+
"do_lower_case": false,
|
| 21 |
+
"eos_token": "<|endoftext|>",
|
| 22 |
+
"legacy": false,
|
| 23 |
+
"model_max_length": 1000000000000000019884624838656,
|
| 24 |
+
"padding_side": "left",
|
| 25 |
+
"remove_space": false,
|
| 26 |
+
"tokenizer_class": "ChatGLM4Tokenizer"
|
| 27 |
+
}
|