Add files using upload-large-folder tool
Browse files- .gitattributes +2 -0
- LICENSE +29 -0
- README.md +198 -0
- chat_template.jinja +255 -0
- config.json +595 -0
- generation_config.json +12 -0
- model-00001-of-00041.safetensors +3 -0
- model-00002-of-00041.safetensors +3 -0
- model-00003-of-00041.safetensors +3 -0
- model-00004-of-00041.safetensors +3 -0
- model-00005-of-00041.safetensors +3 -0
- model-00006-of-00041.safetensors +3 -0
- model-00007-of-00041.safetensors +3 -0
- model-00008-of-00041.safetensors +3 -0
- model-00009-of-00041.safetensors +3 -0
- model-00010-of-00041.safetensors +3 -0
- model-00011-of-00041.safetensors +3 -0
- model-00012-of-00041.safetensors +3 -0
- model-00013-of-00041.safetensors +3 -0
- model-00014-of-00041.safetensors +3 -0
- model-00015-of-00041.safetensors +3 -0
- model-00016-of-00041.safetensors +3 -0
- model-00017-of-00041.safetensors +3 -0
- model-00018-of-00041.safetensors +3 -0
- model-00019-of-00041.safetensors +3 -0
- model-00020-of-00041.safetensors +3 -0
- model-00021-of-00041.safetensors +3 -0
- model-00022-of-00041.safetensors +3 -0
- model-00023-of-00041.safetensors +3 -0
- model-00024-of-00041.safetensors +3 -0
- model-00025-of-00041.safetensors +3 -0
- model-00026-of-00041.safetensors +3 -0
- model-00027-of-00041.safetensors +3 -0
- model-00028-of-00041.safetensors +3 -0
- model-00029-of-00041.safetensors +3 -0
- model-00030-of-00041.safetensors +3 -0
- model-00031-of-00041.safetensors +3 -0
- model-00032-of-00041.safetensors +3 -0
- model-00033-of-00041.safetensors +3 -0
- model-00034-of-00041.safetensors +3 -0
- model-00035-of-00041.safetensors +3 -0
- model-00036-of-00041.safetensors +3 -0
- model-00037-of-00041.safetensors +3 -0
- model-00038-of-00041.safetensors +3 -0
- model-00039-of-00041.safetensors +3 -0
- model-00040-of-00041.safetensors +3 -0
- model-00041-of-00041.safetensors +3 -0
- model.safetensors.index.json +3 -0
- tokenizer.json +3 -0
- tokenizer_config.json +33 -0
.gitattributes
CHANGED
|
@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
|
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
| 36 |
+
tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 37 |
+
model.safetensors.index.json filter=lfs diff=lfs merge=lfs -text
|
LICENSE
ADDED
|
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
GLM-5.3 License
|
| 2 |
+
|
| 3 |
+
Copyright (c) 2026 Z.AI
|
| 4 |
+
|
| 5 |
+
Permission is hereby granted, free of charge, to any person or entity (the "Licensee") obtaining a copy of this software — including the model weights, parameters, configuration files, inference and training code, and associated documentation (collectively, the "Software") — to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software; to run, deploy, fine-tune, or otherwise modify the Software and create derivative works from it; and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
|
| 6 |
+
|
| 7 |
+
1. The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. The Licensee's use of the Software must comply with applicable laws and regulations.
|
| 8 |
+
|
| 9 |
+
2. "Model as a Service" means giving a third party access to language model inference or fine-tuning (e.g., via API) in a manner that allows such third party to exercise meaningful control over the inputs, parameters, or training data. This does not include (a) end-user products with model capabilities solely embedded within specific features or harnesses, or (b) mere relaying of requests to models hosted by others.
|
| 10 |
+
If the Licensee or any of its affiliates operates a Model as a Service business, and the aggregate revenue of the Licensee and its affiliates exceeds 10 billion US dollars (or the equivalent in other currencies) in total over any consecutive 12 months, the Licensee must pass Z.AI's security review before using the Software or its derivative works for any commercial purpose. The scope and method of the security review shall be reasonably determined by Z.AI.
|
| 11 |
+
|
| 12 |
+
3. THE SOFTWARE AND ANY OUTPUT AND RESULTS THEREFROM ARE PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL Z.AI OR ITS AFFILIATES OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
|
| 13 |
+
|
| 14 |
+
For any questions regarding this license, please contact glmlicense@z.ai.
|
| 15 |
+
|
| 16 |
+
-----
|
| 17 |
+
|
| 18 |
+
版权所有 (c) 2026 Z.AI
|
| 19 |
+
|
| 20 |
+
特此免费授予任何获得本软件副本的个人或实体("被许可方")——包括模型权重、参数、配置文件、推理和训练代码及相关文档(统称"软件")——不受限制地处理本软件的权利,包括但不限于:使用、复制、修改、合并、发布、分发、再许可和/或销售软件副本;运行、部署、微调或以其他方式修改软件并创建衍生作品;以及允许获得软件的其他人行使上述权利,但须遵守以下条件:
|
| 21 |
+
|
| 22 |
+
1. 上述版权声明和本许可声明应包含在软件的所有副本或实质性部分中。被许可方对软件的使用必须符合适用法律法规。
|
| 23 |
+
|
| 24 |
+
2. "模型即服务"指以允许第三方对输入、参数或训练数据行使实质性控制的方式,向第三方提供语言模型推理或微调服务(如通过API)。不包括:(a) 模型能力仅嵌入特定功能或框架中的终端用户产品,或(b) 单纯转发请求至他人托管的模型。
|
| 25 |
+
若被许可方或其关联方运营"模型即服务"业务,且被许可方及关联方在任意连续12个月内累计总收入超过100亿美元(或等值其他货币),则被许可方在使用软件或其衍生作品进行任何商业用途之前,须通过Z.AI的安全审查。安全审查的范围和方式由Z.AI合理确定。
|
| 26 |
+
|
| 27 |
+
3. 软件及其任何输出和结果均按"现状"提供,不附带任何形式的保证,无论是明示还是暗示,包括但不限于适销性、特定用途适用性和不侵权的保证。在任何情况下,Z.AI或其关联方或版权持有人均不对任何索赔、损害或其他责任承担责任,无论该责任是基于合同、侵权或其他方式,因软件或使用软件而产生或与之相关。
|
| 28 |
+
|
| 29 |
+
如对本许可有任何疑问,请联系 glmlicense@z.ai。
|
README.md
ADDED
|
@@ -0,0 +1,198 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: other
|
| 3 |
+
license_name: glm-5.3
|
| 4 |
+
license_link: LICENSE
|
| 5 |
+
base_model: zai-org/GLM-5.3-BF16
|
| 6 |
+
base_model_relation: quantized
|
| 7 |
+
pipeline_tag: text-generation
|
| 8 |
+
language: [en, zh]
|
| 9 |
+
tags: [glm, glm-5, moe, quantized, w4afp8, int4, fp8, awq, sglang]
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
# GLM-5.3-W4AFP8-AWQ
|
| 13 |
+
|
| 14 |
+
A W4AFP8 quantization of [GLM-5.3](https://huggingface.co/zai-org/GLM-5.3) by Bitdeer AI.
|
| 15 |
+
Routed experts are 4-bit integer (group size 128, AWQ-calibrated); attention, shared experts
|
| 16 |
+
and the dense layers are block FP8. The checkpoint is about 400 GB, so the full 753B-parameter
|
| 17 |
+
model serves on a single 8×H100 node, and it matches the published scores of the unquantized
|
| 18 |
+
model on the benchmarks below.
|
| 19 |
+
|
| 20 |
+
For most uses we recommend its sibling, [GLM-5.3-W4AFP8-EP-GPTQ](https://huggingface.co/BitdeerAI/GLM-5.3-W4AFP8-EP-GPTQ), scores the same and generates 5-18% fewer output tokens per answer.
|
| 21 |
+
|
| 22 |
+
## Model overview
|
| 23 |
+
|
| 24 |
+
| | |
|
| 25 |
+
|---|---|
|
| 26 |
+
| Base model | [zai-org/GLM-5.3-BF16](https://huggingface.co/zai-org/GLM-5.3-BF16) (revision `304b8051`) |
|
| 27 |
+
| Architecture | `GlmMoeDsaForCausalLM`, 78 layers + 1 MTP layer, 256 routed experts per MoE layer |
|
| 28 |
+
| Parameters | 753B |
|
| 29 |
+
| Quantization | W4AFP8: INT4 routed experts, FP8 activations and non-expert weights |
|
| 30 |
+
| Method | AWQ |
|
| 31 |
+
| Checkpoint size | 399.7 GB (including the 5.3 GB MTP draft layer) |
|
| 32 |
+
| Context length | 1,048,576 tokens (model maximum) |
|
| 33 |
+
| Runtime | SGLang ≥ 0.5.17, `--quantization w4afp8` |
|
| 34 |
+
|
| 35 |
+
## Quantization
|
| 36 |
+
|
| 37 |
+
Routed experts are quantized to INT4 with AWQ.
|
| 38 |
+
|
| 39 |
+
| Component | Precision |
|
| 40 |
+
|---|---|
|
| 41 |
+
| Routed experts, layers 3–77 | INT4, group size 128, AWQ |
|
| 42 |
+
| Attention (MLA) projections | Block FP8 (128×128) |
|
| 43 |
+
| DSA indexer (`wk`, `wq_b`) | Block FP8 (SGLang does not load it in BF16) |
|
| 44 |
+
| Shared experts; dense MLP layers 0–2 | Block FP8 |
|
| 45 |
+
| Router, `lm_head`, embeddings, norms | Unchanged (BF16) |
|
| 46 |
+
| MTP draft layer 78 | INT4 experts + FP8, round-to-nearest (not calibrated) |
|
| 47 |
+
|
| 48 |
+
- **Calibration data:** 256 samples × 2,048 tokens from
|
| 49 |
+
[UltraChat 200k](https://huggingface.co/datasets/HuggingFaceH4/ultrachat_200k), with all
|
| 50 |
+
256 experts calibrated in every layer.
|
| 51 |
+
- The tokenizer and generation config are unchanged from the base model. The chat template is
|
| 52 |
+
the current upstream GLM-5.3 template (see [Notes](#notes)).
|
| 53 |
+
|
| 54 |
+
## Evaluation
|
| 55 |
+
|
| 56 |
+
Three W4AFP8 checkpoints evaluated with the same harness, prompts and serving setup: our EP-GPTQ
|
| 57 |
+
and AWQ checkpoints, and [PhalaCloud/GLM-5.3-W4AFP8](https://huggingface.co/PhalaCloud/GLM-5.3-W4AFP8).
|
| 58 |
+
|
| 59 |
+
### Following Artificial Analysis' methodology
|
| 60 |
+
|
| 61 |
+
| Benchmark | Ours (EP-GPTQ) | **This model** | PhalaCloud | AA published (unquantized) |
|
| 62 |
+
|---|---:|---:|---:|---:|
|
| 63 |
+
| GPQA Diamond, pass@1 | 91.21 ± 0.88 | **91.52 ± 0.87** | 91.11 ± 0.85 | ~91.7 |
|
| 64 |
+
| AA-LCR v1.1, pass@1 | 78.33 ± 2.4 | **79.00 ± 2.4** | 80.00 ± 2.3 | 80.0 |
|
| 65 |
+
|
| 66 |
+
± is one standard error. These are our own reproductions of
|
| 67 |
+
[Artificial Analysis' published methodology](https://artificialanalysis.ai/methodology), not
|
| 68 |
+
official Artificial Analysis runs. GPQA Diamond: 198 questions × 5 repeats. AA-LCR: 100 questions
|
| 69 |
+
× 3 repeats, around 97k input tokens each, LLM-judged. On AA-LCR all three checkpoints answered
|
| 70 |
+
the same 300 items; paired McNemar tests find no difference between any pair (p = 0.55–0.87).
|
| 71 |
+
|
| 72 |
+
### Output tokens at equal accuracy
|
| 73 |
+
|
| 74 |
+
| Mean completion tokens per answer | Ours (EP-GPTQ) | **This model** | PhalaCloud |
|
| 75 |
+
|---|---:|---:|---:|
|
| 76 |
+
| GPQA Diamond | 13,944 | **14,757** | 16,332 |
|
| 77 |
+
| AA-LCR v1.1 | 5,203 | **6,363** | 5,656 |
|
| 78 |
+
|
| 79 |
+
Our EP-GPTQ sibling reaches the same scores with shorter reasoning: 5.5% and 18.2% fewer output tokens than this model. On a short-answer suite the difference disappears.
|
| 80 |
+
|
| 81 |
+
### lm-evaluation-harness
|
| 82 |
+
|
| 83 |
+
| Task | Ours (EP-GPTQ) | **This model** | PhalaCloud |
|
| 84 |
+
|---|---:|---:|---:|
|
| 85 |
+
| GSM8K | 97.57 ± 0.42 | **97.35 ± 0.44** | 96.82 ± 0.48 |
|
| 86 |
+
| IFEval | 88.91 ± 1.35 | **90.76 ± 1.25** | 90.20 ± 1.28 |
|
| 87 |
+
| MMLU | 86.84 ± 0.27 | **86.63 ± 0.28** | 86.81 ± 0.27 |
|
| 88 |
+
| ARC Challenge | 68.86 ± 1.35 | **69.37 ± 1.35** | 70.31 ± 1.34 |
|
| 89 |
+
| HellaSwag | 88.97 ± 0.31 | **89.20 ± 0.31** | 89.35 ± 0.31 |
|
| 90 |
+
| TruthfulQA MC2 | 61.77 ± 1.45 | **62.88 ± 1.46** | 62.54 ± 1.46 |
|
| 91 |
+
|
| 92 |
+
Full test sets, lm-evaluation-harness 0.4.10 against the SGLang endpoint. No gap between the
|
| 93 |
+
checkpoints reaches 2 points.
|
| 94 |
+
|
| 95 |
+
### Terminal-Bench
|
| 96 |
+
|
| 97 |
+
| Benchmark | **This model** | PhalaCloud | GLM-5.3 unquantized |
|
| 98 |
+
|---|---:|---:|---:|
|
| 99 |
+
| Terminal-Bench | **85.4 ± 2.3** | 82.4 ± 5.3 | 85.4 ± 3.4 |
|
| 100 |
+
|
| 101 |
+
Mean over 3 seeds; ± is the standard deviation across seeds, not a standard error. Terminal-Bench is
|
| 102 |
+
large, so only 3 seeds were run, and at this variance a 3-seed standard deviation is only a rough
|
| 103 |
+
guide. This model matched the unquantized model's mean.
|
| 104 |
+
|
| 105 |
+
All evaluations used SGLang 0.5.17, FP8 KV cache, temperature 1.0, top_p 0.95, and a
|
| 106 |
+
131,072-token output limit for the reasoning benchmarks.
|
| 107 |
+
|
| 108 |
+
## Usage
|
| 109 |
+
|
| 110 |
+
### Serving on one 8×H100 node
|
| 111 |
+
|
| 112 |
+
```bash
|
| 113 |
+
python -m sglang.launch_server \
|
| 114 |
+
--model-path BitdeerAI/GLM-5.3-W4AFP8-AWQ \
|
| 115 |
+
--quantization w4afp8 \
|
| 116 |
+
--tp 8 \
|
| 117 |
+
--kv-cache-dtype fp8_e4m3 \
|
| 118 |
+
--disable-shared-experts-fusion \
|
| 119 |
+
--context-length 164800 \
|
| 120 |
+
--mem-fraction-static 0.75 \
|
| 121 |
+
--chunked-prefill-size 2048 \
|
| 122 |
+
--reasoning-parser glm45 \
|
| 123 |
+
--tool-call-parser glm47 \
|
| 124 |
+
--trust-remote-code
|
| 125 |
+
```
|
| 126 |
+
|
| 127 |
+
On 8×H100-80GB this leaves room for a KV cache of about 165k tokens. For the full 1M-token
|
| 128 |
+
context, serve across two 8×H100 nodes (`--tp 16 --nnodes 2`; we measured a 598,848-token KV pool
|
| 129 |
+
at `--mem-fraction-static 0.80`), or use a node with more memory per GPU, such as 8×H200.
|
| 130 |
+
|
| 131 |
+
### Speculative decoding (MTP)
|
| 132 |
+
|
| 133 |
+
The checkpoint includes GLM-5.3's MTP draft layer. To enable EAGLE speculative decoding, add:
|
| 134 |
+
|
| 135 |
+
```bash
|
| 136 |
+
--speculative-algorithm EAGLE \
|
| 137 |
+
--speculative-num-steps 3 \
|
| 138 |
+
--speculative-eagle-topk 1 \
|
| 139 |
+
--speculative-num-draft-tokens 4
|
| 140 |
+
```
|
| 141 |
+
|
| 142 |
+
We validated these flags on the two-node serve. We have not measured the speedup.
|
| 143 |
+
|
| 144 |
+
### Sending requests
|
| 145 |
+
|
| 146 |
+
Use the base model's recommended sampling: temperature 1.0, top_p 0.95.
|
| 147 |
+
|
| 148 |
+
```python
|
| 149 |
+
from openai import OpenAI
|
| 150 |
+
|
| 151 |
+
client = OpenAI(base_url="http://localhost:30000/v1", api_key="EMPTY")
|
| 152 |
+
response = client.chat.completions.create(
|
| 153 |
+
model="BitdeerAI/GLM-5.3-W4AFP8-AWQ",
|
| 154 |
+
messages=[{"role": "user", "content": "Explain expert parallelism in two sentences."}],
|
| 155 |
+
temperature=1.0,
|
| 156 |
+
top_p=0.95,
|
| 157 |
+
max_tokens=32768,
|
| 158 |
+
)
|
| 159 |
+
print(response.choices[0].message.reasoning_content)
|
| 160 |
+
print(response.choices[0].message.content)
|
| 161 |
+
```
|
| 162 |
+
|
| 163 |
+
## Notes
|
| 164 |
+
|
| 165 |
+
- These are quantized weights of GLM-5.3 and inherit the base model's capabilities, limitations
|
| 166 |
+
and biases.
|
| 167 |
+
- The MTP draft layer is quantized with round-to-nearest, not calibrated. The main model verifies
|
| 168 |
+
every drafted token, so this affects speculative-decoding acceptance, not outputs.
|
| 169 |
+
- Benchmark results are our reproductions and may differ from official leaderboard figures.
|
| 170 |
+
- Our evaluations ran with the base model's chat template as of 2026-08-28. This repository ships
|
| 171 |
+
the current upstream template, which adds a fix for tool-calling conversations (tool-result
|
| 172 |
+
reordering and messages with empty content); plain chat prompts render identically.
|
| 173 |
+
|
| 174 |
+
## License
|
| 175 |
+
|
| 176 |
+
This model is a derivative of GLM-5.3 and is released under the same
|
| 177 |
+
[GLM-5.3 License](LICENSE). Copyright belongs to Z.ai; see `LICENSE` for the full terms.
|
| 178 |
+
|
| 179 |
+
## Citation
|
| 180 |
+
|
| 181 |
+
If you use this model, please cite GLM-5:
|
| 182 |
+
|
| 183 |
+
```bibtex
|
| 184 |
+
@misc{glm5team2026glm5vibecodingagentic,
|
| 185 |
+
title={GLM-5: from Vibe Coding to Agentic Engineering},
|
| 186 |
+
author={GLM-5-Team and : and Aohan Zeng and Xin Lv and Zhenyu Hou and Zhengxiao Du and Qinkai Zheng and Bin Chen and Da Yin and Chendi Ge and Chenghua Huang and Chengxing Xie and Chenzheng Zhu and Congfeng Yin and Cunxiang Wang and Gengzheng Pan and Hao Zeng and Haoke Zhang and Haoran Wang and Huilong Chen and Jiajie Zhang and Jian Jiao and Jiaqi Guo and Jingsen Wang and Jingzhao Du and Jinzhu Wu and Kedong Wang and Lei Li and Lin Fan and Lucen Zhong and Mingdao Liu and Mingming Zhao and Pengfan Du and Qian Dong and Rui Lu and Shuang-Li and Shulin Cao and Song Liu and Ting Jiang and Xiaodong Chen and Xiaohan Zhang and Xuancheng Huang and Xuezhen Dong and Yabo Xu and Yao Wei and Yifan An and Yilin Niu and Yitong Zhu and Yuanhao Wen and Yukuo Cen and Yushi Bai and Zhongpei Qiao and Zihan Wang and Zikang Wang and Zilin Zhu and Ziqiang Liu and Zixuan Li and Bojie Wang and Bosi Wen and Can Huang and Changpeng Cai and Chao Yu and Chen Li and Chengwei Hu and Chenhui Zhang and Dan Zhang and Daoyan Lin and Dayong Yang and Di Wang and Ding Ai and Erle Zhu and Fangzhou Yi and Feiyu Chen and Guohong Wen and Hailong Sun and Haisha Zhao and Haiyi Hu and Hanchen Zhang and Hanrui Liu and Hanyu Zhang and Hao Peng and Hao Tai and Haobo Zhang and He Liu and Hongwei Wang and Hongxi Yan and Hongyu Ge and Huan Liu and Huanpeng Chu and Jia'ni Zhao and Jiachen Wang and Jiajing Zhao and Jiamin Ren and Jiapeng Wang and Jiaxin Zhang and Jiayi Gui and Jiayue Zhao and Jijie Li and Jing An and Jing Li and Jingwei Yuan and Jinhua Du and Jinxin Liu and Junkai Zhi and Junwen Duan and Kaiyue Zhou and Kangjian Wei and Ke Wang and Keyun Luo and Laiqiang Zhang and Leigang Sha and Liang Xu and Lindong Wu and Lintao Ding and Lu Chen and Minghao Li and Nianyi Lin and Pan Ta and Qiang Zou and Rongjun Song and Ruiqi Yang and Shangqing Tu and Shangtong Yang and Shaoxiang Wu and Shengyan Zhang and Shijie Li and Shuang Li and Shuyi Fan and Wei Qin and Wei Tian and Weining Zhang and Wenbo Yu and Wenjie Liang and Xiang Kuang and Xiangmeng Cheng and Xiangyang Li and Xiaoquan Yan and Xiaowei Hu and Xiaoying Ling and Xing Fan and Xingye Xia and Xinyuan Zhang and Xinze Zhang and Xirui Pan and Xu Zou and Xunkai Zhang and Yadi Liu and Yandong Wu and Yanfu Li and Yidong Wang and Yifan Zhu and Yijun Tan and Yilin Zhou and Yiming Pan and Ying Zhang and Yinpei Su and Yipeng Geng and Yong Yan and Yonglin Tan and Yuean Bi and Yuhan Shen and Yuhao Yang and Yujiang Li and Yunan Liu and Yunqing Wang and Yuntao Li and Yurong Wu and Yutao Zhang and Yuxi Duan and Yuxuan Zhang and Zezhen Liu and Zhengtao Jiang and Zhenhe Yan and Zheyu Zhang and Zhixiang Wei and Zhuo Chen and Zhuoer Feng and Zijun Yao and Ziwei Chai and Ziyuan Wang and Zuzhou Zhang and Bin Xu and Minlie Huang and Hongning Wang and Juanzi Li and Yuxiao Dong and Jie Tang},
|
| 187 |
+
year={2026},
|
| 188 |
+
eprint={2602.15763},
|
| 189 |
+
archivePrefix={arXiv},
|
| 190 |
+
primaryClass={cs.LG},
|
| 191 |
+
url={https://arxiv.org/abs/2602.15763},
|
| 192 |
+
}
|
| 193 |
+
```
|
| 194 |
+
|
| 195 |
+
## Acknowledgements
|
| 196 |
+
|
| 197 |
+
Thanks to the GLM team at Z.ai for releasing GLM-5.3. Served with
|
| 198 |
+
[SGLang](https://github.com/sgl-project/sglang).
|
chat_template.jinja
ADDED
|
@@ -0,0 +1,255 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[gMASK]<sop>
|
| 2 |
+
{%- set effective_reasoning_effort = reasoning_effort if reasoning_effort is defined and reasoning_effort in ['low', 'high'] else 'max' -%}
|
| 3 |
+
{%- if effective_reasoning_effort is not none -%}<|system|>Reasoning Effort: {{ effective_reasoning_effort | capitalize }}{%- endif -%}
|
| 4 |
+
{%- set clear_thinking = clear_thinking if clear_thinking is defined else false -%}
|
| 5 |
+
{%- if tools -%}
|
| 6 |
+
{%- macro tool_to_json(tool) -%}
|
| 7 |
+
{%- set ns_tool = namespace(first=true) -%}
|
| 8 |
+
{{ '{' -}}
|
| 9 |
+
{%- for k, v in tool.items() -%}
|
| 10 |
+
{%- if k != 'defer_loading' and k != 'strict' -%}
|
| 11 |
+
{%- if not ns_tool.first -%}{{- ', ' -}}{%- endif -%}
|
| 12 |
+
{%- set ns_tool.first = false -%}
|
| 13 |
+
"{{ k }}": {{ v | tojson(ensure_ascii=False) }}
|
| 14 |
+
{%- endif -%}
|
| 15 |
+
{%- endfor -%}
|
| 16 |
+
{{- '}' -}}
|
| 17 |
+
{%- endmacro -%}
|
| 18 |
+
{%- macro tool_references_to_response(refs) -%}
|
| 19 |
+
{{- '<tool_response><tools>\n' -}}
|
| 20 |
+
{%- for tr in refs -%}
|
| 21 |
+
{%- for tool in tools -%}
|
| 22 |
+
{%- if 'function' in tool -%}
|
| 23 |
+
{%- set tool = tool['function'] -%}
|
| 24 |
+
{%- endif -%}
|
| 25 |
+
{%- if tool.name == tr.name -%}
|
| 26 |
+
{{- tool_to_json(tool) + '\n' -}}
|
| 27 |
+
{%- endif -%}
|
| 28 |
+
{%- endfor -%}
|
| 29 |
+
{%- endfor -%}
|
| 30 |
+
{{- '</tools></tool_response>' -}}
|
| 31 |
+
{%- endmacro -%}
|
| 32 |
+
<|system|>
|
| 33 |
+
# Tools
|
| 34 |
+
|
| 35 |
+
You may call one or more functions to assist with the user query.
|
| 36 |
+
|
| 37 |
+
You are provided with function signatures within <tools></tools> XML tags:
|
| 38 |
+
<tools>
|
| 39 |
+
{% for tool in tools %}
|
| 40 |
+
{%- if 'function' in tool -%}
|
| 41 |
+
{%- set tool = tool['function'] -%}
|
| 42 |
+
{%- endif -%}
|
| 43 |
+
{% if tool.defer_loading is not defined or not tool.defer_loading %}
|
| 44 |
+
{{ tool_to_json(tool) }}
|
| 45 |
+
{% endif %}
|
| 46 |
+
{% endfor %}
|
| 47 |
+
</tools>
|
| 48 |
+
|
| 49 |
+
For each function call, output the function name and arguments within the following XML format:
|
| 50 |
+
<tool_call>{function-name}<arg_key>{arg-key-1}</arg_key><arg_value>{arg-value-1}</arg_value><arg_key>{arg-key-2}</arg_key><arg_value>{arg-value-2}</arg_value>...</tool_call>{%- endif -%}
|
| 51 |
+
{%- macro visible_text(content) -%}
|
| 52 |
+
{%- if content is string -%}
|
| 53 |
+
{{- content }}
|
| 54 |
+
{%- elif content is iterable and content is not mapping -%}
|
| 55 |
+
{%- for item in content -%}
|
| 56 |
+
{%- if item is mapping and item.type == 'text' -%}
|
| 57 |
+
{{- item.text }}
|
| 58 |
+
{%- elif item is string -%}
|
| 59 |
+
{{- item }}
|
| 60 |
+
{%- elif item is mapping and item.type in ['image', 'image_url', 'video', 'video_url', 'audio', 'audio_url', 'input_audio'] -%}
|
| 61 |
+
{%- set media_type = item.type | replace('_url', '') | replace('input_', '') -%}
|
| 62 |
+
{{- "<reminder>You are unable to process this " ~ media_type ~ " because you don't have multi-modal input ability. Try different methods.</reminder>" }}
|
| 63 |
+
{%- endif -%}
|
| 64 |
+
{%- endfor -%}
|
| 65 |
+
{%- elif content is not none -%}
|
| 66 |
+
{{- content }}
|
| 67 |
+
{%- endif -%}
|
| 68 |
+
{%- endmacro -%}
|
| 69 |
+
{%- macro tool_response(text) -%}
|
| 70 |
+
{{- '<tool_response>' + text + '</tool_response>' -}}
|
| 71 |
+
{%- endmacro -%}
|
| 72 |
+
{%- macro render_tool_response(m) -%}
|
| 73 |
+
{%- if m.content is string -%}
|
| 74 |
+
{{- tool_response(m.content) -}}
|
| 75 |
+
{%- elif m.content and m.content is not mapping and m.content.0.type == "tool_reference" -%}
|
| 76 |
+
{{- tool_references_to_response(m.content) -}}
|
| 77 |
+
{%- elif is_list_of_outputs(m) -%}
|
| 78 |
+
{%- for tr in m.content -%}
|
| 79 |
+
{%- if tr.output is iterable and tr.output is not string and tr.output is not mapping and tr.output and tr.output.0.type == "tool_reference" -%}
|
| 80 |
+
{{- tool_references_to_response(tr.output) -}}
|
| 81 |
+
{%- else -%}
|
| 82 |
+
{{- tool_response(visible_text(tr.output)) -}}
|
| 83 |
+
{%- endif -%}
|
| 84 |
+
{%- endfor -%}
|
| 85 |
+
{%- else -%}
|
| 86 |
+
{{- tool_response(visible_text(m.content)) -}}
|
| 87 |
+
{%- endif -%}
|
| 88 |
+
{%- endmacro -%}
|
| 89 |
+
{%- macro id_of(obj) -%}
|
| 90 |
+
{%- if obj.tool_call_id -%}
|
| 91 |
+
{{- obj.tool_call_id -}}
|
| 92 |
+
{%- elif obj.id -%}
|
| 93 |
+
{{- obj.id -}}
|
| 94 |
+
{%- endif -%}
|
| 95 |
+
{%- endmacro -%}
|
| 96 |
+
{%- macro is_list_of_outputs(m) -%}
|
| 97 |
+
{%- if m.content and m.content.0.output is defined -%}1{%- endif -%}
|
| 98 |
+
{%- endmacro -%}
|
| 99 |
+
{%- macro has_dup_tool_result_id(lo, hi, target) -%}
|
| 100 |
+
{%- set ns_cnt = namespace(n=0) -%}
|
| 101 |
+
{%- for k in range(lo, hi + 1) -%}
|
| 102 |
+
{%- set m = messages[k] -%}
|
| 103 |
+
{%- if is_list_of_outputs(m) -%}
|
| 104 |
+
{%- for entry in m.content -%}
|
| 105 |
+
{%- if id_of(entry) == target -%}
|
| 106 |
+
{%- set ns_cnt.n = ns_cnt.n + 1 -%}
|
| 107 |
+
{%- endif -%}
|
| 108 |
+
{%- endfor -%}
|
| 109 |
+
{%- elif id_of(m) == target -%}
|
| 110 |
+
{%- set ns_cnt.n = ns_cnt.n + 1 -%}
|
| 111 |
+
{%- endif -%}
|
| 112 |
+
{%- if ns_cnt.n > 1 -%}{%- break -%}{%- endif -%}
|
| 113 |
+
{%- endfor -%}
|
| 114 |
+
{%- if ns_cnt.n > 1 -%}1{%- endif -%}
|
| 115 |
+
{%- endmacro -%}
|
| 116 |
+
{%- macro tc_id_exists(tcs, target) -%}
|
| 117 |
+
{%- set ns_f = namespace(found=false) -%}
|
| 118 |
+
{%- for tc in tcs -%}
|
| 119 |
+
{%- if id_of(tc) == target -%}
|
| 120 |
+
{%- set ns_f.found = true -%}
|
| 121 |
+
{%- break -%}
|
| 122 |
+
{%- endif -%}
|
| 123 |
+
{%- endfor -%}
|
| 124 |
+
{%- if ns_f.found -%}1{%- endif -%}
|
| 125 |
+
{%- endmacro -%}
|
| 126 |
+
{%- set ns = namespace(last_user_index=-1) -%}
|
| 127 |
+
{%- for m in messages %}
|
| 128 |
+
{%- if m.role == 'user' %}
|
| 129 |
+
{%- set ns.last_user_index = loop.index0 -%}
|
| 130 |
+
{%- endif %}
|
| 131 |
+
{%- endfor %}
|
| 132 |
+
{%- for m in messages -%}
|
| 133 |
+
{%- if m.role == 'user' -%}<|user|>{{ visible_text(m.content) }}
|
| 134 |
+
{%- elif m.role == 'assistant' -%}
|
| 135 |
+
<|assistant|>
|
| 136 |
+
{%- set content = visible_text(m.content) %}
|
| 137 |
+
{%- if m.reasoning_content is string %}
|
| 138 |
+
{%- set reasoning_content = m.reasoning_content %}
|
| 139 |
+
{%- elif '</think>' in content %}
|
| 140 |
+
{%- set reasoning_content = content.split('</think>')[0].split('<think>')[-1] %}
|
| 141 |
+
{%- set content = content.split('</think>')[-1] %}
|
| 142 |
+
{%- endif %}
|
| 143 |
+
{%- if (not clear_thinking or loop.index0 > ns.last_user_index) and reasoning_content is defined -%}
|
| 144 |
+
{{ '<think>' + reasoning_content + '</think>'}}
|
| 145 |
+
{%- else -%}
|
| 146 |
+
{{ '<think></think>' }}
|
| 147 |
+
{%- endif -%}
|
| 148 |
+
{%- if content.strip() -%}
|
| 149 |
+
{{ content.strip() }}
|
| 150 |
+
{%- endif -%}
|
| 151 |
+
{% if m.tool_calls %}
|
| 152 |
+
{% for tc in m.tool_calls %}
|
| 153 |
+
{%- if tc.function %}
|
| 154 |
+
{%- set tc = tc.function %}
|
| 155 |
+
{%- endif %}
|
| 156 |
+
{{- '<tool_call>' + tc.name -}}
|
| 157 |
+
{% set _args = tc.arguments %}{% for k, v in _args.items() %}<arg_key>{{ k }}</arg_key><arg_value>{{ v | tojson(ensure_ascii=False) if v is not string else v }}</arg_value>{% endfor %}</tool_call>{% endfor %}
|
| 158 |
+
{% endif %}
|
| 159 |
+
{%- elif m.role == 'tool' -%}
|
| 160 |
+
{%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
|
| 161 |
+
{{- '<|observation|>' -}}
|
| 162 |
+
{%- set block_start = loop.index0 -%}
|
| 163 |
+
{%- set ns_blk = namespace(end=block_start) -%}
|
| 164 |
+
{%- for j in range(block_start, messages|length) -%}
|
| 165 |
+
{%- if messages[j].role == 'tool' -%}
|
| 166 |
+
{%- set ns_blk.end = j -%}
|
| 167 |
+
{%- else -%}
|
| 168 |
+
{%- break -%}
|
| 169 |
+
{%- endif -%}
|
| 170 |
+
{%- endfor -%}
|
| 171 |
+
{%- set ns_a = namespace(tool_calls=none) -%}
|
| 172 |
+
{%- if block_start > 0 and messages[block_start - 1].role == 'assistant' and messages[block_start - 1].tool_calls -%}
|
| 173 |
+
{%- set ns_a.tool_calls = messages[block_start - 1].tool_calls -%}
|
| 174 |
+
{%- endif -%}
|
| 175 |
+
{%- set ns_chk = namespace(can_sort=true) -%}
|
| 176 |
+
{%- if not ns_a.tool_calls -%}
|
| 177 |
+
{%- set ns_chk.can_sort = false -%}
|
| 178 |
+
{%- else -%}
|
| 179 |
+
{%- for k in range(block_start, ns_blk.end + 1) -%}
|
| 180 |
+
{%- if not ns_chk.can_sort -%}{%- break -%}{%- endif -%}
|
| 181 |
+
{%- set m = messages[k] -%}
|
| 182 |
+
{%- if is_list_of_outputs(m) -%}
|
| 183 |
+
{%- for entry in m.content -%}
|
| 184 |
+
{%- if not ns_chk.can_sort -%}{%- break -%}{%- endif -%}
|
| 185 |
+
{%- set eid = id_of(entry) -%}
|
| 186 |
+
{%- if not eid -%}
|
| 187 |
+
{%- set ns_chk.can_sort = false -%}
|
| 188 |
+
{%- elif has_dup_tool_result_id(block_start, ns_blk.end, eid) -%}
|
| 189 |
+
{%- set ns_chk.can_sort = false -%}
|
| 190 |
+
{%- elif not tc_id_exists(ns_a.tool_calls, eid) -%}
|
| 191 |
+
{%- set ns_chk.can_sort = false -%}
|
| 192 |
+
{%- endif -%}
|
| 193 |
+
{%- endfor -%}
|
| 194 |
+
{%- else -%}
|
| 195 |
+
{%- set tk_id = id_of(m) -%}
|
| 196 |
+
{%- if not tk_id -%}
|
| 197 |
+
{%- set ns_chk.can_sort = false -%}
|
| 198 |
+
{%- elif has_dup_tool_result_id(block_start, ns_blk.end, tk_id) -%}
|
| 199 |
+
{%- set ns_chk.can_sort = false -%}
|
| 200 |
+
{%- elif not tc_id_exists(ns_a.tool_calls, tk_id) -%}
|
| 201 |
+
{%- set ns_chk.can_sort = false -%}
|
| 202 |
+
{%- endif -%}
|
| 203 |
+
{%- endif -%}
|
| 204 |
+
{%- endfor -%}
|
| 205 |
+
{%- for i in range(ns_a.tool_calls | length) -%}
|
| 206 |
+
{%- if not ns_chk.can_sort -%}{%- break -%}{%- endif -%}
|
| 207 |
+
{%- set tc_id = id_of(ns_a.tool_calls[i]) -%}
|
| 208 |
+
{%- if not tc_id -%}
|
| 209 |
+
{%- set ns_chk.can_sort = false -%}
|
| 210 |
+
{%- endif -%}
|
| 211 |
+
{%- for j in range(i + 1, ns_a.tool_calls | length) -%}
|
| 212 |
+
{%- if id_of(ns_a.tool_calls[j]) == tc_id -%}
|
| 213 |
+
{%- set ns_chk.can_sort = false -%}
|
| 214 |
+
{%- break -%}
|
| 215 |
+
{%- endif -%}
|
| 216 |
+
{%- endfor -%}
|
| 217 |
+
{%- endfor -%}
|
| 218 |
+
{%- endif -%}
|
| 219 |
+
{%- if ns_chk.can_sort -%}
|
| 220 |
+
{%- for tc in ns_a.tool_calls -%}
|
| 221 |
+
{%- set tc_id = id_of(tc) -%}
|
| 222 |
+
{%- for k in range(block_start, ns_blk.end + 1) -%}
|
| 223 |
+
{%- set m = messages[k] -%}
|
| 224 |
+
{%- if is_list_of_outputs(m) -%}
|
| 225 |
+
{%- for entry in m.content -%}
|
| 226 |
+
{%- set eid = id_of(entry) -%}
|
| 227 |
+
{%- if eid == tc_id -%}
|
| 228 |
+
{%- if entry.output is iterable and entry.output is not string and entry.output is not mapping and entry.output and entry.output.0.type == "tool_reference" -%}
|
| 229 |
+
{{- tool_references_to_response(entry.output) -}}
|
| 230 |
+
{%- else -%}
|
| 231 |
+
{{- tool_response(visible_text(entry.output)) -}}
|
| 232 |
+
{%- endif -%}
|
| 233 |
+
{%- endif -%}
|
| 234 |
+
{%- endfor -%}
|
| 235 |
+
{%- else -%}
|
| 236 |
+
{%- set tk_id = id_of(m) -%}
|
| 237 |
+
{%- if tk_id == tc_id -%}
|
| 238 |
+
{{- render_tool_response(m) -}}
|
| 239 |
+
{%- endif -%}
|
| 240 |
+
{%- endif -%}
|
| 241 |
+
{%- endfor -%}
|
| 242 |
+
{%- endfor -%}
|
| 243 |
+
{%- else -%}
|
| 244 |
+
{%- for k in range(block_start, ns_blk.end + 1) -%}
|
| 245 |
+
{{- render_tool_response(messages[k]) -}}
|
| 246 |
+
{%- endfor -%}
|
| 247 |
+
{%- endif -%}
|
| 248 |
+
{% endif -%}
|
| 249 |
+
{%- elif m.role == 'system' -%}
|
| 250 |
+
<|system|>{{ visible_text(m.content) }}
|
| 251 |
+
{%- endif -%}
|
| 252 |
+
{%- endfor -%}
|
| 253 |
+
{%- if add_generation_prompt -%}
|
| 254 |
+
<|assistant|>{{- '<think>' -}}
|
| 255 |
+
{%- endif -%}
|
config.json
ADDED
|
@@ -0,0 +1,595 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"GlmMoeDsaForCausalLM"
|
| 4 |
+
],
|
| 5 |
+
"attention_bias": false,
|
| 6 |
+
"attention_dropout": 0.0,
|
| 7 |
+
"bos_token_id": 0,
|
| 8 |
+
"dtype": "bfloat16",
|
| 9 |
+
"eos_token_id": [
|
| 10 |
+
154820,
|
| 11 |
+
154827,
|
| 12 |
+
154829
|
| 13 |
+
],
|
| 14 |
+
"ep_size": 1,
|
| 15 |
+
"first_k_dense_replace": 3,
|
| 16 |
+
"head_dim": 64,
|
| 17 |
+
"hidden_act": "silu",
|
| 18 |
+
"hidden_size": 6144,
|
| 19 |
+
"index_head_dim": 128,
|
| 20 |
+
"index_n_heads": 32,
|
| 21 |
+
"index_share_for_mtp_iteration": true,
|
| 22 |
+
"index_skip_topk_offset": 3,
|
| 23 |
+
"index_topk": 2048,
|
| 24 |
+
"index_topk_freq": 4,
|
| 25 |
+
"index_topk_pattern": null,
|
| 26 |
+
"indexer_rope_interleave": true,
|
| 27 |
+
"indexer_types": [
|
| 28 |
+
"full",
|
| 29 |
+
"full",
|
| 30 |
+
"full",
|
| 31 |
+
"shared",
|
| 32 |
+
"shared",
|
| 33 |
+
"shared",
|
| 34 |
+
"full",
|
| 35 |
+
"shared",
|
| 36 |
+
"shared",
|
| 37 |
+
"shared",
|
| 38 |
+
"full",
|
| 39 |
+
"shared",
|
| 40 |
+
"shared",
|
| 41 |
+
"shared",
|
| 42 |
+
"full",
|
| 43 |
+
"shared",
|
| 44 |
+
"shared",
|
| 45 |
+
"shared",
|
| 46 |
+
"full",
|
| 47 |
+
"shared",
|
| 48 |
+
"shared",
|
| 49 |
+
"shared",
|
| 50 |
+
"full",
|
| 51 |
+
"shared",
|
| 52 |
+
"shared",
|
| 53 |
+
"shared",
|
| 54 |
+
"full",
|
| 55 |
+
"shared",
|
| 56 |
+
"shared",
|
| 57 |
+
"shared",
|
| 58 |
+
"full",
|
| 59 |
+
"shared",
|
| 60 |
+
"shared",
|
| 61 |
+
"shared",
|
| 62 |
+
"full",
|
| 63 |
+
"shared",
|
| 64 |
+
"shared",
|
| 65 |
+
"shared",
|
| 66 |
+
"full",
|
| 67 |
+
"shared",
|
| 68 |
+
"shared",
|
| 69 |
+
"shared",
|
| 70 |
+
"full",
|
| 71 |
+
"shared",
|
| 72 |
+
"shared",
|
| 73 |
+
"shared",
|
| 74 |
+
"full",
|
| 75 |
+
"shared",
|
| 76 |
+
"shared",
|
| 77 |
+
"shared",
|
| 78 |
+
"full",
|
| 79 |
+
"shared",
|
| 80 |
+
"shared",
|
| 81 |
+
"shared",
|
| 82 |
+
"full",
|
| 83 |
+
"shared",
|
| 84 |
+
"shared",
|
| 85 |
+
"shared",
|
| 86 |
+
"full",
|
| 87 |
+
"shared",
|
| 88 |
+
"shared",
|
| 89 |
+
"shared",
|
| 90 |
+
"full",
|
| 91 |
+
"shared",
|
| 92 |
+
"shared",
|
| 93 |
+
"shared",
|
| 94 |
+
"full",
|
| 95 |
+
"shared",
|
| 96 |
+
"shared",
|
| 97 |
+
"shared",
|
| 98 |
+
"full",
|
| 99 |
+
"shared",
|
| 100 |
+
"shared",
|
| 101 |
+
"shared",
|
| 102 |
+
"full",
|
| 103 |
+
"shared",
|
| 104 |
+
"shared",
|
| 105 |
+
"shared"
|
| 106 |
+
],
|
| 107 |
+
"initializer_range": 0.02,
|
| 108 |
+
"intermediate_size": 12288,
|
| 109 |
+
"kv_lora_rank": 512,
|
| 110 |
+
"layer_types": [
|
| 111 |
+
"deepseek_sparse_attention",
|
| 112 |
+
"deepseek_sparse_attention",
|
| 113 |
+
"deepseek_sparse_attention",
|
| 114 |
+
"deepseek_sparse_attention",
|
| 115 |
+
"deepseek_sparse_attention",
|
| 116 |
+
"deepseek_sparse_attention",
|
| 117 |
+
"deepseek_sparse_attention",
|
| 118 |
+
"deepseek_sparse_attention",
|
| 119 |
+
"deepseek_sparse_attention",
|
| 120 |
+
"deepseek_sparse_attention",
|
| 121 |
+
"deepseek_sparse_attention",
|
| 122 |
+
"deepseek_sparse_attention",
|
| 123 |
+
"deepseek_sparse_attention",
|
| 124 |
+
"deepseek_sparse_attention",
|
| 125 |
+
"deepseek_sparse_attention",
|
| 126 |
+
"deepseek_sparse_attention",
|
| 127 |
+
"deepseek_sparse_attention",
|
| 128 |
+
"deepseek_sparse_attention",
|
| 129 |
+
"deepseek_sparse_attention",
|
| 130 |
+
"deepseek_sparse_attention",
|
| 131 |
+
"deepseek_sparse_attention",
|
| 132 |
+
"deepseek_sparse_attention",
|
| 133 |
+
"deepseek_sparse_attention",
|
| 134 |
+
"deepseek_sparse_attention",
|
| 135 |
+
"deepseek_sparse_attention",
|
| 136 |
+
"deepseek_sparse_attention",
|
| 137 |
+
"deepseek_sparse_attention",
|
| 138 |
+
"deepseek_sparse_attention",
|
| 139 |
+
"deepseek_sparse_attention",
|
| 140 |
+
"deepseek_sparse_attention",
|
| 141 |
+
"deepseek_sparse_attention",
|
| 142 |
+
"deepseek_sparse_attention",
|
| 143 |
+
"deepseek_sparse_attention",
|
| 144 |
+
"deepseek_sparse_attention",
|
| 145 |
+
"deepseek_sparse_attention",
|
| 146 |
+
"deepseek_sparse_attention",
|
| 147 |
+
"deepseek_sparse_attention",
|
| 148 |
+
"deepseek_sparse_attention",
|
| 149 |
+
"deepseek_sparse_attention",
|
| 150 |
+
"deepseek_sparse_attention",
|
| 151 |
+
"deepseek_sparse_attention",
|
| 152 |
+
"deepseek_sparse_attention",
|
| 153 |
+
"deepseek_sparse_attention",
|
| 154 |
+
"deepseek_sparse_attention",
|
| 155 |
+
"deepseek_sparse_attention",
|
| 156 |
+
"deepseek_sparse_attention",
|
| 157 |
+
"deepseek_sparse_attention",
|
| 158 |
+
"deepseek_sparse_attention",
|
| 159 |
+
"deepseek_sparse_attention",
|
| 160 |
+
"deepseek_sparse_attention",
|
| 161 |
+
"deepseek_sparse_attention",
|
| 162 |
+
"deepseek_sparse_attention",
|
| 163 |
+
"deepseek_sparse_attention",
|
| 164 |
+
"deepseek_sparse_attention",
|
| 165 |
+
"deepseek_sparse_attention",
|
| 166 |
+
"deepseek_sparse_attention",
|
| 167 |
+
"deepseek_sparse_attention",
|
| 168 |
+
"deepseek_sparse_attention",
|
| 169 |
+
"deepseek_sparse_attention",
|
| 170 |
+
"deepseek_sparse_attention",
|
| 171 |
+
"deepseek_sparse_attention",
|
| 172 |
+
"deepseek_sparse_attention",
|
| 173 |
+
"deepseek_sparse_attention",
|
| 174 |
+
"deepseek_sparse_attention",
|
| 175 |
+
"deepseek_sparse_attention",
|
| 176 |
+
"deepseek_sparse_attention",
|
| 177 |
+
"deepseek_sparse_attention",
|
| 178 |
+
"deepseek_sparse_attention",
|
| 179 |
+
"deepseek_sparse_attention",
|
| 180 |
+
"deepseek_sparse_attention",
|
| 181 |
+
"deepseek_sparse_attention",
|
| 182 |
+
"deepseek_sparse_attention",
|
| 183 |
+
"deepseek_sparse_attention",
|
| 184 |
+
"deepseek_sparse_attention",
|
| 185 |
+
"deepseek_sparse_attention",
|
| 186 |
+
"deepseek_sparse_attention",
|
| 187 |
+
"deepseek_sparse_attention",
|
| 188 |
+
"deepseek_sparse_attention"
|
| 189 |
+
],
|
| 190 |
+
"max_position_embeddings": 1048576,
|
| 191 |
+
"mlp_bias": false,
|
| 192 |
+
"mlp_layer_types": [
|
| 193 |
+
"dense",
|
| 194 |
+
"dense",
|
| 195 |
+
"dense",
|
| 196 |
+
"sparse",
|
| 197 |
+
"sparse",
|
| 198 |
+
"sparse",
|
| 199 |
+
"sparse",
|
| 200 |
+
"sparse",
|
| 201 |
+
"sparse",
|
| 202 |
+
"sparse",
|
| 203 |
+
"sparse",
|
| 204 |
+
"sparse",
|
| 205 |
+
"sparse",
|
| 206 |
+
"sparse",
|
| 207 |
+
"sparse",
|
| 208 |
+
"sparse",
|
| 209 |
+
"sparse",
|
| 210 |
+
"sparse",
|
| 211 |
+
"sparse",
|
| 212 |
+
"sparse",
|
| 213 |
+
"sparse",
|
| 214 |
+
"sparse",
|
| 215 |
+
"sparse",
|
| 216 |
+
"sparse",
|
| 217 |
+
"sparse",
|
| 218 |
+
"sparse",
|
| 219 |
+
"sparse",
|
| 220 |
+
"sparse",
|
| 221 |
+
"sparse",
|
| 222 |
+
"sparse",
|
| 223 |
+
"sparse",
|
| 224 |
+
"sparse",
|
| 225 |
+
"sparse",
|
| 226 |
+
"sparse",
|
| 227 |
+
"sparse",
|
| 228 |
+
"sparse",
|
| 229 |
+
"sparse",
|
| 230 |
+
"sparse",
|
| 231 |
+
"sparse",
|
| 232 |
+
"sparse",
|
| 233 |
+
"sparse",
|
| 234 |
+
"sparse",
|
| 235 |
+
"sparse",
|
| 236 |
+
"sparse",
|
| 237 |
+
"sparse",
|
| 238 |
+
"sparse",
|
| 239 |
+
"sparse",
|
| 240 |
+
"sparse",
|
| 241 |
+
"sparse",
|
| 242 |
+
"sparse",
|
| 243 |
+
"sparse",
|
| 244 |
+
"sparse",
|
| 245 |
+
"sparse",
|
| 246 |
+
"sparse",
|
| 247 |
+
"sparse",
|
| 248 |
+
"sparse",
|
| 249 |
+
"sparse",
|
| 250 |
+
"sparse",
|
| 251 |
+
"sparse",
|
| 252 |
+
"sparse",
|
| 253 |
+
"sparse",
|
| 254 |
+
"sparse",
|
| 255 |
+
"sparse",
|
| 256 |
+
"sparse",
|
| 257 |
+
"sparse",
|
| 258 |
+
"sparse",
|
| 259 |
+
"sparse",
|
| 260 |
+
"sparse",
|
| 261 |
+
"sparse",
|
| 262 |
+
"sparse",
|
| 263 |
+
"sparse",
|
| 264 |
+
"sparse",
|
| 265 |
+
"sparse",
|
| 266 |
+
"sparse",
|
| 267 |
+
"sparse",
|
| 268 |
+
"sparse",
|
| 269 |
+
"sparse",
|
| 270 |
+
"sparse"
|
| 271 |
+
],
|
| 272 |
+
"model_type": "glm_moe_dsa",
|
| 273 |
+
"moe_intermediate_size": 2048,
|
| 274 |
+
"moe_layer_freq": 1,
|
| 275 |
+
"moe_router_dtype": "float32",
|
| 276 |
+
"n_group": 1,
|
| 277 |
+
"n_routed_experts": 256,
|
| 278 |
+
"n_shared_experts": 1,
|
| 279 |
+
"norm_topk_prob": true,
|
| 280 |
+
"num_attention_heads": 64,
|
| 281 |
+
"num_experts_per_tok": 8,
|
| 282 |
+
"num_hidden_layers": 78,
|
| 283 |
+
"num_key_value_heads": 64,
|
| 284 |
+
"num_nextn_predict_layers": 1,
|
| 285 |
+
"pad_token_id": 154820,
|
| 286 |
+
"pretraining_tp": 1,
|
| 287 |
+
"q_lora_rank": 2048,
|
| 288 |
+
"qk_head_dim": 256,
|
| 289 |
+
"qk_nope_head_dim": 192,
|
| 290 |
+
"qk_rope_head_dim": 64,
|
| 291 |
+
"quantization_config": {
|
| 292 |
+
"quant_method": "w4afp8",
|
| 293 |
+
"group_size": 128,
|
| 294 |
+
"ignored_layers": [
|
| 295 |
+
"model.layers.0.self_attn.indexer.weights_proj",
|
| 296 |
+
"model.layers.1.self_attn.indexer.weights_proj",
|
| 297 |
+
"model.layers.2.self_attn.indexer.weights_proj",
|
| 298 |
+
"model.layers.3.mlp.gate",
|
| 299 |
+
"model.layers.4.mlp.gate",
|
| 300 |
+
"model.layers.5.mlp.gate",
|
| 301 |
+
"model.layers.6.self_attn.indexer.weights_proj",
|
| 302 |
+
"model.layers.6.mlp.gate",
|
| 303 |
+
"model.layers.7.mlp.gate",
|
| 304 |
+
"model.layers.8.mlp.gate",
|
| 305 |
+
"model.layers.9.mlp.gate",
|
| 306 |
+
"model.layers.10.self_attn.indexer.weights_proj",
|
| 307 |
+
"model.layers.10.mlp.gate",
|
| 308 |
+
"model.layers.11.mlp.gate",
|
| 309 |
+
"model.layers.12.mlp.gate",
|
| 310 |
+
"model.layers.13.mlp.gate",
|
| 311 |
+
"model.layers.14.self_attn.indexer.weights_proj",
|
| 312 |
+
"model.layers.14.mlp.gate",
|
| 313 |
+
"model.layers.15.mlp.gate",
|
| 314 |
+
"model.layers.16.mlp.gate",
|
| 315 |
+
"model.layers.17.mlp.gate",
|
| 316 |
+
"model.layers.18.self_attn.indexer.weights_proj",
|
| 317 |
+
"model.layers.18.mlp.gate",
|
| 318 |
+
"model.layers.19.mlp.gate",
|
| 319 |
+
"model.layers.20.mlp.gate",
|
| 320 |
+
"model.layers.21.mlp.gate",
|
| 321 |
+
"model.layers.22.self_attn.indexer.weights_proj",
|
| 322 |
+
"model.layers.22.mlp.gate",
|
| 323 |
+
"model.layers.23.mlp.gate",
|
| 324 |
+
"model.layers.24.mlp.gate",
|
| 325 |
+
"model.layers.25.mlp.gate",
|
| 326 |
+
"model.layers.26.self_attn.indexer.weights_proj",
|
| 327 |
+
"model.layers.26.mlp.gate",
|
| 328 |
+
"model.layers.27.mlp.gate",
|
| 329 |
+
"model.layers.28.mlp.gate",
|
| 330 |
+
"model.layers.29.mlp.gate",
|
| 331 |
+
"model.layers.30.self_attn.indexer.weights_proj",
|
| 332 |
+
"model.layers.30.mlp.gate",
|
| 333 |
+
"model.layers.31.mlp.gate",
|
| 334 |
+
"model.layers.32.mlp.gate",
|
| 335 |
+
"model.layers.33.mlp.gate",
|
| 336 |
+
"model.layers.34.self_attn.indexer.weights_proj",
|
| 337 |
+
"model.layers.34.mlp.gate",
|
| 338 |
+
"model.layers.35.mlp.gate",
|
| 339 |
+
"model.layers.36.mlp.gate",
|
| 340 |
+
"model.layers.37.mlp.gate",
|
| 341 |
+
"model.layers.38.self_attn.indexer.weights_proj",
|
| 342 |
+
"model.layers.38.mlp.gate",
|
| 343 |
+
"model.layers.39.mlp.gate",
|
| 344 |
+
"model.layers.40.mlp.gate",
|
| 345 |
+
"model.layers.41.mlp.gate",
|
| 346 |
+
"model.layers.42.self_attn.indexer.weights_proj",
|
| 347 |
+
"model.layers.42.mlp.gate",
|
| 348 |
+
"model.layers.43.mlp.gate",
|
| 349 |
+
"model.layers.44.mlp.gate",
|
| 350 |
+
"model.layers.45.mlp.gate",
|
| 351 |
+
"model.layers.46.self_attn.indexer.weights_proj",
|
| 352 |
+
"model.layers.46.mlp.gate",
|
| 353 |
+
"model.layers.47.mlp.gate",
|
| 354 |
+
"model.layers.48.mlp.gate",
|
| 355 |
+
"model.layers.49.mlp.gate",
|
| 356 |
+
"model.layers.50.self_attn.indexer.weights_proj",
|
| 357 |
+
"model.layers.50.mlp.gate",
|
| 358 |
+
"model.layers.51.mlp.gate",
|
| 359 |
+
"model.layers.52.mlp.gate",
|
| 360 |
+
"model.layers.53.mlp.gate",
|
| 361 |
+
"model.layers.54.self_attn.indexer.weights_proj",
|
| 362 |
+
"model.layers.54.mlp.gate",
|
| 363 |
+
"model.layers.55.mlp.gate",
|
| 364 |
+
"model.layers.56.mlp.gate",
|
| 365 |
+
"model.layers.57.mlp.gate",
|
| 366 |
+
"model.layers.58.self_attn.indexer.weights_proj",
|
| 367 |
+
"model.layers.58.mlp.gate",
|
| 368 |
+
"model.layers.59.mlp.gate",
|
| 369 |
+
"model.layers.60.mlp.gate",
|
| 370 |
+
"model.layers.61.mlp.gate",
|
| 371 |
+
"model.layers.62.self_attn.indexer.weights_proj",
|
| 372 |
+
"model.layers.62.mlp.gate",
|
| 373 |
+
"model.layers.63.mlp.gate",
|
| 374 |
+
"model.layers.64.mlp.gate",
|
| 375 |
+
"model.layers.65.mlp.gate",
|
| 376 |
+
"model.layers.66.self_attn.indexer.weights_proj",
|
| 377 |
+
"model.layers.66.mlp.gate",
|
| 378 |
+
"model.layers.67.mlp.gate",
|
| 379 |
+
"model.layers.68.mlp.gate",
|
| 380 |
+
"model.layers.69.mlp.gate",
|
| 381 |
+
"model.layers.70.self_attn.indexer.weights_proj",
|
| 382 |
+
"model.layers.70.mlp.gate",
|
| 383 |
+
"model.layers.71.mlp.gate",
|
| 384 |
+
"model.layers.72.mlp.gate",
|
| 385 |
+
"model.layers.73.mlp.gate",
|
| 386 |
+
"model.layers.74.self_attn.indexer.weights_proj",
|
| 387 |
+
"model.layers.74.mlp.gate",
|
| 388 |
+
"model.layers.75.mlp.gate",
|
| 389 |
+
"model.layers.76.mlp.gate",
|
| 390 |
+
"model.layers.77.mlp.gate",
|
| 391 |
+
"lm_head",
|
| 392 |
+
"model.layers.0.self_attn.indexer.k_norm",
|
| 393 |
+
"model.layers.0.self_attn.kv_a_layernorm",
|
| 394 |
+
"model.layers.0.self_attn.q_a_layernorm",
|
| 395 |
+
"model.layers.1.self_attn.indexer.k_norm",
|
| 396 |
+
"model.layers.1.self_attn.kv_a_layernorm",
|
| 397 |
+
"model.layers.1.self_attn.q_a_layernorm",
|
| 398 |
+
"model.layers.10.self_attn.indexer.k_norm",
|
| 399 |
+
"model.layers.10.self_attn.kv_a_layernorm",
|
| 400 |
+
"model.layers.10.self_attn.q_a_layernorm",
|
| 401 |
+
"model.layers.11.self_attn.kv_a_layernorm",
|
| 402 |
+
"model.layers.11.self_attn.q_a_layernorm",
|
| 403 |
+
"model.layers.12.self_attn.kv_a_layernorm",
|
| 404 |
+
"model.layers.12.self_attn.q_a_layernorm",
|
| 405 |
+
"model.layers.13.self_attn.kv_a_layernorm",
|
| 406 |
+
"model.layers.13.self_attn.q_a_layernorm",
|
| 407 |
+
"model.layers.14.self_attn.indexer.k_norm",
|
| 408 |
+
"model.layers.14.self_attn.kv_a_layernorm",
|
| 409 |
+
"model.layers.14.self_attn.q_a_layernorm",
|
| 410 |
+
"model.layers.15.self_attn.kv_a_layernorm",
|
| 411 |
+
"model.layers.15.self_attn.q_a_layernorm",
|
| 412 |
+
"model.layers.16.self_attn.kv_a_layernorm",
|
| 413 |
+
"model.layers.16.self_attn.q_a_layernorm",
|
| 414 |
+
"model.layers.17.self_attn.kv_a_layernorm",
|
| 415 |
+
"model.layers.17.self_attn.q_a_layernorm",
|
| 416 |
+
"model.layers.18.self_attn.indexer.k_norm",
|
| 417 |
+
"model.layers.18.self_attn.kv_a_layernorm",
|
| 418 |
+
"model.layers.18.self_attn.q_a_layernorm",
|
| 419 |
+
"model.layers.19.self_attn.kv_a_layernorm",
|
| 420 |
+
"model.layers.19.self_attn.q_a_layernorm",
|
| 421 |
+
"model.layers.2.self_attn.indexer.k_norm",
|
| 422 |
+
"model.layers.2.self_attn.kv_a_layernorm",
|
| 423 |
+
"model.layers.2.self_attn.q_a_layernorm",
|
| 424 |
+
"model.layers.20.self_attn.kv_a_layernorm",
|
| 425 |
+
"model.layers.20.self_attn.q_a_layernorm",
|
| 426 |
+
"model.layers.21.self_attn.kv_a_layernorm",
|
| 427 |
+
"model.layers.21.self_attn.q_a_layernorm",
|
| 428 |
+
"model.layers.22.self_attn.indexer.k_norm",
|
| 429 |
+
"model.layers.22.self_attn.kv_a_layernorm",
|
| 430 |
+
"model.layers.22.self_attn.q_a_layernorm",
|
| 431 |
+
"model.layers.23.self_attn.kv_a_layernorm",
|
| 432 |
+
"model.layers.23.self_attn.q_a_layernorm",
|
| 433 |
+
"model.layers.24.self_attn.kv_a_layernorm",
|
| 434 |
+
"model.layers.24.self_attn.q_a_layernorm",
|
| 435 |
+
"model.layers.25.self_attn.kv_a_layernorm",
|
| 436 |
+
"model.layers.25.self_attn.q_a_layernorm",
|
| 437 |
+
"model.layers.26.self_attn.indexer.k_norm",
|
| 438 |
+
"model.layers.26.self_attn.kv_a_layernorm",
|
| 439 |
+
"model.layers.26.self_attn.q_a_layernorm",
|
| 440 |
+
"model.layers.27.self_attn.kv_a_layernorm",
|
| 441 |
+
"model.layers.27.self_attn.q_a_layernorm",
|
| 442 |
+
"model.layers.28.self_attn.kv_a_layernorm",
|
| 443 |
+
"model.layers.28.self_attn.q_a_layernorm",
|
| 444 |
+
"model.layers.29.self_attn.kv_a_layernorm",
|
| 445 |
+
"model.layers.29.self_attn.q_a_layernorm",
|
| 446 |
+
"model.layers.3.self_attn.kv_a_layernorm",
|
| 447 |
+
"model.layers.3.self_attn.q_a_layernorm",
|
| 448 |
+
"model.layers.30.self_attn.indexer.k_norm",
|
| 449 |
+
"model.layers.30.self_attn.kv_a_layernorm",
|
| 450 |
+
"model.layers.30.self_attn.q_a_layernorm",
|
| 451 |
+
"model.layers.31.self_attn.kv_a_layernorm",
|
| 452 |
+
"model.layers.31.self_attn.q_a_layernorm",
|
| 453 |
+
"model.layers.32.self_attn.kv_a_layernorm",
|
| 454 |
+
"model.layers.32.self_attn.q_a_layernorm",
|
| 455 |
+
"model.layers.33.self_attn.kv_a_layernorm",
|
| 456 |
+
"model.layers.33.self_attn.q_a_layernorm",
|
| 457 |
+
"model.layers.34.self_attn.indexer.k_norm",
|
| 458 |
+
"model.layers.34.self_attn.kv_a_layernorm",
|
| 459 |
+
"model.layers.34.self_attn.q_a_layernorm",
|
| 460 |
+
"model.layers.35.self_attn.kv_a_layernorm",
|
| 461 |
+
"model.layers.35.self_attn.q_a_layernorm",
|
| 462 |
+
"model.layers.36.self_attn.kv_a_layernorm",
|
| 463 |
+
"model.layers.36.self_attn.q_a_layernorm",
|
| 464 |
+
"model.layers.37.self_attn.kv_a_layernorm",
|
| 465 |
+
"model.layers.37.self_attn.q_a_layernorm",
|
| 466 |
+
"model.layers.38.self_attn.indexer.k_norm",
|
| 467 |
+
"model.layers.38.self_attn.kv_a_layernorm",
|
| 468 |
+
"model.layers.38.self_attn.q_a_layernorm",
|
| 469 |
+
"model.layers.39.self_attn.kv_a_layernorm",
|
| 470 |
+
"model.layers.39.self_attn.q_a_layernorm",
|
| 471 |
+
"model.layers.4.self_attn.kv_a_layernorm",
|
| 472 |
+
"model.layers.4.self_attn.q_a_layernorm",
|
| 473 |
+
"model.layers.40.self_attn.kv_a_layernorm",
|
| 474 |
+
"model.layers.40.self_attn.q_a_layernorm",
|
| 475 |
+
"model.layers.41.self_attn.kv_a_layernorm",
|
| 476 |
+
"model.layers.41.self_attn.q_a_layernorm",
|
| 477 |
+
"model.layers.42.self_attn.indexer.k_norm",
|
| 478 |
+
"model.layers.42.self_attn.kv_a_layernorm",
|
| 479 |
+
"model.layers.42.self_attn.q_a_layernorm",
|
| 480 |
+
"model.layers.43.self_attn.kv_a_layernorm",
|
| 481 |
+
"model.layers.43.self_attn.q_a_layernorm",
|
| 482 |
+
"model.layers.44.self_attn.kv_a_layernorm",
|
| 483 |
+
"model.layers.44.self_attn.q_a_layernorm",
|
| 484 |
+
"model.layers.45.self_attn.kv_a_layernorm",
|
| 485 |
+
"model.layers.45.self_attn.q_a_layernorm",
|
| 486 |
+
"model.layers.46.self_attn.indexer.k_norm",
|
| 487 |
+
"model.layers.46.self_attn.kv_a_layernorm",
|
| 488 |
+
"model.layers.46.self_attn.q_a_layernorm",
|
| 489 |
+
"model.layers.47.self_attn.kv_a_layernorm",
|
| 490 |
+
"model.layers.47.self_attn.q_a_layernorm",
|
| 491 |
+
"model.layers.48.self_attn.kv_a_layernorm",
|
| 492 |
+
"model.layers.48.self_attn.q_a_layernorm",
|
| 493 |
+
"model.layers.49.self_attn.kv_a_layernorm",
|
| 494 |
+
"model.layers.49.self_attn.q_a_layernorm",
|
| 495 |
+
"model.layers.5.self_attn.kv_a_layernorm",
|
| 496 |
+
"model.layers.5.self_attn.q_a_layernorm",
|
| 497 |
+
"model.layers.50.self_attn.indexer.k_norm",
|
| 498 |
+
"model.layers.50.self_attn.kv_a_layernorm",
|
| 499 |
+
"model.layers.50.self_attn.q_a_layernorm",
|
| 500 |
+
"model.layers.51.self_attn.kv_a_layernorm",
|
| 501 |
+
"model.layers.51.self_attn.q_a_layernorm",
|
| 502 |
+
"model.layers.52.self_attn.kv_a_layernorm",
|
| 503 |
+
"model.layers.52.self_attn.q_a_layernorm",
|
| 504 |
+
"model.layers.53.self_attn.kv_a_layernorm",
|
| 505 |
+
"model.layers.53.self_attn.q_a_layernorm",
|
| 506 |
+
"model.layers.54.self_attn.indexer.k_norm",
|
| 507 |
+
"model.layers.54.self_attn.kv_a_layernorm",
|
| 508 |
+
"model.layers.54.self_attn.q_a_layernorm",
|
| 509 |
+
"model.layers.55.self_attn.kv_a_layernorm",
|
| 510 |
+
"model.layers.55.self_attn.q_a_layernorm",
|
| 511 |
+
"model.layers.56.self_attn.kv_a_layernorm",
|
| 512 |
+
"model.layers.56.self_attn.q_a_layernorm",
|
| 513 |
+
"model.layers.57.self_attn.kv_a_layernorm",
|
| 514 |
+
"model.layers.57.self_attn.q_a_layernorm",
|
| 515 |
+
"model.layers.58.self_attn.indexer.k_norm",
|
| 516 |
+
"model.layers.58.self_attn.kv_a_layernorm",
|
| 517 |
+
"model.layers.58.self_attn.q_a_layernorm",
|
| 518 |
+
"model.layers.59.self_attn.kv_a_layernorm",
|
| 519 |
+
"model.layers.59.self_attn.q_a_layernorm",
|
| 520 |
+
"model.layers.6.self_attn.indexer.k_norm",
|
| 521 |
+
"model.layers.6.self_attn.kv_a_layernorm",
|
| 522 |
+
"model.layers.6.self_attn.q_a_layernorm",
|
| 523 |
+
"model.layers.60.self_attn.kv_a_layernorm",
|
| 524 |
+
"model.layers.60.self_attn.q_a_layernorm",
|
| 525 |
+
"model.layers.61.self_attn.kv_a_layernorm",
|
| 526 |
+
"model.layers.61.self_attn.q_a_layernorm",
|
| 527 |
+
"model.layers.62.self_attn.indexer.k_norm",
|
| 528 |
+
"model.layers.62.self_attn.kv_a_layernorm",
|
| 529 |
+
"model.layers.62.self_attn.q_a_layernorm",
|
| 530 |
+
"model.layers.63.self_attn.kv_a_layernorm",
|
| 531 |
+
"model.layers.63.self_attn.q_a_layernorm",
|
| 532 |
+
"model.layers.64.self_attn.kv_a_layernorm",
|
| 533 |
+
"model.layers.64.self_attn.q_a_layernorm",
|
| 534 |
+
"model.layers.65.self_attn.kv_a_layernorm",
|
| 535 |
+
"model.layers.65.self_attn.q_a_layernorm",
|
| 536 |
+
"model.layers.66.self_attn.indexer.k_norm",
|
| 537 |
+
"model.layers.66.self_attn.kv_a_layernorm",
|
| 538 |
+
"model.layers.66.self_attn.q_a_layernorm",
|
| 539 |
+
"model.layers.67.self_attn.kv_a_layernorm",
|
| 540 |
+
"model.layers.67.self_attn.q_a_layernorm",
|
| 541 |
+
"model.layers.68.self_attn.kv_a_layernorm",
|
| 542 |
+
"model.layers.68.self_attn.q_a_layernorm",
|
| 543 |
+
"model.layers.69.self_attn.kv_a_layernorm",
|
| 544 |
+
"model.layers.69.self_attn.q_a_layernorm",
|
| 545 |
+
"model.layers.7.self_attn.kv_a_layernorm",
|
| 546 |
+
"model.layers.7.self_attn.q_a_layernorm",
|
| 547 |
+
"model.layers.70.self_attn.indexer.k_norm",
|
| 548 |
+
"model.layers.70.self_attn.kv_a_layernorm",
|
| 549 |
+
"model.layers.70.self_attn.q_a_layernorm",
|
| 550 |
+
"model.layers.71.self_attn.kv_a_layernorm",
|
| 551 |
+
"model.layers.71.self_attn.q_a_layernorm",
|
| 552 |
+
"model.layers.72.self_attn.kv_a_layernorm",
|
| 553 |
+
"model.layers.72.self_attn.q_a_layernorm",
|
| 554 |
+
"model.layers.73.self_attn.kv_a_layernorm",
|
| 555 |
+
"model.layers.73.self_attn.q_a_layernorm",
|
| 556 |
+
"model.layers.74.self_attn.indexer.k_norm",
|
| 557 |
+
"model.layers.74.self_attn.kv_a_layernorm",
|
| 558 |
+
"model.layers.74.self_attn.q_a_layernorm",
|
| 559 |
+
"model.layers.75.self_attn.kv_a_layernorm",
|
| 560 |
+
"model.layers.75.self_attn.q_a_layernorm",
|
| 561 |
+
"model.layers.76.self_attn.kv_a_layernorm",
|
| 562 |
+
"model.layers.76.self_attn.q_a_layernorm",
|
| 563 |
+
"model.layers.77.self_attn.kv_a_layernorm",
|
| 564 |
+
"model.layers.77.self_attn.q_a_layernorm",
|
| 565 |
+
"model.layers.8.self_attn.kv_a_layernorm",
|
| 566 |
+
"model.layers.8.self_attn.q_a_layernorm",
|
| 567 |
+
"model.layers.9.self_attn.kv_a_layernorm",
|
| 568 |
+
"model.layers.9.self_attn.q_a_layernorm",
|
| 569 |
+
"model.layers.0.input_layernorm",
|
| 570 |
+
"model.layers.0.post_attention_layernorm",
|
| 571 |
+
"model.layers.1.input_layernorm",
|
| 572 |
+
"model.layers.1.post_attention_layernorm",
|
| 573 |
+
"model.layers.2.input_layernorm",
|
| 574 |
+
"model.layers.2.post_attention_layernorm",
|
| 575 |
+
"model.layers.78.eh_proj",
|
| 576 |
+
"model.layers.78.mlp.gate",
|
| 577 |
+
"model.layers.78.self_attn.indexer.weights_proj"
|
| 578 |
+
]
|
| 579 |
+
},
|
| 580 |
+
"rms_norm_eps": 1e-05,
|
| 581 |
+
"rope_interleave": true,
|
| 582 |
+
"rope_parameters": {
|
| 583 |
+
"rope_theta": 8000000,
|
| 584 |
+
"rope_type": "default"
|
| 585 |
+
},
|
| 586 |
+
"routed_scaling_factor": 2.5,
|
| 587 |
+
"scoring_func": "sigmoid",
|
| 588 |
+
"tie_word_embeddings": false,
|
| 589 |
+
"topk_group": 1,
|
| 590 |
+
"topk_method": "noaux_tc",
|
| 591 |
+
"transformers_version": "5.14.1",
|
| 592 |
+
"use_cache": true,
|
| 593 |
+
"v_head_dim": 256,
|
| 594 |
+
"vocab_size": 154880
|
| 595 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"eos_token_id": [
|
| 4 |
+
154820,
|
| 5 |
+
154827,
|
| 6 |
+
154829
|
| 7 |
+
],
|
| 8 |
+
"pad_token_id": 154820,
|
| 9 |
+
"temperature": 1.0,
|
| 10 |
+
"top_p": 0.95,
|
| 11 |
+
"transformers_version": "5.12.0"
|
| 12 |
+
}
|
model-00001-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:34a1bf1c4615cf38c10616e0207aca407de9761300990bdb10f07c10c6193628
|
| 3 |
+
size 9973453440
|
model-00002-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:47236458df34066c60c913eefba6c01eed4a3d1237d4de830fd6d842f6cd8bdd
|
| 3 |
+
size 10001905908
|
model-00003-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ec5bd8c530876c393c07a1869215e5b80b88ecf747636c84b8e219cbe5fb4f8e
|
| 3 |
+
size 9992012398
|
model-00004-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:77054f2f8b7cf5f5f18263e4905fcd0a658288894e5d4ca575edda48128e7311
|
| 3 |
+
size 10001905868
|
model-00005-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:435986cedee8d75fd3fc0e511d30fe9a9622e10ec48328e690cffe263f8685ea
|
| 3 |
+
size 9992016246
|
model-00006-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6871200b9f89129a6231ff7e040db196bff3f8d094c694f6776ad8f013ac05f6
|
| 3 |
+
size 10001910172
|
model-00007-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e8e28958001d6331b5e1da6fa45d7a344dcbf7c4a5bb031c26e2beb8f7e65a7c
|
| 3 |
+
size 9992016686
|
model-00008-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0cd24cb2ddfbdb4d1954e411ffe51609ad569f52db5f5f0563975519b7c1e1e9
|
| 3 |
+
size 10001910164
|
model-00009-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4ef3f4134d7df9e4ebd962c762659b51ff57f9d4bea4733a431603f628d92b03
|
| 3 |
+
size 9992016710
|
model-00010-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:911da2200774c544c645a47ea36063ff4d5211dd12e3831aa63723fc5651a38d
|
| 3 |
+
size 10001910172
|
model-00011-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:79a743ac0e67eafdfaba9a2d1ae5514806801f22a77788b2ef044217b564dbdb
|
| 3 |
+
size 9992016630
|
model-00012-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fbd6fe3d4ad051b2b485fc86e5b42503af14bf4785e5d2157fdc6c7ed9f87ed2
|
| 3 |
+
size 10001910100
|
model-00013-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:de7b645395a48a67a87439201f917792b7e2839f2e434d75f0b42471b64b795e
|
| 3 |
+
size 9992016638
|
model-00014-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bc0bf3a6939215bbaa11412981df64b097bac4d84a921ed8768d9bb9cb28f940
|
| 3 |
+
size 10001335716
|
model-00015-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dbfe809c2fd7cf4eef43d22717c72e7ff19e195ea5bfd5f48b37c9e673eb3398
|
| 3 |
+
size 10004117692
|
model-00016-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d9050a0f7c3d153f9404ab6a1ff4d76a22f49cffab4548025d6e311843a8c22e
|
| 3 |
+
size 9992016926
|
model-00017-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f4f1a6672e0afc7fa38ad9cbc32ac19c2201a6db08cf928576c59a6071580dc0
|
| 3 |
+
size 10001910340
|
model-00018-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:30b03d1fd07a8ad28e83f8851f9815192f79d4215d04da1fd3e1fab5112aff45
|
| 3 |
+
size 9992016878
|
model-00019-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9d271691692b94c56b6b6b8dd6e99769852dfe44b59c07d377511d7298fd78d7
|
| 3 |
+
size 10001910252
|
model-00020-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8e83a8e3e44de28dc2f9ddb09f54cf55963e50a7e9013dfe55116be91d2b2bbc
|
| 3 |
+
size 9992016694
|
model-00021-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:400d28cd64726ba80bc0f22cbfa0d369727d806420fefacc58a263ce33b77385
|
| 3 |
+
size 10001910164
|
model-00022-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e0996de1777b6af8677bb0c524ca90fb0aba62e799aba3f4094f7ccb07710abc
|
| 3 |
+
size 9992016710
|
model-00023-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b909b819a069a7caa0e90dc7e07d65ca9921d33cca3bedb0ca9996ccc5580dd9
|
| 3 |
+
size 10001910164
|
model-00024-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4c2c09d615ece885f1cb26223ecdf2325b51c324f122854dc56c18013f4e3c75
|
| 3 |
+
size 9992016638
|
model-00025-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bcffa76ee34044e198c035ab3a9ce23c6a1063eaff061927e085a7708b7076f9
|
| 3 |
+
size 10001910108
|
model-00026-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c7b6b1c6d994defae73e81732b1f50c9caa0de61e78047b16a5a1b0d26278e15
|
| 3 |
+
size 9992016622
|
model-00027-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d428b3aaf9ee24f6f672ddb7beea2667d71ad59e7301293f864b2491c779cdba
|
| 3 |
+
size 10001910108
|
model-00028-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8aebd8f274bbbf6e5476492957e77ae1de1ab379c27eccebd2d122c2f45aec83
|
| 3 |
+
size 9993649742
|
model-00029-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5da71348392087555c99a52067111bdfe29798a59a62827e3736c44e8f47afbf
|
| 3 |
+
size 10001910396
|
model-00030-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:aeaee1fb7ca4c3d91b517a1c98e78ff44bb92e678307cea00ccfe36df8695983
|
| 3 |
+
size 9992016862
|
model-00031-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8fbb528c38f8974593e46811575babbc337cfe80b3ea3ad4d0a8fdbdd6928d52
|
| 3 |
+
size 10001910340
|
model-00032-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:31c72cb5f80bf3f490281aa4e79a66f341d0c47860da9cc3674c3cfff5f33cf7
|
| 3 |
+
size 9992016862
|
model-00033-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8b5b7bdc579c22783cd739f3cf5511478cc5f6f502d0bfabaa16a928b4ae3ab9
|
| 3 |
+
size 10001910164
|
model-00034-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:945763bdeda8dd368750d069a49c2b5b8eb835a7d7552446bbfa99f6c797c440
|
| 3 |
+
size 9992016686
|
model-00035-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:afa12b0d23a7dfc173a72789be4eb38332df4024ea61ff0a3694ca6e85b5e250
|
| 3 |
+
size 10001910164
|
model-00036-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b703d1b8619183fb53313f3663bdd3a9ea5efef1ef11d01f4df4dfc903bbe83e
|
| 3 |
+
size 9992016710
|
model-00037-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4f2e595acc741b0d79d7c4dc5a5db6365f5725928d4cace3b4280e68abd01b3f
|
| 3 |
+
size 10001910164
|
model-00038-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cb42d2906b21a6d641fc9119078a8b02d823a11512265bf5ea19e84fe9266fb9
|
| 3 |
+
size 9992016638
|
model-00039-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8a793ca1fd0b232f06c8f74f0b6ed8cab6137a52de6d117e08b1c59dda13e092
|
| 3 |
+
size 10001910108
|
model-00040-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5dce546672378d49dc07e1202862a9bbaa308c7dbb9bfc7d60ee97a78da02af9
|
| 3 |
+
size 4495039668
|
model-00041-of-00041.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:688db6c3d922e74018eaa2ab0b41855a28c6d5a303c29aecce398b6fcfbb1fbf
|
| 3 |
+
size 5349725032
|
model.safetensors.index.json
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7bf935e27669908e3721101ee92447439e69adf12878ece9001fef0373071831
|
| 3 |
+
size 16951579
|
tokenizer.json
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:19e773648cb4e65de8660ea6365e10acca112d42a854923df93db4a6f333a82d
|
| 3 |
+
size 20217442
|
tokenizer_config.json
ADDED
|
@@ -0,0 +1,33 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"backend": "tokenizers",
|
| 3 |
+
"clean_up_tokenization_spaces": false,
|
| 4 |
+
"do_lower_case": false,
|
| 5 |
+
"eos_token": "<|endoftext|>",
|
| 6 |
+
"extra_special_tokens": [
|
| 7 |
+
"<|endoftext|>",
|
| 8 |
+
"[MASK]",
|
| 9 |
+
"[gMASK]",
|
| 10 |
+
"[sMASK]",
|
| 11 |
+
"<sop>",
|
| 12 |
+
"<eop>",
|
| 13 |
+
"<|system|>",
|
| 14 |
+
"<|user|>",
|
| 15 |
+
"<|assistant|>",
|
| 16 |
+
"<|observation|>",
|
| 17 |
+
"<|begin_of_image|>",
|
| 18 |
+
"<|end_of_image|>",
|
| 19 |
+
"<|begin_of_video|>",
|
| 20 |
+
"<|end_of_video|>",
|
| 21 |
+
"<|begin_of_audio|>",
|
| 22 |
+
"<|end_of_audio|>",
|
| 23 |
+
"<|begin_of_transcription|>",
|
| 24 |
+
"<|end_of_transcription|>"
|
| 25 |
+
],
|
| 26 |
+
"is_local": true,
|
| 27 |
+
"model_max_length": 1048576,
|
| 28 |
+
"model_specific_special_tokens": {},
|
| 29 |
+
"pad_token": "<|endoftext|>",
|
| 30 |
+
"padding_side": "left",
|
| 31 |
+
"remove_space": false,
|
| 32 |
+
"tokenizer_class": "TokenizersBackend"
|
| 33 |
+
}
|