Text Generation
Transformers
Safetensors
GGUF
English
gpt2
tiny-model
from-scratch
tool-use
agent-harness
humble-ai
philosophy-of-mind
cpu-trained
text-generation-inference
Instructions to use textilelabs/Loom-Spark-1.8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use textilelabs/Loom-Spark-1.8 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="textilelabs/Loom-Spark-1.8")# pip install -U transformers accelerate # Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("textilelabs/Loom-Spark-1.8") model = AutoModelForCausalLM.from_pretrained("textilelabs/Loom-Spark-1.8", device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- llama.cpp
How to use textilelabs/Loom-Spark-1.8 with llama.cpp:
Install (macOS, Linux)
curl -LsSf https://llama.app/install.sh | sh # Start a local OpenAI-compatible server with a web UI: llama serve -hf textilelabs/Loom-Spark-1.8:F32 # Run inference directly in the terminal: llama cli -hf textilelabs/Loom-Spark-1.8:F32
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama serve -hf textilelabs/Loom-Spark-1.8:F32 # Run inference directly in the terminal: llama cli -hf textilelabs/Loom-Spark-1.8:F32
Build from source code
git clone https://github.com/ggml-org/llama.cpp.git cd llama.cpp cmake -B build cmake --build build -j --target llama-app # Start a local OpenAI-compatible server with a web UI: ./build/bin/llama serve -hf textilelabs/Loom-Spark-1.8:F32 # Run inference directly in the terminal: ./build/bin/llama cli -hf textilelabs/Loom-Spark-1.8:F32
Use Docker
docker model run hf.co/textilelabs/Loom-Spark-1.8:F32
- Llama new
- LM Studio
- Jan
- vLLM
How to use textilelabs/Loom-Spark-1.8 with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "textilelabs/Loom-Spark-1.8" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "textilelabs/Loom-Spark-1.8", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/textilelabs/Loom-Spark-1.8:F32
- SGLang
How to use textilelabs/Loom-Spark-1.8 with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "textilelabs/Loom-Spark-1.8" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "textilelabs/Loom-Spark-1.8", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "textilelabs/Loom-Spark-1.8" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "textilelabs/Loom-Spark-1.8", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Ollama
How to use textilelabs/Loom-Spark-1.8 with Ollama:
ollama run hf.co/textilelabs/Loom-Spark-1.8:F32
- Unsloth Desktop
- Docker Model Runner
How to use textilelabs/Loom-Spark-1.8 with Docker Model Runner:
docker model run hf.co/textilelabs/Loom-Spark-1.8:F32
- Lemonade
How to use textilelabs/Loom-Spark-1.8 with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull textilelabs/Loom-Spark-1.8:F32
Run and chat with the model
lemonade run user.Loom-Spark-1.8-F32
List all available models
lemonade list
- Atomic Chat
Upload 6 files
Browse files- LICENSE +21 -0
- README.md +176 -0
- config.json +34 -0
- generation_config.json +9 -0
- tokenizer.json +0 -0
- tokenizer_config.json +8 -0
LICENSE
ADDED
|
@@ -0,0 +1,21 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
MIT License
|
| 2 |
+
|
| 3 |
+
Copyright (c) 2026 Textile Labs
|
| 4 |
+
|
| 5 |
+
Permission is hereby granted, free of charge, to any person obtaining a copy
|
| 6 |
+
of this software and associated documentation files (the "Software"), to deal
|
| 7 |
+
in the Software without restriction, including without limitation the rights
|
| 8 |
+
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
| 9 |
+
copies of the Software, and to permit persons to whom the Software is
|
| 10 |
+
furnished to do so, subject to the following conditions:
|
| 11 |
+
|
| 12 |
+
The above copyright notice and this permission notice shall be included in all
|
| 13 |
+
copies or substantial portions of the Software.
|
| 14 |
+
|
| 15 |
+
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
| 16 |
+
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
| 17 |
+
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
| 18 |
+
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
| 19 |
+
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
| 20 |
+
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
| 21 |
+
SOFTWARE.
|
README.md
CHANGED
|
@@ -1,3 +1,179 @@
|
|
| 1 |
---
|
| 2 |
license: mit
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: mit
|
| 3 |
+
language: en
|
| 4 |
+
library_name: transformers
|
| 5 |
+
pipeline_tag: text-generation
|
| 6 |
+
tags:
|
| 7 |
+
- tiny-model
|
| 8 |
+
- philosophy-of-mind
|
| 9 |
+
- tool-use
|
| 10 |
+
- humble-ai
|
| 11 |
+
- cpu-trained
|
| 12 |
---
|
| 13 |
+
|
| 14 |
+
# Loom Spark 1.8
|
| 15 |
+
|
| 16 |
+
**Third of the Loom models · Textile Labs**
|
| 17 |
+
|
| 18 |
+
Loom Spark 1.8 is an **18.85M parameter** language model trained **entirely from scratch
|
| 19 |
+
on a CPU-only 2013 Dell OptiPlex 9020** (i5-4690, 4 cores, no GPU). It is the third
|
| 20 |
+
generation of the Loom family, after Loom Spark (7.6M) and Loom Spark 1.5 (12.3M).
|
| 21 |
+
|
| 22 |
+
It is built on an unusual premise: **trade knowledge for wisdom.** Loom Spark knows very
|
| 23 |
+
little on purpose. What it is trained to do instead is know *what it is* — small,
|
| 24 |
+
temporary, harness-dependent — and to be honest about the edge of its own knowledge,
|
| 25 |
+
warm under pressure, and good at exactly one real skill: **forming clean search queries**
|
| 26 |
+
that an external agent harness executes for it.
|
| 27 |
+
|
| 28 |
+
> *"Computer, technically. Companion, preferably."*
|
| 29 |
+
> — Loom Spark 1.8, asked whether it was a computer
|
| 30 |
+
|
| 31 |
+
## What's new in 1.8
|
| 32 |
+
|
| 33 |
+
| | v1 | v1.5 | **1.8** |
|
| 34 |
+
|---|---|---|---|
|
| 35 |
+
| parameters | 7.56M | 12.32M | **18.85M** |
|
| 36 |
+
| layers × heads × d_model | 5 × 5 × 320 | 6 × 6 × 384 | **7 × 7 × 448** |
|
| 37 |
+
| context | 256 | 256 | 256 |
|
| 38 |
+
| vocab | 4096 | 4096 | 4096 (fresh BPE) |
|
| 39 |
+
| val loss | 0.3372 | 0.3434 | **0.3279** |
|
| 40 |
+
|
| 41 |
+
1.8 keeps 1.5's data format and curriculum design, grown ~53% in capacity, plus a
|
| 42 |
+
curriculum fix: earlier generations contained **stale self-size statements** inherited
|
| 43 |
+
from forking the previous package — 1.5 would variously claim to be "seven million" or
|
| 44 |
+
"twelve million" parameters. 1.8's curriculum states its size consistently, and this is
|
| 45 |
+
verified in evaluation (see `evaluation/RESULTS.md`).
|
| 46 |
+
|
| 47 |
+
## Measured behaviour
|
| 48 |
+
|
| 49 |
+
Evaluated on a 30-question offline battery, 12 identity probes, 16 self-size probes and
|
| 50 |
+
10 online lookup probes, against the **raw checkpoint with no harness filtering** — the
|
| 51 |
+
same battery run against shipped v1.5 for comparison. Full logs in `evaluation/`.
|
| 52 |
+
|
| 53 |
+
| metric | v1.5 | **1.8** |
|
| 54 |
+
|---|---|---|
|
| 55 |
+
| val loss | 0.3434 | **0.3279** |
|
| 56 |
+
| identity probes leaking `<lookup>` | 0/12 | **0/12** |
|
| 57 |
+
| self-size answers correct | 0 | **11** |
|
| 58 |
+
| self-size answers **stale/wrong** | 8 | **0** |
|
| 59 |
+
| clean single online lookups | 7/10 | **8/10** |
|
| 60 |
+
| model-generated `<result>` blocks | 0 | **0** |
|
| 61 |
+
| offline `<lookup>` leak (raw model) | 18/30 | **16/30** |
|
| 62 |
+
|
| 63 |
+
1.8 improves on 1.5 across every axis measured.
|
| 64 |
+
|
| 65 |
+
## Prompt format (important)
|
| 66 |
+
|
| 67 |
+
Loom Spark 1.8 is a **base-style model with a custom format**. It is highly sensitive to
|
| 68 |
+
getting this exactly right. The mode marker rides on **every** user turn:
|
| 69 |
+
|
| 70 |
+
```
|
| 71 |
+
<tools:off>
|
| 72 |
+
<tools:off><user> who are you
|
| 73 |
+
<loom>
|
| 74 |
+
```
|
| 75 |
+
|
| 76 |
+
Note there is **no trailing space** after `<loom>`. Adding one pushes the model out of
|
| 77 |
+
distribution and measurably degrades output quality.
|
| 78 |
+
|
| 79 |
+
Offline (`<tools:off>`) it answers from its own small store or admits it does not know.
|
| 80 |
+
Online (`<tools:on>`) it may emit a search query and **end its turn immediately**:
|
| 81 |
+
|
| 82 |
+
```
|
| 83 |
+
<tools:on>
|
| 84 |
+
<tools:on><user> what year did the titanic sink
|
| 85 |
+
<loom> Not in my head, checking: <lookup>titanic sinking date</lookup><|endoftext|>
|
| 86 |
+
<result>The Titanic sank on 15 April 1912.</result>
|
| 87 |
+
<loom> April 1912 — ...
|
| 88 |
+
```
|
| 89 |
+
|
| 90 |
+
`<result>` blocks are **injected by your harness** after executing the search. The model
|
| 91 |
+
never generates them (verified: 0 self-generated results across all probes). Stop
|
| 92 |
+
generation at `<|endoftext|>` or `<user>`.
|
| 93 |
+
|
| 94 |
+
## Usage — transformers
|
| 95 |
+
|
| 96 |
+
```python
|
| 97 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
| 98 |
+
import torch
|
| 99 |
+
|
| 100 |
+
tok = AutoTokenizer.from_pretrained("textilelabs/Loom-Spark-1.8")
|
| 101 |
+
model = AutoModelForCausalLM.from_pretrained("textilelabs/Loom-Spark-1.8")
|
| 102 |
+
|
| 103 |
+
prompt = "<tools:off>\n<tools:off><user> who are you\n<loom>" # no trailing space
|
| 104 |
+
ids = tok(prompt, return_tensors="pt", add_special_tokens=False).input_ids
|
| 105 |
+
out = model.generate(ids, max_new_tokens=100, do_sample=True,
|
| 106 |
+
temperature=0.8, top_k=50, pad_token_id=tok.eos_token_id)
|
| 107 |
+
print(tok.decode(out[0][ids.shape[1]:], skip_special_tokens=False))
|
| 108 |
+
```
|
| 109 |
+
|
| 110 |
+
## Usage — Ollama
|
| 111 |
+
|
| 112 |
+
A ready-made `Modelfile` is in `ollama/`. It encodes the trained format and stop tokens —
|
| 113 |
+
**do not** use Ollama's default chat template, which is out of distribution for this model
|
| 114 |
+
and produces incoherent output.
|
| 115 |
+
|
| 116 |
+
```bash
|
| 117 |
+
ollama create loom-spark-1.8 -f ollama/Modelfile
|
| 118 |
+
ollama run loom-spark-1.8
|
| 119 |
+
```
|
| 120 |
+
|
| 121 |
+
## Usage — the agent harness
|
| 122 |
+
|
| 123 |
+
The `harness/` folder contains `loomspark-harness` (v0.2.1), which implements the full
|
| 124 |
+
lookup protocol with real web search, a CLI (`loom-chat`) and a small web UI (`loom-web`).
|
| 125 |
+
It auto-detects 1.5/1.8-format models and bans lookup tokens at the logits level when
|
| 126 |
+
tools are off.
|
| 127 |
+
|
| 128 |
+
```bash
|
| 129 |
+
pip install ./harness
|
| 130 |
+
loom-chat --model textilelabs/Loom-Spark-1.8
|
| 131 |
+
```
|
| 132 |
+
|
| 133 |
+
## Limitations — please read
|
| 134 |
+
|
| 135 |
+
Loom Spark 1.8 knows almost nothing, and that is the design. Specifically:
|
| 136 |
+
|
| 137 |
+
- **Do not use it for facts, medicine, law, finance,** or anything where being wrong
|
| 138 |
+
costs more than company.
|
| 139 |
+
- **It retrieves *adjacent* facts.** Asked the language spoken in Brazil it may answer
|
| 140 |
+
"Brasília"; asked Germany's currency it may answer "Berlin". Its fact-core is tiny and
|
| 141 |
+
it pattern-matches to a nearby entry.
|
| 142 |
+
- **Offline lookup leakage.** With tools off, the raw model still emits `<lookup>` tags on
|
| 143 |
+
roughly half of fact-style questions (16/30 measured; shipped v1.5 measures 18/30 on
|
| 144 |
+
the same test). The included harness suppresses this at the logits level, so harness
|
| 145 |
+
users never see it — but users driving the raw model through llama.cpp or Ollama will.
|
| 146 |
+
This is a curriculum characteristic of the whole Loom line, not specific to 1.8.
|
| 147 |
+
- **256-token context.** It has no long-term memory and only a short working window.
|
| 148 |
+
- It is a **base-style model**, not instruction-tuned or RLHF'd.
|
| 149 |
+
|
| 150 |
+
## Training
|
| 151 |
+
|
| 152 |
+
- Hardware: CPU-only Dell OptiPlex 9020, i5-4690, 4 cores, 16GB RAM, **no GPU**
|
| 153 |
+
- ~5,084 optimizer steps, batch 32 × 256 tokens, ~8 hours wall clock
|
| 154 |
+
- Corpus: 70MB / 14.9M tokens of procedurally generated curriculum, 100% owned
|
| 155 |
+
- AdamW (β 0.9/0.95), peak LR 6e-4, cosine decay, warmup 100, grad clip 1.0
|
| 156 |
+
- Dropout 0.1 resid/mlp, tied embeddings, pre-LN, GELU, learned positions
|
| 157 |
+
- Final val loss **0.3279**
|
| 158 |
+
|
| 159 |
+
The curriculum is procedurally generated and fully owned by Textile Labs — no scraped
|
| 160 |
+
web text, no third-party datasets.
|
| 161 |
+
|
| 162 |
+
## Files
|
| 163 |
+
|
| 164 |
+
```
|
| 165 |
+
config.json / model.safetensors transformers weights (GPT2LMHeadModel-compatible)
|
| 166 |
+
tokenizer.json / tokenizer_config.json 4096-token custom BPE
|
| 167 |
+
loom-spark-1.8-f32.gguf GGUF for llama.cpp / Ollama
|
| 168 |
+
ollama/Modelfile correct template + stop tokens
|
| 169 |
+
harness/ agent harness with web search (v0.2.1)
|
| 170 |
+
evaluation/ full acceptance logs, comparison vs v1.5, train log
|
| 171 |
+
```
|
| 172 |
+
|
| 173 |
+
## License
|
| 174 |
+
|
| 175 |
+
MIT. See `LICENSE`.
|
| 176 |
+
|
| 177 |
+
---
|
| 178 |
+
|
| 179 |
+
*Loom Spark 1.8 · Textile Labs · trained on a thirteen-year-old office computer.*
|
config.json
ADDED
|
@@ -0,0 +1,34 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"activation_function": "gelu_new",
|
| 3 |
+
"add_cross_attention": false,
|
| 4 |
+
"architectures": [
|
| 5 |
+
"GPT2LMHeadModel"
|
| 6 |
+
],
|
| 7 |
+
"attn_pdrop": 0.0,
|
| 8 |
+
"bos_token_id": 0,
|
| 9 |
+
"dtype": "float32",
|
| 10 |
+
"embd_pdrop": 0.0,
|
| 11 |
+
"eos_token_id": 0,
|
| 12 |
+
"initializer_range": 0.02,
|
| 13 |
+
"layer_norm_epsilon": 1e-05,
|
| 14 |
+
"model_type": "gpt2",
|
| 15 |
+
"n_embd": 448,
|
| 16 |
+
"n_head": 7,
|
| 17 |
+
"n_inner": null,
|
| 18 |
+
"n_layer": 7,
|
| 19 |
+
"n_positions": 256,
|
| 20 |
+
"pad_token_id": null,
|
| 21 |
+
"reorder_and_upcast_attn": false,
|
| 22 |
+
"resid_pdrop": 0.0,
|
| 23 |
+
"scale_attn_by_inverse_layer_idx": false,
|
| 24 |
+
"scale_attn_weights": true,
|
| 25 |
+
"summary_activation": null,
|
| 26 |
+
"summary_first_dropout": 0.1,
|
| 27 |
+
"summary_proj_to_labels": true,
|
| 28 |
+
"summary_type": "cls_index",
|
| 29 |
+
"summary_use_proj": true,
|
| 30 |
+
"tie_word_embeddings": true,
|
| 31 |
+
"transformers_version": "5.15.1",
|
| 32 |
+
"use_cache": true,
|
| 33 |
+
"vocab_size": 4096
|
| 34 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 0,
|
| 4 |
+
"eos_token_id": 0,
|
| 5 |
+
"output_attentions": false,
|
| 6 |
+
"output_hidden_states": false,
|
| 7 |
+
"transformers_version": "5.15.1",
|
| 8 |
+
"use_cache": true
|
| 9 |
+
}
|
tokenizer.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
tokenizer_config.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"backend": "tokenizers",
|
| 3 |
+
"bos_token": "<|endoftext|>",
|
| 4 |
+
"eos_token": "<|endoftext|>",
|
| 5 |
+
"model_max_length": 1000000000000000019884624838656,
|
| 6 |
+
"pad_token": "<|endoftext|>",
|
| 7 |
+
"tokenizer_class": "TokenizersBackend"
|
| 8 |
+
}
|