Instructions to use plasmova/nova-v3 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use plasmova/nova-v3 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="plasmova/nova-v3", trust_remote_code=True)# pip install -U transformers accelerate # Load model directly from transformers import AutoModelForCausalLM model = AutoModelForCausalLM.from_pretrained("plasmova/nova-v3", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- llama.cpp
How to use plasmova/nova-v3 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 plasmova/nova-v3:F16 # Run inference directly in the terminal: llama cli -hf plasmova/nova-v3:F16
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama serve -hf plasmova/nova-v3:F16 # Run inference directly in the terminal: llama cli -hf plasmova/nova-v3:F16
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 plasmova/nova-v3:F16 # Run inference directly in the terminal: ./build/bin/llama cli -hf plasmova/nova-v3:F16
Use Docker
docker model run hf.co/plasmova/nova-v3:F16
- Llama new
- LM Studio
- Jan
- vLLM
How to use plasmova/nova-v3 with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "plasmova/nova-v3" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "plasmova/nova-v3", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/plasmova/nova-v3:F16
- SGLang
How to use plasmova/nova-v3 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 "plasmova/nova-v3" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "plasmova/nova-v3", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'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 "plasmova/nova-v3" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "plasmova/nova-v3", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Ollama
How to use plasmova/nova-v3 with Ollama:
ollama run hf.co/plasmova/nova-v3:F16
- Unsloth Desktop
- Docker Model Runner
How to use plasmova/nova-v3 with Docker Model Runner:
docker model run hf.co/plasmova/nova-v3:F16
- Lemonade
How to use plasmova/nova-v3 with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull plasmova/nova-v3:F16
Run and chat with the model
lemonade run user.nova-v3-F16
List all available models
lemonade list
- Atomic Chat
Upload Nova v3 Transformers model
Browse files- README.md +55 -0
- config.json +24 -0
- configuration_nova.py +30 -0
- generation_config.json +6 -0
- model.safetensors +3 -0
- modeling_nova.py +164 -0
- special_tokens_map.json +4 -0
- tokenizer.json +0 -0
- tokenizer_config.json +98 -0
README.md
ADDED
|
@@ -0,0 +1,55 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: transformers
|
| 3 |
+
tags:
|
| 4 |
+
- nova
|
| 5 |
+
- text-generation
|
| 6 |
+
- custom-code
|
| 7 |
+
---
|
| 8 |
+
|
| 9 |
+
# Nova v3
|
| 10 |
+
|
| 11 |
+
Nova v3 is a decoder-only causal language model exported from the native Nova v3 checkpoint. This repository includes the model weights, tokenizer, and the custom Transformers configuration and modeling code needed to load it.
|
| 12 |
+
|
| 13 |
+
## Model details
|
| 14 |
+
|
| 15 |
+
- Vocabulary size: 32,768
|
| 16 |
+
- Hidden size: 640
|
| 17 |
+
- Transformer layers: 26
|
| 18 |
+
- Attention heads: 10 query heads and 5 key/value heads
|
| 19 |
+
- Maximum context: 2,048 tokens
|
| 20 |
+
- Checkpoint step: 3,500
|
| 21 |
+
|
| 22 |
+
The exported model does not use a KV cache. Generation recomputes the full prefix at each step, so it is slower than cached Transformers models.
|
| 23 |
+
|
| 24 |
+
## Load with Transformers
|
| 25 |
+
|
| 26 |
+
The model uses custom code in this repository. Review that code before enabling `trust_remote_code`.
|
| 27 |
+
|
| 28 |
+
```python
|
| 29 |
+
import torch
|
| 30 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
| 31 |
+
|
| 32 |
+
model_id = "YOUR_USERNAME/nova-v3"
|
| 33 |
+
tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
|
| 34 |
+
model = AutoModelForCausalLM.from_pretrained(
|
| 35 |
+
model_id,
|
| 36 |
+
trust_remote_code=True,
|
| 37 |
+
dtype=torch.float32,
|
| 38 |
+
)
|
| 39 |
+
|
| 40 |
+
inputs = tokenizer("Hello!", return_tensors="pt")
|
| 41 |
+
output = model.generate(
|
| 42 |
+
**inputs,
|
| 43 |
+
max_new_tokens=64,
|
| 44 |
+
do_sample=False,
|
| 45 |
+
)
|
| 46 |
+
print(tokenizer.decode(output[0], skip_special_tokens=True))
|
| 47 |
+
```
|
| 48 |
+
|
| 49 |
+
## Evaluation
|
| 50 |
+
|
| 51 |
+
The local v3 checkpoint scored 37.6% normalized accuracy on a zero-shot 500-example sample of HellaSwag validation. This is a limited local benchmark run, not a full validation-set result.
|
| 52 |
+
|
| 53 |
+
## License and training data
|
| 54 |
+
|
| 55 |
+
No license or training-data statement was included with the source checkpoint. These should be added by the model publisher before public distribution.
|
config.json
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"NovaForCausalLM"
|
| 4 |
+
],
|
| 5 |
+
"dtype": "float32",
|
| 6 |
+
"eos_token_id": 10,
|
| 7 |
+
"head_dim": 64,
|
| 8 |
+
"hidden_size": 640,
|
| 9 |
+
"intermediate_size": 1920,
|
| 10 |
+
"max_position_embeddings": 2048,
|
| 11 |
+
"model_type": "nova",
|
| 12 |
+
"num_attention_heads": 10,
|
| 13 |
+
"num_hidden_layers": 26,
|
| 14 |
+
"num_key_value_heads": 5,
|
| 15 |
+
"pad_token_id": 10,
|
| 16 |
+
"rope_theta": 100000.0,
|
| 17 |
+
"transformers_version": "4.57.5",
|
| 18 |
+
"use_cache": false,
|
| 19 |
+
"vocab_size": 32768,
|
| 20 |
+
"auto_map": {
|
| 21 |
+
"AutoConfig": "configuration_nova.NovaConfig",
|
| 22 |
+
"AutoModelForCausalLM": "modeling_nova.NovaForCausalLM"
|
| 23 |
+
}
|
| 24 |
+
}
|
configuration_nova.py
ADDED
|
@@ -0,0 +1,30 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
from transformers import PretrainedConfig
|
| 2 |
+
|
| 3 |
+
|
| 4 |
+
class NovaConfig(PretrainedConfig):
|
| 5 |
+
model_type = "nova"
|
| 6 |
+
|
| 7 |
+
def __init__(
|
| 8 |
+
self,
|
| 9 |
+
vocab_size=32768,
|
| 10 |
+
hidden_size=640,
|
| 11 |
+
num_hidden_layers=26,
|
| 12 |
+
num_attention_heads=10,
|
| 13 |
+
num_key_value_heads=5,
|
| 14 |
+
head_dim=64,
|
| 15 |
+
intermediate_size=1920,
|
| 16 |
+
max_position_embeddings=2048,
|
| 17 |
+
rope_theta=100000.0,
|
| 18 |
+
**kwargs,
|
| 19 |
+
):
|
| 20 |
+
super().__init__(tie_word_embeddings=True, **kwargs)
|
| 21 |
+
self.vocab_size = vocab_size
|
| 22 |
+
self.hidden_size = hidden_size
|
| 23 |
+
self.num_hidden_layers = num_hidden_layers
|
| 24 |
+
self.num_attention_heads = num_attention_heads
|
| 25 |
+
self.num_key_value_heads = num_key_value_heads
|
| 26 |
+
self.head_dim = head_dim
|
| 27 |
+
self.intermediate_size = intermediate_size
|
| 28 |
+
self.max_position_embeddings = max_position_embeddings
|
| 29 |
+
self.rope_theta = rope_theta
|
| 30 |
+
|
generation_config.json
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"eos_token_id": 10,
|
| 4 |
+
"pad_token_id": 10,
|
| 5 |
+
"transformers_version": "4.57.5"
|
| 6 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6050e8ddaaf5e9a3aa4d3e1ab8fa8d5d04908f1a5acd9cd210f8634ef503143c
|
| 3 |
+
size 595243360
|
modeling_nova.py
ADDED
|
@@ -0,0 +1,164 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import torch
|
| 2 |
+
import torch.nn as nn
|
| 3 |
+
import torch.nn.functional as F
|
| 4 |
+
from transformers import GenerationMixin, PreTrainedModel
|
| 5 |
+
from transformers.modeling_outputs import CausalLMOutput
|
| 6 |
+
|
| 7 |
+
try:
|
| 8 |
+
from .configuration_nova import NovaConfig
|
| 9 |
+
except ImportError:
|
| 10 |
+
from configuration_nova import NovaConfig
|
| 11 |
+
|
| 12 |
+
|
| 13 |
+
def _rope(x, cos, sin):
|
| 14 |
+
a, b = x.chunk(2, dim=-1)
|
| 15 |
+
return torch.cat((a * cos - b * sin, b * cos + a * sin), dim=-1)
|
| 16 |
+
|
| 17 |
+
|
| 18 |
+
class NovaRMSNorm(nn.Module):
|
| 19 |
+
def __init__(self, dim):
|
| 20 |
+
super().__init__()
|
| 21 |
+
self.w = nn.Parameter(torch.ones(dim))
|
| 22 |
+
|
| 23 |
+
def forward(self, x):
|
| 24 |
+
return F.rms_norm(x, (x.size(-1),), self.w, 1e-6)
|
| 25 |
+
|
| 26 |
+
|
| 27 |
+
class NovaBlock(nn.Module):
|
| 28 |
+
def __init__(self, config):
|
| 29 |
+
super().__init__()
|
| 30 |
+
d, h, kv, hd, ff = (
|
| 31 |
+
config.hidden_size,
|
| 32 |
+
config.num_attention_heads,
|
| 33 |
+
config.num_key_value_heads,
|
| 34 |
+
config.head_dim,
|
| 35 |
+
config.intermediate_size,
|
| 36 |
+
)
|
| 37 |
+
self.h, self.kv, self.hd = h, kv, hd
|
| 38 |
+
self.n1, self.n2 = NovaRMSNorm(d), NovaRMSNorm(d)
|
| 39 |
+
self.qn, self.kn = NovaRMSNorm(hd), NovaRMSNorm(hd)
|
| 40 |
+
self.wq = nn.Linear(d, h * hd, bias=False)
|
| 41 |
+
self.wk = nn.Linear(d, kv * hd, bias=False)
|
| 42 |
+
self.wv = nn.Linear(d, kv * hd, bias=False)
|
| 43 |
+
self.wo = nn.Linear(h * hd, d, bias=False)
|
| 44 |
+
self.gate = nn.Linear(d, ff, bias=False)
|
| 45 |
+
self.up = nn.Linear(d, ff, bias=False)
|
| 46 |
+
self.down = nn.Linear(ff, d, bias=False)
|
| 47 |
+
|
| 48 |
+
def forward(self, x, cos, sin, attention_mask):
|
| 49 |
+
batch, length, _ = x.shape
|
| 50 |
+
h = self.n1(x)
|
| 51 |
+
q = self.wq(h).view(batch, length, self.h, self.hd)
|
| 52 |
+
k = self.wk(h).view(batch, length, self.kv, self.hd)
|
| 53 |
+
v = self.wv(h).view(batch, length, self.kv, self.hd)
|
| 54 |
+
q = _rope(self.qn(q), cos, sin).to(v.dtype).transpose(1, 2)
|
| 55 |
+
k = _rope(self.kn(k), cos, sin).to(v.dtype).transpose(1, 2)
|
| 56 |
+
v = v.transpose(1, 2)
|
| 57 |
+
repeat = self.h // self.kv
|
| 58 |
+
k = k.repeat_interleave(repeat, dim=1)
|
| 59 |
+
v = v.repeat_interleave(repeat, dim=1)
|
| 60 |
+
causal = torch.ones(length, length, dtype=torch.bool, device=x.device).tril()
|
| 61 |
+
mask = causal[None, None, :, :]
|
| 62 |
+
if attention_mask is not None:
|
| 63 |
+
mask = mask & attention_mask[:, None, None, :].to(torch.bool)
|
| 64 |
+
out = F.scaled_dot_product_attention(q, k, v, attn_mask=mask)
|
| 65 |
+
x = x + self.wo(out.transpose(1, 2).reshape(batch, length, -1))
|
| 66 |
+
h = self.n2(x)
|
| 67 |
+
return x + self.down(F.silu(self.gate(h)) * self.up(h))
|
| 68 |
+
|
| 69 |
+
|
| 70 |
+
class NovaModel(PreTrainedModel):
|
| 71 |
+
config_class = NovaConfig
|
| 72 |
+
base_model_prefix = "model"
|
| 73 |
+
_no_split_modules = ["NovaBlock"]
|
| 74 |
+
|
| 75 |
+
def __init__(self, config):
|
| 76 |
+
super().__init__(config)
|
| 77 |
+
self.emb = nn.Embedding(config.vocab_size, config.hidden_size)
|
| 78 |
+
self.blocks = nn.ModuleList(NovaBlock(config) for _ in range(config.num_hidden_layers))
|
| 79 |
+
self.norm = NovaRMSNorm(config.hidden_size)
|
| 80 |
+
frequencies = torch.outer(
|
| 81 |
+
torch.arange(config.max_position_embeddings, dtype=torch.float32),
|
| 82 |
+
1 / config.rope_theta ** (
|
| 83 |
+
torch.arange(0, config.head_dim, 2, dtype=torch.float32) / config.head_dim
|
| 84 |
+
),
|
| 85 |
+
)
|
| 86 |
+
self.register_buffer("cos", frequencies.cos(), persistent=False)
|
| 87 |
+
self.register_buffer("sin", frequencies.sin(), persistent=False)
|
| 88 |
+
self.post_init()
|
| 89 |
+
|
| 90 |
+
def get_input_embeddings(self):
|
| 91 |
+
return self.emb
|
| 92 |
+
|
| 93 |
+
def set_input_embeddings(self, value):
|
| 94 |
+
self.emb = value
|
| 95 |
+
|
| 96 |
+
def forward(
|
| 97 |
+
self,
|
| 98 |
+
input_ids=None,
|
| 99 |
+
attention_mask=None,
|
| 100 |
+
position_ids=None,
|
| 101 |
+
inputs_embeds=None,
|
| 102 |
+
**kwargs,
|
| 103 |
+
):
|
| 104 |
+
if inputs_embeds is None:
|
| 105 |
+
inputs_embeds = self.emb(input_ids)
|
| 106 |
+
batch, length, _ = inputs_embeds.shape
|
| 107 |
+
if position_ids is None:
|
| 108 |
+
position_ids = torch.arange(length, device=inputs_embeds.device).unsqueeze(0).expand(batch, -1)
|
| 109 |
+
cos = self.cos[position_ids].unsqueeze(-2)
|
| 110 |
+
sin = self.sin[position_ids].unsqueeze(-2)
|
| 111 |
+
hidden = inputs_embeds
|
| 112 |
+
for block in self.blocks:
|
| 113 |
+
hidden = block(hidden, cos, sin, attention_mask)
|
| 114 |
+
return self.norm(hidden)
|
| 115 |
+
|
| 116 |
+
|
| 117 |
+
class NovaForCausalLM(PreTrainedModel, GenerationMixin):
|
| 118 |
+
config_class = NovaConfig
|
| 119 |
+
base_model_prefix = "model"
|
| 120 |
+
_tied_weights_keys = []
|
| 121 |
+
|
| 122 |
+
def __init__(self, config):
|
| 123 |
+
super().__init__(config)
|
| 124 |
+
self.model = NovaModel(config)
|
| 125 |
+
self.post_init()
|
| 126 |
+
|
| 127 |
+
def get_input_embeddings(self):
|
| 128 |
+
return self.model.emb
|
| 129 |
+
|
| 130 |
+
def set_input_embeddings(self, value):
|
| 131 |
+
self.model.set_input_embeddings(value)
|
| 132 |
+
|
| 133 |
+
def get_output_embeddings(self):
|
| 134 |
+
return None
|
| 135 |
+
|
| 136 |
+
def prepare_inputs_for_generation(self, input_ids, attention_mask=None, **kwargs):
|
| 137 |
+
# This model has no KV cache; each generation step recomputes the full prefix.
|
| 138 |
+
return {"input_ids": input_ids, "attention_mask": attention_mask}
|
| 139 |
+
|
| 140 |
+
def forward(
|
| 141 |
+
self,
|
| 142 |
+
input_ids=None,
|
| 143 |
+
attention_mask=None,
|
| 144 |
+
position_ids=None,
|
| 145 |
+
inputs_embeds=None,
|
| 146 |
+
labels=None,
|
| 147 |
+
**kwargs,
|
| 148 |
+
):
|
| 149 |
+
hidden = self.model(
|
| 150 |
+
input_ids=input_ids,
|
| 151 |
+
attention_mask=attention_mask,
|
| 152 |
+
position_ids=position_ids,
|
| 153 |
+
inputs_embeds=inputs_embeds,
|
| 154 |
+
)
|
| 155 |
+
logits = F.linear(hidden, self.model.emb.weight)
|
| 156 |
+
loss = None
|
| 157 |
+
if labels is not None:
|
| 158 |
+
loss = F.cross_entropy(
|
| 159 |
+
logits[..., :-1, :].contiguous().view(-1, logits.size(-1)),
|
| 160 |
+
labels[..., 1:].contiguous().view(-1),
|
| 161 |
+
ignore_index=-100,
|
| 162 |
+
)
|
| 163 |
+
return CausalLMOutput(loss=loss, logits=logits)
|
| 164 |
+
|
special_tokens_map.json
ADDED
|
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"eos_token": "<|endoftext|>",
|
| 3 |
+
"pad_token": "<|endoftext|>"
|
| 4 |
+
}
|
tokenizer.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
tokenizer_config.json
ADDED
|
@@ -0,0 +1,98 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"added_tokens_decoder": {
|
| 3 |
+
"0": {
|
| 4 |
+
"content": "<|start_of_line|>",
|
| 5 |
+
"lstrip": false,
|
| 6 |
+
"normalized": false,
|
| 7 |
+
"rstrip": false,
|
| 8 |
+
"single_word": false,
|
| 9 |
+
"special": true
|
| 10 |
+
},
|
| 11 |
+
"1": {
|
| 12 |
+
"content": "<|end_of_line|>",
|
| 13 |
+
"lstrip": false,
|
| 14 |
+
"normalized": false,
|
| 15 |
+
"rstrip": false,
|
| 16 |
+
"single_word": false,
|
| 17 |
+
"special": true
|
| 18 |
+
},
|
| 19 |
+
"2": {
|
| 20 |
+
"content": "<|system|>",
|
| 21 |
+
"lstrip": false,
|
| 22 |
+
"normalized": false,
|
| 23 |
+
"rstrip": false,
|
| 24 |
+
"single_word": false,
|
| 25 |
+
"special": true
|
| 26 |
+
},
|
| 27 |
+
"3": {
|
| 28 |
+
"content": "<|/system|>",
|
| 29 |
+
"lstrip": false,
|
| 30 |
+
"normalized": false,
|
| 31 |
+
"rstrip": false,
|
| 32 |
+
"single_word": false,
|
| 33 |
+
"special": true
|
| 34 |
+
},
|
| 35 |
+
"4": {
|
| 36 |
+
"content": "<|user|>",
|
| 37 |
+
"lstrip": false,
|
| 38 |
+
"normalized": false,
|
| 39 |
+
"rstrip": false,
|
| 40 |
+
"single_word": false,
|
| 41 |
+
"special": true
|
| 42 |
+
},
|
| 43 |
+
"5": {
|
| 44 |
+
"content": "<|/user|>",
|
| 45 |
+
"lstrip": false,
|
| 46 |
+
"normalized": false,
|
| 47 |
+
"rstrip": false,
|
| 48 |
+
"single_word": false,
|
| 49 |
+
"special": true
|
| 50 |
+
},
|
| 51 |
+
"6": {
|
| 52 |
+
"content": "<|assistant|>",
|
| 53 |
+
"lstrip": false,
|
| 54 |
+
"normalized": false,
|
| 55 |
+
"rstrip": false,
|
| 56 |
+
"single_word": false,
|
| 57 |
+
"special": true
|
| 58 |
+
},
|
| 59 |
+
"7": {
|
| 60 |
+
"content": "<|/assistant|>",
|
| 61 |
+
"lstrip": false,
|
| 62 |
+
"normalized": false,
|
| 63 |
+
"rstrip": false,
|
| 64 |
+
"single_word": false,
|
| 65 |
+
"special": true
|
| 66 |
+
},
|
| 67 |
+
"8": {
|
| 68 |
+
"content": "<|think|>",
|
| 69 |
+
"lstrip": false,
|
| 70 |
+
"normalized": false,
|
| 71 |
+
"rstrip": false,
|
| 72 |
+
"single_word": false,
|
| 73 |
+
"special": true
|
| 74 |
+
},
|
| 75 |
+
"9": {
|
| 76 |
+
"content": "<|/think|>",
|
| 77 |
+
"lstrip": false,
|
| 78 |
+
"normalized": false,
|
| 79 |
+
"rstrip": false,
|
| 80 |
+
"single_word": false,
|
| 81 |
+
"special": true
|
| 82 |
+
},
|
| 83 |
+
"10": {
|
| 84 |
+
"content": "<|endoftext|>",
|
| 85 |
+
"lstrip": false,
|
| 86 |
+
"normalized": false,
|
| 87 |
+
"rstrip": false,
|
| 88 |
+
"single_word": false,
|
| 89 |
+
"special": true
|
| 90 |
+
}
|
| 91 |
+
},
|
| 92 |
+
"clean_up_tokenization_spaces": false,
|
| 93 |
+
"eos_token": "<|endoftext|>",
|
| 94 |
+
"extra_special_tokens": {},
|
| 95 |
+
"model_max_length": 1000000000000000019884624838656,
|
| 96 |
+
"pad_token": "<|endoftext|>",
|
| 97 |
+
"tokenizer_class": "PreTrainedTokenizerFast"
|
| 98 |
+
}
|