Text Generation
Transformers
Safetensors
English
pretrained-from-scratch
causal-lm
gpt
rope
swiglu
grouped-query-attention
Instructions to use Dikshan1234/LLMPretrain with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Dikshan1234/LLMPretrain with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="Dikshan1234/LLMPretrain")# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("Dikshan1234/LLMPretrain", device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use Dikshan1234/LLMPretrain with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "Dikshan1234/LLMPretrain" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Dikshan1234/LLMPretrain", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/Dikshan1234/LLMPretrain
- SGLang
How to use Dikshan1234/LLMPretrain 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 "Dikshan1234/LLMPretrain" \ --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": "Dikshan1234/LLMPretrain", "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 "Dikshan1234/LLMPretrain" \ --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": "Dikshan1234/LLMPretrain", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use Dikshan1234/LLMPretrain with Docker Model Runner:
docker model run hf.co/Dikshan1234/LLMPretrain
Update model card
Browse files
README.md
CHANGED
|
@@ -51,8 +51,8 @@ Parameter split: **41,943,040** embedding
|
|
| 51 |
|
| 52 |
| | |
|
| 53 |
|---|---|
|
| 54 |
-
| Tokens seen |
|
| 55 |
-
| Steps | 76,
|
| 56 |
| Schedule | warmup-stable-decay (WSD) |
|
| 57 |
| Optimizer | fused AdamW, betas (0.9, 0.95), wd 0.1 on 2D params only |
|
| 58 |
| Precision | bf16 autocast, fp32 master weights |
|
|
@@ -68,8 +68,8 @@ directly comparable.
|
|
| 68 |
| Metric | This model | GPT-2 124M |
|
| 69 |
|---|---|---|
|
| 70 |
| Parameters | 351,650,304 | 124,439,808 |
|
| 71 |
-
| Validation loss | 1.
|
| 72 |
-
| **Bits per byte** | **0.
|
| 73 |
| HellaSwag (0-shot, len-norm) | _pending_ | 29.55% |
|
| 74 |
|
| 75 |
|
|
@@ -113,48 +113,35 @@ and are short enough to read.
|
|
| 113 |
## Sample generations
|
| 114 |
|
| 115 |
```
|
| 116 |
-
===== step
|
| 117 |
|
| 118 |
--- [greedy] 'Once upon a time'
|
| 119 |
-
Once upon a time, the
|
| 120 |
|
| 121 |
--- [t=0.8] 'Once upon a time'
|
| 122 |
-
Once upon a time
|
| 123 |
-
|
| 124 |
-
|
| 125 |
-
As long as you have a good tree, there is value to the tree.
|
| 126 |
-
And so, we
|
| 127 |
|
| 128 |
--- [greedy] 'The capital of France is'
|
| 129 |
-
The capital of France is divided into three classes of shares, Class A shares, Class B shares and Class C shares. Class A shares are entitled to one vote per share
|
| 130 |
|
| 131 |
-
The Company
|
| 132 |
|
| 133 |
--- [t=0.8] 'The capital of France is'
|
| 134 |
-
The capital of France is
|
| 135 |
-
|
| 136 |
-
The Company’s registered address is located at “Hirac Plaza”, 10th floor, 12, rue de Rue de Rue de Grangville, 106000 Paris, France. The Company’s telephone number is +33 (0)1-33-83000-3450.
|
| 137 |
-
|
| 138 |
-
The Company’s common stock is issued in registered form. The registrar and transfer agent is Vstock Transfer, LLC
|
| 139 |
|
| 140 |
--- [greedy] 'In 1969, humans first'
|
| 141 |
-
In 1969, humans first began to use
|
| 142 |
|
| 143 |
--- [t=0.8] 'In 1969, humans first'
|
| 144 |
-
In 1969, humans first
|
| 145 |
-
The
|
| 146 |
-
In
|
|
|
|
| 147 |
|
| 148 |
--- [greedy] 'The three states of matter are'
|
| 149 |
-
The three states of matter are the
|
| 150 |
-
The state of matter is the state of being.
|
| 151 |
-
The state of matter is the state of being.
|
| 152 |
-
The state of matter is the state of being.
|
| 153 |
-
The state of matter is the state of being.
|
| 154 |
-
The state of matter is the state of being.
|
| 155 |
-
The state of matter is the state of being.
|
| 156 |
-
The state of matter is the state of being.
|
| 157 |
-
The state of matter is the state of being
|
| 158 |
```
|
| 159 |
|
| 160 |
## Limitations
|
|
|
|
| 51 |
|
| 52 |
| | |
|
| 53 |
|---|---|
|
| 54 |
+
| Tokens seen | 40,109,656,064 |
|
| 55 |
+
| Steps | 76,500 |
|
| 56 |
| Schedule | warmup-stable-decay (WSD) |
|
| 57 |
| Optimizer | fused AdamW, betas (0.9, 0.95), wd 0.1 on 2D params only |
|
| 58 |
| Precision | bf16 autocast, fp32 master weights |
|
|
|
|
| 68 |
| Metric | This model | GPT-2 124M |
|
| 69 |
|---|---|---|
|
| 70 |
| Parameters | 351,650,304 | 124,439,808 |
|
| 71 |
+
| Validation loss | 1.4946 | — |
|
| 72 |
+
| **Bits per byte** | **0.5065** | — |
|
| 73 |
| HellaSwag (0-shot, len-norm) | _pending_ | 29.55% |
|
| 74 |
|
| 75 |
|
|
|
|
| 113 |
## Sample generations
|
| 114 |
|
| 115 |
```
|
| 116 |
+
===== step 76500 =====
|
| 117 |
|
| 118 |
--- [greedy] 'Once upon a time'
|
| 119 |
+
Once upon a time, the Company’s principal source of liquidity was cash on hand. The Company’s principal source of liquidity is cash flow from operations. The Company’s principal source of liquidity is cash flow from operations. The Company’s principal source of liquidity is cash flow from operations. The Company’s principal source of liquidity is cash flow from operations. The Company’s principal source of liquidity is cash flow from operations. The Company’s principal source of liquidity is cash flow from operations. The Company’s principal source of liquidity is cash flow from operations. The Company’s principal source of liquidity
|
| 120 |
|
| 121 |
--- [t=0.8] 'Once upon a time'
|
| 122 |
+
Once upon a time I had been in possession of my body and my physical body had been restored to the body of a man, I never had any ill effects on others. I have received no injuries to the body or to any other body of men from that person, no hurt on the person, no injury to any thing, no hurt at any time.
|
| 123 |
+
And in the mean time there have been no deaths, no any injuries, no any deaths.
|
| 124 |
+
Now, if I had to die in another man, I would not have been in possession of my body, my physical body, and my physical body in
|
|
|
|
|
|
|
| 125 |
|
| 126 |
--- [greedy] 'The capital of France is'
|
| 127 |
+
The capital of France is divided into three classes of shares, Class A shares, Class B shares and Class C shares. Class A shares are entitled to one vote per share, Class B shares are entitled to ten votes per share, and Class C shares are entitled to one vote per share. The holders of Class A shares and Class B shares are entitled to share equally in the profits of the Company and dividends are declared and paid in accordance with the Company’s articles of association.
|
| 128 |
|
| 129 |
+
The Company has authorized and issued 1,000,000,000 shares of Class A shares, 1,000,000,000 shares
|
| 130 |
|
| 131 |
--- [t=0.8] 'The capital of France is'
|
| 132 |
+
The capital of France is regulated by the Financial Services Authority (FSA). The FSA is responsible for supervising the financial activity of banks in France. The FSA’s primary regulator of banks in France is the Autorité des marchés financiers du Marche (AMF), a public corporation, and the bank’s legal name is Banco Bilbao Vizcaya Argentaria. The Autorité des marchés financiers du Marche is responsible for supervising the financial activities of banks in France. The Autorité des marchés financiers du Marche is similar
|
|
|
|
|
|
|
|
|
|
|
|
|
| 133 |
|
| 134 |
--- [greedy] 'In 1969, humans first'
|
| 135 |
+
In 1969, humans first began to use the term “genotype” to describe the genetic makeup of an organism. In 1972, the term “genotype” was introduced to describe the genetic makeup of an organism. In 1973, the term “phenotype” was introduced to describe the genetic makeup of an organism. In 1976, the term “genotype” was introduced to describe the genetic makeup of an organism. In 1977, the term “genotype” was introduced to describe the genetic makeup of an organism. In 1978, the term “phenotype” was introduced to describe the genetic makeup of an organism. In 1979
|
| 136 |
|
| 137 |
--- [t=0.8] 'In 1969, humans first'
|
| 138 |
+
In 1969, humans first became attracted to firewood in North America, following a banshi in Japan.
|
| 139 |
+
The Japanese introduced the term “firewood” as a generic term in their language of Japan, and as a synonym for any wood that burns.
|
| 140 |
+
In the early 1990s, the United States adopted the term “forest and woodland ecosystem,” which was adopted in part to address the problems caused by the deforestation and degradation of forests.
|
| 141 |
+
The term “forest and woodland ecosystem” was used to refer to the rich natural ecosystems rich in biodiversity, including forests, wetlands, grasslands, shrublands, tund
|
| 142 |
|
| 143 |
--- [greedy] 'The three states of matter are'
|
| 144 |
+
The three states of matter are distinguished from each other primarily by their relative density. The density of a particular substance is the mass of that substance divided by the total mass of all the substances that are in that substance. For example, water has a density of 1.0 g/cm3, bu
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 145 |
```
|
| 146 |
|
| 147 |
## Limitations
|