Intel/Qwen3-235B-A22B-Instruct-2507-gguf-q2ks-mixed-AutoRound

Model Details

This model is a mixed gguf:q2ks of Qwen/Qwen3-235B-A22B-Instruct-2507 generated by intel/auto-round algorithm. Embedding layer and lm-head layer are fallback to 8 bits and non expert layers are fallback to 4 bits. Please refer to Section Generate the model for more details.

How To Use

LLamacpp Inference

./llama-cli -hf Intel/Qwen3-235B-A22B-Instruct-2507-gguf-q2ks-mixed-AutoRound:q2_k_s --conversation

"""
> Hi
Hello! How can I help you today? 😊

> code a flappy bird in Python
Sure! Here's a simple **Flappy Bird** game implemented in Python using the **Pygame** library.

---

### 🔧 Requirements:
Make sure you have **Pygame** installed:

```bash
pip install pygame
```

---

### 🐤 Flappy Bird in Python

```python
import pygame
import random
import sys

# Initialize Pygame
pygame.init()

# Screen dimensions
WIDTH, HEIGHT = 400, 600
screen = pygame.display.set_mode((WIDTH, HEIGHT))
pygame.display.set_caption("Flappy Bird")

# Colors
WHITE = (255, 255, 255)
BLACK = (0, 0, 0)
GREEN = (0, 128, 0)
SKY_BLUE = (135, 206, 235)

# Game settings
GRAVITY = 0.5
FLAP_STRENGTH = -10
PIPE_SPEED = 3
PIPE_GAP = 150
PIPE_FREQUENCY = 1500  # milliseconds

# Bird class
class Bird:
    def __init__(self):
        self.x = 50
        self.y = HEIGHT // 2
        self.velocity = 0
        self.width = 20
        self.height = 20

    def flap(self):
        self.velocity = FLAP_STRENGTH

    def update(self):
        self.velocity += GRAVITY
        self.y += self.velocity
        if self.y < 0:
            self.y = 0
        if self.y > HEIGHT:
            self.y = HEIGHT

    def draw(self):
        pygame.draw.circle(screen, BLACK, (self.x, int(self.y)), self.width // 2)

# Pipe class
class Pipe:
    def __init__(self):
        self.x = WIDTH
        self.top = random.randint(50, HEIGHT - PIPE_GAP - 50)
        self.bottom = self.top + PIPE_GAP
        self.width = 50
        self.passed = False

    def update(self):
        self.x -= PIPE_SPEED
        return self.x > -self.width

    def draw(self):
        pygame.draw.rect(screen, GREEN, (self.x, 0, self.width, self.top))
        pygame.draw.rect(screen, GREEN, (self.x, self.bottom, self.width, HEIGHT - self.bottom))

    def collide(self, bird):
        if bird.x + bird.width // 2 > self.x and bird.x - bird.width // 2 < self.x + self.width:
            if bird.y - bird.height // 2 < self.top or bird.y + bird.height // 2 > self.bottom:
                return True
        return False

# Main game loop
def game():
    bird = Bird()
    pipes = []
    score = 0
    clock = pygame.time.Clock()
    last_pipe = pygame.time.get_ticks()

    running = True
    game_active = True

    font = pygame.font.SysFont(None, 55)

    while running:
        screen.fill(SKY_BLUE)
        dt = clock.tick(60)
        current_time = pygame.time.get_ticks()

        # Event handling
        for event in pygame.event.get():
            if event.type == pygame.QUIT:
                pygame.quit()
                sys.exit()
            if event.type == pygame.KEYDOWN:
                if event.key == pygame.K_SPACE and game_active:
                    bird.flap()
                if event.key == pygame.K_r and not game_active:
                    game()

        if game_active:
            # Update bird
            bird.update()

            # Generate new pipes
            if current_time - last_pipe > PIPE_FREQUENCY:
                pipes.append(Pipe())
                last_pipe = current_time

            # Update and draw pipes
            pipes = [p for p in pipes if p.update()]
            for pipe in pipes:
                pipe.draw()
                if pipe.collide(bird):
                    game_active = False

                # Score logic
                if pipe.x + pipe.width < bird.x and not pipe.passed:
                    score += 1
                    pipe.passed = True

            # Draw bird
            bird.draw()

            # Check if bird hits the ground or flies too high
            if bird.y >= HEIGHT or bird.y <= 0:
                game_active = False

            # Display score
            score_text = font.render(f"Score: {score}", True, BLACK)
            screen.blit(score_text, (10, 10))

        else:
            # Game over screen
            game_over_text = font.render("Game Over!", True, BLACK)
            restart_text = font.render("Press 'R' to Restart", True, BLACK)
            screen.blit(game_over_text, (WIDTH // 2 - 90, HEIGHT // 2 - 50))
            screen.blit(restart_text, (WIDTH // 2 - 150, HEIGHT // 2 + 10))

        pygame.display.update()

# Start the game
game()
```

---

### 🎮 How to Play:
- Press **Spacebar** to flap the bird upward.
- Avoid hitting the pipes or the ground.
- When you die, press **'R'** to restart.

---

### ✅ Features:
- Gravity and flap mechanics.
- Random pipe generation.
- Collision detection.
- Score tracking.
- Simple graphics using shapes.

---

Let me know if you'd like to add images, sounds, or improve the visuals! 🐦💨


"""

Generate the model

auto-round>0.5.1

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
from auto_round import AutoRound

model_name = "Qwen/Qwen3-235B-A22B-Instruct-2507" 

model = AutoModelForCausalLM.from_pretrained(model_name,
                                             device_map="cpu", torch_dtype="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)
layer_config = {}
for n, m in model.named_modules():
    if n == "lm_head" or isinstance(m,torch.nn.Embedding):
        layer_config[n] = {"bits": 8}
    elif isinstance(m, torch.nn.Linear) and (not "expert" in n or "shared_experts" in n) and n != "lm_head":
        layer_config[n] = {"bits": 4}

autoround = AutoRound(model, tokenizer, iters=0, layer_config=layer_config, nsamples=512)
autoround.quantize_and_save("/models/Qwen3-235B-A22B-Instruct-2507-q2ks", format="gguf:q2_k_s")

Ethical Considerations and Limitations

The model can produce factually incorrect output, and should not be relied on to produce factually accurate information. Because of the limitations of the pretrained model and the finetuning datasets, it is possible that this model could generate lewd, biased or otherwise offensive outputs.

Therefore, before deploying any applications of the model, developers should perform safety testing.

Caveats and Recommendations

Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model.

Here are a couple of useful links to learn more about Intel's AI software:

• Intel Neural Compressor link

Disclaimer

The license on this model does not constitute legal advice. We are not responsible for the actions of third parties who use this model. Please consult an attorney before using this model for commercial purposes.

Cite

@article{cheng2023optimize, title={Optimize weight rounding via signed gradient descent for the quantization of llms}, author={Cheng, Wenhua and Zhang, Weiwei and Shen, Haihao and Cai, Yiyang and He, Xin and Lv, Kaokao and Liu, Yi}, journal={arXiv preprint arXiv:2309.05516}, year={2023} }

arxiv github