danyaljj

Yes — and CoreWeave’s 8× RTX PRO 6000 Blackwell Server Edition is actually a very interesting addition on price. It gives you 768 GB of total VRAM at a much lower hourly rate than HGX H100/H200. CoreWeave currently lists the 8-GPU high-memory node at $20/hr on-demand and about $11/hr spot. ([CoreWeave][1])

The catch is important: RTX PRO 6000 Blackwell is PCIe-based, not an 8-GPU NVSwitch fabric like HGX H100/H200/B200. NVIDIA’s reference architecture shows an 8-GPU RTX PRO server using PCIe switching, while HGX-class systems are designed around NVLink/NVSwitch. ([NVIDIA][2])

Here’s the updated shortlist:

Rank	Provider	GPU setup	Total VRAM	GPU fabric	Pricing model	$/hour	$/week	Best for
1	**CoreW

danyaljj

Yes — and **CoreWeave’s 8× RTX PRO 6000 Blackwell Server Edition** is actually a very interesting addition on price. It gives you **768 GB of total VRAM** at a much lower hourly rate than HGX H100/H200. CoreWeave currently lists the 8-GPU high-memory node at **$20/hr on-demand** and about **$11/hr spot**. ([CoreWeave][1])

The catch is important: **RTX PRO 6000 Blackwell is PCIe-based, not an 8-GPU NVSwitch fabric like HGX H100/H200/B200.** NVIDIA’s reference architecture shows an 8-GPU RTX PRO server using PCIe switching, while HGX-class systems are designed around NVLink/NVSwitch. ([NVIDIA][2])

Here’s the updated shortlist:

|   Rank | Provider      | GPU setup                      |   Total VRAM | GPU fabric            | Pricing model  |     $/hour |      $/week | Best for                                |
| -----: | ------------- | ------------------------------ | -----------: | --------------------- | -------------- | ---------: | ----------: | --------------------------------------- |
|  **1** | **CoreW

danyaljj

from transformers import pipeline

# Qwen/Qwen2.5-0.5B
# EleutherAI/pythia-410m
# HuggingFaceTB/SmolLM2-360M
# Qwen/Qwen2-1.5B
# Qwen/Qwen2.5-1.5B
gen = pipeline("text-generation", model="Qwen/Qwen2.5-1.5B")

prompt = "The best university in the world is" 

danyaljj

from transformers import pipeline

# Qwen/Qwen2.5-0.5B
# EleutherAI/pythia-410m
# HuggingFaceTB/SmolLM2-360M
# Qwen/Qwen2-1.5B
# Qwen/Qwen2.5-1.5B
gen = pipeline("text-generation", model="Qwen/Qwen2.5-1.5B")

prompt = "The best university in the world is" 

danyaljj

"""
Split messages JSON into one file per conversation (turn).

Usage:
    python analysis/1.split_conversations.py messages_202606221221.json [output_dir] [--since 2026-06-01]

Each file is named:  YYYY-MM-DDTHH-MM_<slug>.json
The trace payloads are fully deserialized — no double-JSON-encoded strings.
"""

danyaljj

"""
Split messages JSON into one file per conversation (turn).

Usage:
    python analysis/1.split_conversations.py messages_202606221221.json [output_dir] [--since 2026-06-01]

Each file is named:  YYYY-MM-DDTHH-MM_<slug>.json
The trace payloads are fully deserialized — no double-JSON-encoded strings.
"""

danyaljj

Option Explicit

Sub AddElements()
Dim shp As Shape

Dim i As Integer, n As Integer
n = ActivePresentation.Slides.Count
For i = 1 To n
    Dim s As Slide
    Set s = ActivePresentation.Slides(i)

danyaljj

import requests

PAPER_ID = "0a2c5dfa8b273c0130a0b59be7ce5dba575303be"
BASE_URL = "https://api.semanticscholar.org/graph/v1"
S2_API_KEY = "..."

HEADERS = {"x-api-key": S2_API_KEY}
PAPER_FIELDS = "title,abstract,year,authors,citationCount,referenceCount,externalIds,venue,publicationDate"
REF_FIELDS = "title,authors,year,citationCount,externalIds"

danyaljj

{
  "Health & Clinical Sciences": {
    "Dentistry": {
      "General Dentistry": [
        "Dental Research and COVID-19"
      ],
      "Oral Surgery": [
        "Dental Anxiety and Anesthesia Techniques",
        "Dental Implant Techniques and Outcomes",
        "Dental Radiography and Imaging",

danyaljj

Dataset / Platform	Scope (what it includes)	Approx. size	Bulk access	License / reuse for products & demos
Semantic Scholar Open Research Corpus (S2ORC)	Papers (metadata, abstracts, references); full text only for OA subset	~80M papers; ~8–12M full text	Static snapshots (≤2020) + AP